AI platforms are shipping major capability updates every week. Your brain needs 66 days to form a new habit. Here's the framework that closes the gap, and the four traits that separate those who compound advantages from those who fall behind.

“Which AI tool should I learn?”

Wrong question.

You finally figure out a workflow. It clicks. You feel competent for maybe 72 hours. Then the notification drops: “What’s New in [insert platform].” And the tool you just learned? It’s different now. The feature you mastered got replaced, merged, or buried under three new ones you’ve never seen.

So you start over. Again.

This keeps happening because we’ve been framing the problem wrong. The question was never “which tool.” The tool doesn’t matter. By the time you’ve built muscle memory around it, the platform has shipped two updates that changed how it works. The real question is: how do you build the capacity to adapt continuously when the tools never stop changing?

That’s what this piece answers.

The Practitioner’s View

I manage technology across 120+ franchise locations, 2,500+ employees, in 20+ states. When I say “iteration fatigue,” I’m not pulling from a whitepaper. I’m watching it happen in real time across a distributed workforce where “one more thing to learn” hits different when you’re already running at capacity.

So I went looking for the data to confirm what I was seeing on the ground. What I found was worse than I expected.

Between November 17 and December 11, 2025, four frontier AI companies released flagship models in a 24-day window. Grok 4.1, Gemini 3, Claude Opus 4.5, GPT-5.2. Boom, boom, boom, boom. In early 2026, OpenAI shipped GPT-5.3 and GPT-5.4 two days apart, with no public explanation. METR’s research shows AI capability doubling times have compressed from roughly 7 months to under 3 months when you isolate the most recent models. Epoch AI identified a structural breakpoint around April 2024 where the rate of frontier improvement nearly doubled.

Mary Meeker used the word “unprecedented” on 51 pages of a 340-page report. Fifty-one times. When Mary Meeker runs out of adjectives, pay attention.

What You’ll Walk Away With

We’re covering five things:

1. See the acceleration quantified. Actual data on how fast AI platforms are iterating and why it feels like drinking from a fire hose.

2. Understand why your brain fights it. The behavioral science behind change resistance (it’s not a character flaw, it’s biology).

3. Learn the 66-Day Problem. The structural mismatch between habit formation and AI release cadences that nobody is talking about.

4. Get the A.L.I.C. framework. Four traits that separate those who compound advantages from those who fall behind, with clear ownership: two on employees, two on leadership.

5. Walk away with a starting point. What to do this week, not this quarter.

The Big Reframe

The competitive moat is no longer what you know. It’s how fast you can learn, unlearn, and relearn.

That sentence deserves a second read. Because everything most organizations are doing right now is built on the opposite assumption:

Read that last row again. If your strategy is “pick the right tool,” you’ve already lost the thread. The right tool today is the wrong tool in six weeks. The muscle to adapt continuously is the only durable advantage.

And that muscle has a name.

A.L.I.C.: The Four Survival Traits

Four traits. Two owned by employees. Two owned by leadership. All four non-negotiable.

1. Adaptability (Employee-Owned)

The willingness to adjust your behavior, workflows, and mental models without waiting for permission. Not a personality trait. A strategic competency. The World Economic Forum ranks resilience and adaptability as the #2 most important skill globally, right behind analytical thinking (which 70% of employers call essential). Gallup’s 2025 data shows teams with high adaptability are 36% more productive and 32% more engaged.

The employee who treats each platform iteration as a compounding advantage (not a disruption) becomes irreplaceable. The one who waits for the dust to settle? The dust isn’t settling.

2. Learning (Leadership-Owned)

The organizational commitment to build continuous learning infrastructure. Not a quarterly training event. A weekly operating rhythm. 39% of existing worker skills will transform or become outdated by 2030, according to the WEF. Organizations that focus on cultural change see 5.3x better transformation outcomes than those focused on technology alone (McKinsey).

The critical insight most leaders miss: people follow what you do with AI, not what you say about it. If leadership isn’t visibly learning and adapting, nobody else will either. You go first.

3. Ideation (Employee-Owned)

The capacity for original, associational thinking. Connecting unrelated concepts in ways AI cannot. Wharton research found something fascinating: AI improves the quality of individual ideas but reduces the diversity of ideas across groups. It makes us individually better but collectively more similar.

That’s the trap. And it’s also the opportunity. The employee who brings novel combinations of experience, domain knowledge, and pattern recognition to AI-assisted workflows creates value that no model can replicate. AI can’t do your lived experience. It can’t improvise. It can’t add a knowledge domain on the fly because something reminded it of a conversation from three years ago. Feed that advantage. Read outside your lane. Talk to people who think nothing like you.

4. Change (Leadership-Owned)

The organizational capability to manage continuous change without burning out the workforce. Seventy percent of transformations fail. Not 30%. Not half. Seventy percent (McKinsey, Gartner, Bain all confirm this range). And the vast majority fail on culture, not technology.

Only 42% of burned-out employees tell their manager. Among those who do, 42% say their manager takes no action. That means the signal is invisible AND the response system is broken. In the Age of Iteration, change management isn’t a project phase. It’s a permanent operating discipline, owned by leadership, measured like any other business function.

The 66-Day Problem

Now here’s why A.L.I.C. isn’t optional.

University College London’s habit formation research (Lally et al.) found it takes a median of 66 days to build automaticity around a new behavior. Range: 18 to 254 days. Complex behavioral changes, like adopting new AI workflows, fall at the upper end of that range.

AI platforms ship major updates every 2 to 4 weeks. Four flagship models in 24 days. Two GPT versions two days apart. Capability doubling time under 3 months and compressing.

Your brain needs 66 days. The platform ships every 14. That’s the whole problem.

You’re being asked to adapt to the next iteration before you’ve automated the previous one. You’re perpetually in the learning curve. The question isn’t whether that’s uncomfortable. It is. The question is whether you treat that discomfort as a sign you’re behind or as the price of staying ahead.

The burnout data confirms the toll. Employee burnout hit 66% in 2025 (Modern Health/Forbes). Deloitte’s 2025 Workforce Intelligence Report identifies cognitive strain, not workload volume, as the #1 burnout driver for the first time. Gen Z burnout: 66%. Millennials: 58%. Boomers: 37%. The people most expected to adopt AI are burning out fastest.

This isn’t sustainable without a framework. That’s what A.L.I.C. is for.

What to Do This Week

For Individuals (the employee side of A.L.I.C.):

Shift from mastery to fluency. You don’t need to learn every feature. You need to recognize when a new capability matters for YOUR work and integrate the pieces that create value. Stop trying to “complete” a tool. There is no complete.

Build a 15-minute weekly learning sprint. Pick the AI tool you use most. Open the changelog or “What’s New” section every Monday. Spend 15 minutes exploring one new capability. That’s it. Small, consistent exposure beats quarterly deep dives every time.

Invest in your ideation edge. Read outside your domain. Have conversations with people who solve different problems than you do. Your weird combination of experience, industry knowledge, and pattern recognition is the one thing AI can’t replicate. Feed that.

Accept productive discomfort as the new normal. The period where the old way is faster than the new way is temporary. Your brain overweights that period (behavioral economists call this hyperbolic discounting). Push through it. The payoff compounds.

For Leaders (the leadership side of A.L.I.C.):

Build iteration reviews into the operating rhythm. Monthly “what changed in our AI tools” sessions. Not IT-led. Leadership-led. You go first. When your team sees you learning in real time, it normalizes the discomfort.

Communicate the “why” behind every workflow change. McKinsey’s #1 transformation failure mode: no compelling “why.” Your people can absorb change if they understand why it matters. Release notes aren’t enough. Context is everything.

Monitor change fatigue proactively. Only 42% of burned-out employees tell their manager. Don’t wait for the signal. Ask. Check in. Build it into your 1-on-1s.

Designate an “iteration scout.” Someone on the team whose job includes tracking platform updates and translating them into practical workflow implications. Not an IT role. A bridge role. The person who reads the changelog so 50 other people don’t have to, and tells the team what actually matters for their work.

What A.L.I.C. Doesn’t Solve

It doesn’t eliminate the discomfort. Building adaptation muscles doesn’t make change painless. It makes the pain productive instead of paralyzing. If you’re looking for a framework that makes the acceleration disappear, this isn’t it. The acceleration isn’t disappearing.

It’s not a silver bullet for broken culture. If your organization has toxic leadership, zero trust, or no psychological safety, A.L.I.C. won’t fix that. The framework assumes a baseline of functional organizational health.

The speed vs. quality tension is real. 2025 was the year of AI speed. 2026 is expected to be the year of AI quality. Not every update deserves your attention. Chasing every release is a recipe for exactly the burnout the framework is designed to prevent. Be selective.

Individual adaptation has a ceiling without organizational support. An employee can build all the Adaptability and Ideation in the world. But if leadership doesn’t invest in Learning infrastructure and Change management, the individual burns out alone. The framework requires both sides of the contract.

If You Only Remember Three Things

1. The acceleration is real and it’s not slowing down. AI capability doubling times have compressed from 7 months to under 3 months. Platforms are shipping major updates weekly. This is the operating environment now.

2. The gap isn’t between you and the technology. It’s between how fast the technology iterates and how fast humans adapt. That’s a design problem, not a willpower problem. 66 days vs. 14 days. The math doesn’t work unless you change the approach.

3. A.L.I.C.: Adaptability, Learning, Ideation, Change. Two on employees. Two on leadership. Both sides show up, or neither side wins. Build the muscles. The iteration isn’t stopping. But it can become your advantage.

Good Luck - Dan

This Week in 30 Seconds

Block fired 4,000 people and said AI did it. Wall Street gave them a 24% raise. Google, Meta, and Accenture are now tracking whether employees use AI and tying it to promotions. The Dallas Fed published wage data showing AI is already splitting the labor market along experience lines. And IBM is tripling junior hiring while everyone else cuts. The AI conversation just shifted from “which tools should we buy?” to “who stays, who goes, and what does the team look like in three years?”

Four stories this week. For each one: the news (what happened), the noise (what everyone’s saying), and the signal (what actually matters for you).

Story 1: Block Fires 4,000 People, Blames AI, and Wall Street Loves It

The News: Jack Dorsey’s Block (Square, Cash App, Afterpay) laid off more than 4,000 employees, roughly 40% of its workforce. Dorsey told shareholders that “intelligence tools have changed what it means to build and run a company.” CFO Amrita Ahuja put it more bluntly: they see an opportunity to “move faster with smaller, highly talented teams using AI to automate more work.” The stock jumped 24%. Block’s Q4 was actually strong ($6.25 billion in revenue, gross profit up 24% year-over-year). This wasn’t a struggling company trimming fat. It was a healthy company making a bet.

The Noise: The AI skeptics see a PR-wrapped mass firing. The AI boosters see the vanguard of a leaner future. Tech media defaults to “is AI coming for YOUR job?” Nobody’s talking about the messy middle.

The Signal: Forbes analyst Ron Shevlin called it clearly: Block inflated its headcount by 160% during the pandemic easy-money era (from ~3,800 in 2019 to over 10,000 by 2025). They’d already cut hundreds in early February. The AI framing is convenient, and it’s great for the stock price. Bloomberg is reporting suspicions of “AI washing,” using AI as a narrative cover for restructuring driven by other pressures.

Two things to watch. First, Block is a financial services company. They handle payments, lending, fraud detection, KYC compliance, regulatory reporting. These are functions where mistakes have legal consequences. Cutting 40% of the workforce while betting on “intelligence tools” for compliance is a massive, untested gamble. Second, Klarna’s CEO made the same claim last year (AI helped shrink workforce by 40%), then Klarna had to rehire workers because the AI couldn’t handle the work. That cautionary tale has a shelf life of about 12 months before we find out if Block follows the same path.

For SMB operators, the Block story is really about what’s coming for you. Every company is about to be asked whether they’re “using AI to be more efficient.” You need an answer that’s more honest than Dorsey’s. And honestly, that conversation is uncomfortable. Nobody wants to map out which tasks on their team could be automated. It feels like building a case against the people you work with every day. But that’s not what it is. The operators who navigate this well can articulate exactly which tasks AI handles, which tasks still need humans, and why their team is the right size for the work. That clarity protects your people more than avoiding the question does.

Your Move: Run a task audit this week. For every role on your team, identify which tasks AI could realistically handle today, which need human judgment, and which fall somewhere in between. Not to justify cuts. To know your actual position when the conversation comes.

Story 2: Use AI or Don’t Get Promoted. The Mandate Is Here.

The News: The shift from “we encourage AI” to “we track and enforce it” happened across multiple major companies this week. Google is factoring AI tool use into software engineer performance reviews. Meta’s system can track how much code an engineer wrote with AI assistance. Amazon AWS managers have dashboards showing individual AI-tool usage. Ring requires all promotion applications to explain how the employee is using AI. Accenture trained 550,000 employees on AI and is now monitoring weekly AI tool log-ins for senior staff. A leaked memo: “regular adoption” of AI tools will be a “visible input to talent discussions” for leadership promotions this summer. KPMG is tracking Copilot usage data and baking it into annual reviews. A September 2025 survey found 58% of companies already require some employees to use AI tools.

The Noise: Tech press frames this as a “future of work” trend piece. Labor advocates call it surveillance. AI evangelists celebrate. The practical management tension gets lost.

The Signal: The companies enforcing AI adoption are solving a real problem. They spent millions on tools nobody uses. The knee-jerk fix: mandate and measure. But IT Brew flagged the risk nobody’s discussing. When you mandate AI use and tie it to career advancement, people game the system. They feed sensitive data into consumer AI products. They find ways to check boxes whether the output is useful or not. One CISO quoted in the piece: mandatory AI drives shadow AI and creates security vulnerabilities.

The smarter play (and the one worth stealing): go to the people actually doing the work, ask where they lose time on repetitive tasks that don’t require judgment, and bring AI to those friction points. That’s adoption through value. Mandating tool use without redesigning work creates the worst version of AI slop: more output, less value, plus security exposure.

If you manage people, this story is also a preview. Within 12 months, someone on your team will ask whether they need to be using AI to get promoted. What’s your answer?

Your Move: Skip the mandate. Have one conversation with your team this week: “Where are you losing time on repetitive work that doesn’t require your judgment?” Whatever they say, that’s your AI adoption starting point.

Story 3: The Dallas Fed Put Numbers on What Everyone Was Feeling

The News: The Federal Reserve Bank of Dallas published research showing AI is splitting the labor market along experience lines. Not a survey. Not predictions from a tech CEO. Wage data. The core finding: for occupations with high experience premiums (where years on the job command significantly higher pay), AI exposure boosted wage growth for experienced workers. For occupations with low experience premiums, AI exposure lowered wage growth for both entry-level and experienced workers.

The mechanism comes down to two types of knowledge. Codifiable knowledge (the stuff you learn from books and school) is something AI replicates well. Tacit knowledge (judgment, intuition, context that comes from years of hands-on work) is something AI can’t replicate yet. Entry-level workers primarily do codifiable tasks. For experienced workers in fields like law, insurance underwriting, and credit analysis (where the experience premium exceeds 100%), those same codifiable tasks are the least valuable part of their work. AI handles the grunt work. Their judgment becomes more valuable.

Separately, employment for workers under 25 has fallen. Not from layoffs. From a collapsing job-finding rate. Young workers entering the labor force are finding fewer doors open.

The Dallas Fed researchers flagged something that should make every operator pause: “Leaving new employees off the job ladder is not sustainable in the long run.” Today’s experienced workers became experienced by doing entry-level codifiable work for years. If AI absorbs that layer, where does the next generation of experienced workers come from?

The Noise: Media coverage splits into “AI is coming for your job” panic or “AI will make you more productive” reassurance. The nuance (it’s doing both, simultaneously, to different people) gets lost.

The Signal: This is the most important data release of the week. It moves the workforce conversation from speculation to measurement. The hiring math changes: you might need fewer junior roles doing routine work, but you probably need to keep (and pay more for) the senior people whose judgment AI can’t replicate. In the short term, that saves money. In the long term, if every employer stops developing junior talent, you’ll be competing for an increasingly scarce pool of experienced workers in 3-5 years. I keep thinking about this one. It’s the kind of problem that doesn’t feel urgent today but becomes very expensive very fast.

The operators who get this right will redesign entry-level roles, not eliminate them. Automate the codifiable tasks. Rebuild the role around learning judgment, context, and client interaction alongside AI.

Your Move: Look at your junior roles. Which daily tasks are codifiable (routine, rule-based, learnable from documentation)? Those are the tasks AI absorbs. The question isn’t whether to eliminate those roles. It’s how to redesign them so junior people learn the tacit knowledge they’ll need to become your next senior hires.

Try This Prompt:

For ChatGPT/Claude:

I manage a team of [X] people at a [type of business]. Here are the entry-level roles on my team and their main responsibilities:

[Role 1]: [list 3-5 key tasks]
[Role 2]: [list 3-5 key tasks]

For each role, categorize every task as:
(1) Codifiable — routine, rule-based, learnable from documentation. AI could handle this now or soon.
(2) Tacit — requires judgment, context, relationship, or pattern recognition built from experience. AI can't replicate this yet.
(3) Hybrid — has elements of both.

Then for each role, suggest how I could redesign it so the person spends less time on codifiable work and more time developing the tacit knowledge they'll need to grow into a senior position. Include:
- Which codifiable tasks to automate or hand to AI
- What new tacit-knowledge responsibilities to add
- How the redesigned role builds a pipeline to senior positions

Be specific to my industry. Don't give generic advice.

For Perplexity:

How are companies redesigning entry-level roles around AI instead of eliminating them? Include examples of businesses that automated routine tasks while keeping junior employees focused on developing judgment, client interaction, and tacit knowledge. Focus on practical approaches from 2025-2026.

Story 4: IBM Is Tripling Junior Hiring While Everyone Else Cuts

The News: While Block, Salesforce, Amazon, and Pinterest cut entry-level roles, IBM announced it’s tripling junior hiring in 2026 across software, consulting, cloud, and HR. IBM’s Chief Human Resources Officer Nickle LaMoreaux: “The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment.” She added: “We are tripling our entry-level hiring, and yes, that is for software developers and all these jobs we’re being told AI can do.”

After integrating AI across operations, IBM found the technology has limits. AI handles codifiable tasks well but can’t replace contextual understanding, customer interaction, and judgment. So they’re rewriting entry-level roles. Software engineers spend less time on routine coding, more time with customers. HR staffers intervene with chatbots rather than answering every question manually.

IBM isn’t alone. Dropbox is expanding intern and new grad programs by 25%. Its Chief People Officer told Bloomberg that Gen Z workers are “biking in the Tour de France” on AI proficiency while the rest of the company has “training wheels.”

The Noise: Most outlets frame this as a feel-good Gen Z hiring story. The labor market narrative stays dominated by layoff headlines.

The Signal: IBM’s move reads differently next to the Dallas Fed data. The Fed showed that experienced workers’ value depends on a pipeline of junior workers going through the learning process. IBM is betting the companies that gut their pipelines now will be scrambling for experienced talent in 3-5 years.

The key distinction: Block eliminated roles. IBM redesigned them. Same technological moment. Opposite workforce strategy.

For SMB operators, this is the strategic question of the year. Are you building your team for 2026 or for 2029? And there’s a practical argument beyond pipeline building. Gen Z workers who are AI-fluent can be force multipliers. A 24-year-old who uses Claude to draft client communications, build reporting automations, and prototype solutions might deliver more value at entry-level pay than a mid-career hire who hasn’t adapted. You’re not just building a pipeline. You’re seeding AI fluency through the people most comfortable with the tools.

Your Move: Before you eliminate any junior role, ask: what would this role look like if we redesigned it around AI instead of replacing it with AI? If the answer is “someone who uses AI to handle routine work while learning the judgment and context they’ll need to grow,” that’s a role worth keeping.

The Pattern

Block is cutting people. Google is mandating AI. The Dallas Fed is measuring the split. IBM is hiring in the opposite direction. Four stories from four completely different corners, and they all land in the same place: AI adoption is now a workforce architecture decision. The operators still treating it like a tool purchase are going to have a rough year. The ones treating it like a team-building question are already in a better position.

The Contrarian Corner

The Block layoffs are being covered as either the dawn of the AI workforce revolution or a cynical excuse for corporate restructuring. Both takes miss it. The real story: “AI made us do it” has become an acceptable reason to cut 40% of your staff, and Wall Street will reward you for saying it. That incentive structure is going to drive layoffs across every industry for the next 12 months. The question for every operator isn’t “will AI replace my team?” It’s “how do I build a team that’s genuinely more capable with AI, so I never have to use a CEO excuse as a strategy?”

Your One Move This Week

Run a task audit on one team. For every role, list the five most time-consuming weekly tasks. Mark each one: codifiable(rule-based, routine, learnable from documentation) or tacit (requires judgment, context, relationship, pattern recognition). The codifiable tasks are where AI creates value now. The tacit tasks are where your people create value AI can’t touch. That split is your workforce architecture map. It tells you what to automate, what to protect, and where to invest in developing your next generation of experienced talent.

Good Lick - Dan

This Week in 30 Seconds

The biggest threat to your AI strategy isn’t the technology. It’s the distance between what you bought and what you’re actually using. Four different research reports landed this week, all saying the same thing: organizations are adopting AI faster than ever, and the gap between adoption and impact is getting wider by the month. Smaller, leaner organizations are the ones closing it. Most of them don’t realize they have that advantage yet.

Four stories this week. For each one: the news (what happened), the noise (what everyone’s saying), and the signal (what actually matters for your business).

73% of Organizations Can’t Ship What They Promise

The News: Camunda’s 2026 State of Agentic Orchestration Report dropped a stat that should make you feel better about your own AI journey: 73% of organizations admit a disconnect between their agentic AI ambitions and what they can actually deploy. Only 11% of agentic AI use cases made it into full production last year. The blockers aren’t technical. They’re trust, transparency, and governance. Businesses are pouring money into AI agents, but most projects stall in pilot mode because leaders don’t trust these systems enough to let them touch mission-critical work.

The Noise: Enterprise analysts are recommending more governance committees, more review boards, more oversight layers. The classic enterprise playbook: slow everything down until it feels safe.

The Signal: The 11% number is the one that matters, and most people are reading it wrong. This isn’t a governance failure. It’s a sequencing failure. Most organizations jumped straight to autonomous AI agents before they’d proven they could manage basic AI workflows. They skipped documenting processes, building feedback loops, and establishing who checks what. Then they wondered why 89% of their experiments died in the sandbox.

The vicious cycle Camunda describes is worth understanding: you experiment with agentic AI, hit roadblocks (trust, transparency, control), pull back, and watch bolder competitors inch ahead. Meanwhile, the wasted budget and frustrated teams make the next experiment even harder to greenlight. The trust deficit compounds just like the adoption gap.

For SMB operators, this is quietly great news. You don’t need an “agentic AI strategy.” You need one workflow that works, with a human checking the output. That puts you ahead of almost nine out of ten enterprise use cases that never left the sandbox.

Your Move: Pick ONE workflow where AI already helps you (even partially). Document the steps, define what “good” looks like, and build a simple weekly review. That’s the trust-building exercise that most organizations are skipping entirely.

The Adoption Gap Is Compounding, and the Clock Is Real

The News: Research from Harvard Business School (via Paul Baier at GAI Insights) shows a widening gap between what AI can deliver and what organizations actually capture. AI capabilities improve every quarter. Most organizations adopt on annual planning cycles. The distance grows monthly. Key finding: companies with clear escalation policies for AI agents scale adoption 3x faster.

The Noise: The usual “AI is moving fast, don’t get left behind” urgency. Enterprise consultants building digital transformation maturity models. Nothing actionable.

The Signal: The compounding math is what matters here. This gap isn’t linear. Every quarter you delay meaningful adoption, the distance between you and competitors who’ve figured it out grows exponentially harder to close. Not because the tools get harder. Because the people using them build operational muscle memory that compounds. Your competitor who’s been using AI daily for six months doesn’t just have six months more experience. They have six months of pattern recognition, workflow refinements, and intuition you can’t shortcut.

Baier identifies who’s most exposed: firms that compete on knowledge work. Law firms, consultancies, financial services, insurance. AI directly augments the work these businesses sell. And the risk spreads wider than those industries. Top performers already recognize that AI proficiency defines career advantage, and they’re choosing employers accordingly. Companies slow to adopt don’t just lose market share. They lose their best people to competitors who move faster.

The buried finding: companies with escalation policies scale 3x faster. Translated out of enterprise language, an “escalation policy” just means answering one question in advance: “When the AI gets it wrong, who checks it before it goes anywhere?” That’s it. That simple decision, made before something breaks, is the difference between teams that use AI daily and teams that dabble once a month.

Your Move: Two things. First, spend 15 minutes every day using AI on real cognitive work (strategy questions, draft communications, market analysis). Not prompts for fun. Real work. Baier’s research is clear: CEO daily AI usage is the single strongest predictor of organizational adoption. Second, write your AI escalation policy in one sentence: “When AI output is wrong or unclear, [name] reviews it before it goes to [audience].” You just did what 3x-faster companies do.

AI Doesn’t Save Time. It Quietly Eats It.

The News: Researchers Aruna Ranganathan and Xingqi Maggie Ye from UC Berkeley’s Haas School of Business published findings from an eight-month ethnographic study at a U.S. tech company with about 200 employees. This wasn’t a survey. They conducted twice-weekly in-person observations, tracked internal communication channels, and ran 40+ in-depth interviews across engineering, product, design, research, and operations. Their conclusion: AI tools consistently intensify work rather than reducing it.

The Noise: Skeptics saying “See? AI is overhyped.” Consultants saying “Just need better change management.” Both are wrong.

The Signal: The methodology matters here, and it’s why this study is the most important AI research of the month. Most AI productivity claims come from two-week pilot studies or self-reported surveys. This is eight months of direct observation. It captures behavior people don’t self-report.

Three forms of intensification emerged, and they feed on each other. First, task expansion: product managers started writing code, researchers took on engineering tasks, people attempted work they would have outsourced or deferred. Nobody asked them to. AI made it possible, so they did it. One participant put it simply: “You had thought that maybe you save some time, you can work less. But then really, you don’t work less.”

Second, blurred work boundaries. AI reduced the friction of starting tasks, which sounds like a benefit. But reduced friction also means reduced barriers. Work slipped into lunch breaks, commutes, and evenings. Not because deadlines demanded it, but because starting felt effortless. (If you lived through the smartphone revolution and watched email colonize every waking hour, you’ve seen this movie before. Except this time it’s not just communication leaking into your evenings. It’s the actual work.)

Third, cognitive overload through multitasking. Workers managed three, four, five parallel AI-assisted threads because each individual stream felt manageable. The aggregate load didn’t. Individual task completion times dropped. Total time spent working increased. That’s the paradox.

The competitive dynamic makes it worse. When your colleague uses AI to take on more, standing still feels like falling behind. Informal expectations escalate without anyone formally raising them. Within months, what AI makes possible becomes what’s expected.

For small teams where everyone’s “doing more” because of AI, check whether quality is actually improving or just output volume. A 3-person team producing 5x the content at half the quality isn’t winning. They’re creating 5x the cleanup work.

Your Move: Run a quick audit: “What work are we doing now that we didn’t do six months ago, specifically because AI made it possible?” If that list includes work outside your core competency, cut one thing this week. AI should sharpen your focus, not scatter it.

Measuring AI ROI Is Broken. Smaller Orgs Are Winning Anyway.

The News: KPMG surveyed 2,500+ global tech execs for their 2026 Global Tech Report. The numbers: 74% say AI is producing value, but only 24% achieve ROI across multiple use cases. 58% say traditional ROI measures don’t work for AI. The surprise: smaller organizations with lean governance saw 3.6x more ROI than larger peers. Early adopters saw 3.2x more.

The Noise: Enterprise analysts treating this as a measurement problem. “We need better KPIs for AI.” Consultants building elaborate ROI frameworks. More complexity for a problem that needs less.

The Signal: Read that number one more time. 3.6x more ROI for smaller organizations. KPMG found something without explicitly saying it: the things that make enterprises “serious” about AI (governance committees, multi-stakeholder review boards, phased approval processes) are killing their returns. Smaller organizations with fewer silos, shorter approval chains, and less bureaucratic overhead capture more value because they iterate faster. Try something. See if it works. Adjust. Repeat.

Guy Holland, KPMG’s global leader of the CIO Center of Excellence, said it directly: AI functions as an enterprise transformation, not a discrete deployment. Value emerges unevenly across automation, adoption, and reinvention phases. Traditional ROI models miss this because they’re designed for projects with clear start and end dates. AI doesn’t work that way.

The 58% who say traditional ROI measures don’t work are right, but not for the reason they think. AI value isn’t hard to measure. It’s emergent. It compounds across workflows over time. Your first month using AI for email drafting saves 20 minutes a day. By month three, you’ve changed how you communicate entirely. Measuring “ROI of the email tool” misses the real return, which lives in the behavioral shift.

KPMG also found that 69% of tech execs made security, scalability, and data standardization trade-offs to move faster, neglecting technical debt and talent gaps. The lesson: speed without a foundation just creates a more expensive mess. The organizations that saw 3.6x ROI didn’t skip governance. They kept it lean.

Your Move: Stop building an AI ROI spreadsheet. Track two things instead: (1) How many workflows include AI as a regular step, not a one-off experiment? (2) Are those workflows improving over time (fewer errors, faster completion, higher quality)? If both numbers trend up, you’re capturing compounding value. If not, you’re renting tools.

The Pattern

Four reports. Four research teams. One conclusion: the AI problem shifted from access to execution, and most organizations haven’t noticed. Everyone has the tools. Almost nobody has redesigned the work around them. The organizations pulling ahead aren’t spending more. They’re iterating faster with less bureaucracy. For SMB operators, that’s your structural advantage. But it has an expiration date.

The Contrarian Corner

The industry is still debating “should we adopt AI?” when the data says 88% of organizations already have. Only 6% see meaningful bottom-line impact. The gap isn’t adoption. It’s workflow design. And every week the conversation stays stuck on adoption is another week the execution gap compounds.

Your One Move This Week

Pick one AI-assisted workflow your team runs regularly. Block 30 minutes to map it: What goes in? What comes out? Who checks it? What happens when it’s wrong? If you can’t answer all four questions, that’s your gap. Closing it is worth more than adopting any new tool.

Try This Prompt:

For ChatGPT/Claude:

I want to audit one AI-assisted workflow in my business. The workflow I want to examine is: [describe it].

Walk me through these four questions:
1. What goes in? (inputs, context, data)
2. What comes out? (deliverable, format, audience)
3. Who checks it? (quality gate, review step)
4. What happens when it's wrong? (escalation, correction, feedback loop)

For any question where my current answer is "nobody" or "nothing," suggest a simple fix I can implement this week. Keep it practical for a small team.

For Perplexity:

What are best practices for auditing AI-assisted business workflows in small companies? Focus on quality checks, escalation policies, and measuring improvement over time.

Good Luck - Dan

Anthropic’s Cowork plugins wiped $1 trillion off software stocks. Both Anthropic and OpenAI dropped new models on the same day. They’re running dueling Super Bowl ads. Here’s the 8-minute version of what actually matters, and the one move you should make this week.

This Week in 30 Seconds

Forget “which AI model is winning?” Wrong question.

Both platforms are now capable enough that the bottleneck isn’t the AI. It’s whether you’ve redesigned your workflows to use it. The race between models is over. The race between you and your competitors just started.

I spent the week reading earnings calls, scanning 15+ sources, testing Cowork plugins, and watching $1 trillion evaporate from software stocks, so you get the version that matters to your business in 8 minutes.

Five stories this week. For each one: the news (what happened), the noise (what everyone’s saying), and the signal (what actually matters).

Before You Read: The Shift That Happened This Week

The shift from “AI as a tool you use” to “AI as a colleague you manage” happened this week. Most operators haven’t noticed yet.

Story 1: The SaaSpocalypse Is Your Leverage Moment

The News: On January 30, Anthropic released 11 open-source plugins for Claude Cowork targeting legal, finance, sales, and marketing workflows. By Tuesday, Thomson Reuters dropped 15.8%. LegalZoom sank 19.7%. The S&P 500 software index shed roughly $1 trillion in market value in a single week. HubSpot is down 39% for the year. Figma 40%. Atlassian 35%.

The Noise: Wall Street is panic-selling on a narrative. “SaaSpocalypse” is the buzzword. Every analyst is picking winners and losers.

The Signal: A $200/month AI tool started doing tasks that companies currently pay $50K to $500K per year for in specialized software licenses. That’s available right now. The enterprise software companies are the ones getting hammered, but the disruption hits SMBs differently. You’re not losing stock value. You’re gaining leverage. The same tools threatening Thomson Reuters’ market cap put enterprise-grade capabilities in the hands of a 10-person firm. But leverage only works if you pick it up.

Your Move: Pick one specialized software tool you’re paying for (or wish you could afford). Spend 30 minutes this week testing whether an AI assistant can handle 60% of that workflow. Not perfectly. Just competently enough to change the math.

Story 2: “Vibe Working” Just Entered the Lexicon

The News: On the same day (February 5), Anthropic dropped Claude Opus 4.6 and OpenAI released GPT-5.3-Codex. Opus features a 1 million token context window (5x larger than its predecessor), “agent teams” that coordinate multiple AI workers in parallel, and PowerPoint integration. OpenAI’s model is their “first model that was instrumental in creating itself.” Anthropic’s head of enterprise product coined the term “vibe working.”

The Noise: Tech media is treating this as a horse race. Benchmark comparisons. Speed tests. Who’s winning?

The Signal: “Vibe working” is the non-technical equivalent of “vibe coding.” Describe the outcome, AI handles execution. The 1M token context window means Claude can hold your entire project in memory (every document, every conversation, every constraint) and work across all of it simultaneously. That’s not a chatbot. That’s a junior analyst who never sleeps. And GPT-5.3-Codex literally helped build itself. I’ll be honest, I’m still sitting with that one. The pace of improvement is about to compound in ways none of us have fully thought through.

Your Move: Block one hour this week to test a real project (not a demo, an actual deliverable) with the latest models. Feed it a complete project brief. The gap between what you think AI can do and what it actually does right now is probably wider than you realize.

Story 3: The Super Bowl Ad War Is Actually a Business Model Decision

The News: Anthropic is running a Super Bowl ad mocking OpenAI for testing ads in ChatGPT. Tagline: “Ads are coming to AI. But not to Claude. Keep Thinking.” They also published a manifesto called “Claude is a space to think,” declaring ads incompatible with a helpful AI assistant. Sam Altman fired back on X, calling the ad “deceptive” and claiming more Texans use ChatGPT for free than use Claude in the entire U.S.

The Noise: Tech Twitter is treating this like a celebrity feud. Memes. Hot takes. Entertainment value is high, substance is low.

The Signal: When your AI assistant is ad-supported, the incentives shift. The tool stops working purely for you and starts working partly for advertisers. Same dynamic that turned social media from a communication tool into an attention-harvesting machine. Anthropic compared Claude to “a notebook or a clean chalkboard.” That’s deliberate positioning as a thinking tool, not a consumer product. For operators building real workflows on AI, that distinction matters more than any benchmark score.

Your Move: You don’t need to switch platforms today. But know why your primary AI platform makes money. That determines whose problem it’s solving, yours or an advertiser’s.

Story 4: Experimentation Without Accountability Is Just Waste

The News: A Forbes piece by Janet Lam argues that companies still treating AI as a “side initiative” in 2026 will face tool overload, unclear ROI, and internal frustration. The core thesis: AI isn’t separate from business strategy. It IS business strategy.

The Noise: The “experimentation to execution” framing is everywhere. Every consulting firm has a version of this article.

The Signal: The gap between companies who’ve internalized this and companies who haven’t is about to become visible, and painful. The organizations pulling ahead aren’t the ones with the most AI tools. They’re the ones where a specific person owns a specific outcome, and AI is measured by whether that outcome improves. Sales cycle reduction. Customer response time. Contract review throughput. Real numbers, attached to real people. More AI tools don’t fix unclear goals. They amplify them.

Your Move: Pick one business outcome. Assign one person. Give them 30 days to use AI to move the number. No committees. No “AI strategy” documents. Just a person, an outcome, and a deadline.

Story 5: Your Next Hire Might Be an AI Agent Owner

The News: Writer’s Chief People Officer published a framework arguing the agentic AI revolution creates entirely new roles. The stat: 92 million jobs may be displaced by 2030, but 170 million new roles will be created. Productivity gains of up to 30% with proper AI agent management.

The Noise: The “new AI roles” conversation has been going for a year. Most of it is vague and aimed at Fortune 500 companies.

The Signal: The most valuable professional skill of 2026 isn’t prompt engineering. It’s managing AI agents the way a manager manages people. Setting goals. Defining constraints. Reviewing output. Knowing when to intervene and when to let it run. This is the shift from the “tool user” era to the “orchestrator” era. For SMBs, you don’t need to create a new job title. The person on your team who’s already doing this work informally? Give them the title. Or at least the time.

Your Move: Identify your team’s strongest AI user. Give them 20% of their time to experiment with AI agents, define success metrics, and report back in 30 days. You’re not creating a new position. You’re recognizing a capability that already exists.

The Numbers That Matter

Market Impact

• $1 trillion in software market value wiped in one week (Reuters)

• WisdomTree Cloud Computing Fund down ~20% YTD (CNBC)

• Thomson Reuters’ worst single-day drop on record: 15.83% (VentureBeat)

Model Capabilities

• Opus 4.6 context window: 1 million tokens, up from 200K (Anthropic)

• Enterprise customers make up ~80% of Anthropic’s business (CNBC)

• GPT-5.3-Codex: “first model instrumental in creating itself” (OpenAI)

The Workforce Shift

• 92 million jobs displaced by 2030; 170 million new roles created (Writer.com)

• Productivity gains of up to 30% with proper AI agent management (Writer.com)

What To Do Based on Your Role

If you run operations: Audit your software stack. Identify the two most expensive specialized tools. Test whether an AI assistant handles 60% of those workflows at a fraction of the cost.

If you run sales or marketing: The marketing automation tools getting hammered in the market are getting hammered because AI can do much of what they do. Test one campaign workflow this week. “Vibe working” means describing the outcome and letting AI handle execution. Try it on your next proposal.

If you handle finance or legal: This is where Cowork’s plugins hit hardest. Contract review, compliance tracking, financial analysis. Test one task this week. Not to replace your judgment. To replace the hours of manual prep before your judgment kicks in.

If you’re thinking about hiring: Before posting that next role, ask: “Could 20% of this role be handled by an AI agent?” If yes, you might need a different kind of hire. Someone who can orchestrate AI, not just do the work manually.

Try This Prompt

Take everything in this briefing and apply it to your specific business.

For ChatGPT / Claude:

I run a [YOUR COMPANY SIZE]-person [YOUR INDUSTRY] firm. I need you to act as a strategic operations advisor helping me respond to this week's shifts in AI.

Here's the context you need:
- AI tools like Claude Cowork and ChatGPT can now handle legal, finance, marketing, and sales workflows that previously required $50K–$500K/year in specialized software licenses
- AI context windows now hold 1M tokens — enough to process an entire project's documents, conversations, and constraints simultaneously
- AI models are coordinating teams of agents that work in parallel on complex tasks
- The shift from "AI as a tool I use" to "AI as an operational colleague I manage" is accelerating across industries

Our current software stack includes: [LIST YOUR 3-5 MOST EXPENSIVE OR TIME-CONSUMING TOOLS]
Our biggest operational bottleneck right now is: [DESCRIBE IN 1-2 SENTENCES]

Based on this, give me:
1. Which 2-3 specialized software tools or manual workflows I should test replacing (or augmenting) with AI this quarter — rank by estimated ROI and ease of testing
2. What role or responsibility should be formalized as our "AI operations" function — who on a team like mine is the right person, and what does 20% of their time on this look like?
3. A 30-day pilot structure: what specifically to test in week 1 vs. week 4, who owns each test, and how to measure success with concrete metrics
4. The single highest-ROI workflow I should test first — walk me through exactly how to set up that test this week

Be specific to my industry. Prioritize quick wins that build confidence and create internal momentum. Flag anything where I'd need to keep a human in the loop for quality or compliance reasons.

For Perplexity:

What are the most cost-effective AI alternatives to expensive specialized business software (legal research, financial analysis, marketing automation) for small and mid-size companies in 2026? Include real examples of SMBs replacing enterprise tools with Claude Cowork or ChatGPT, estimated cost savings, and practical limitations.

The Contrarian Corner

The “SaaSpocalypse” coverage is framed as big companies losing value. The real story is small companies gaining power. Every dollar that comes off Thomson Reuters’ market cap represents capability that’s becoming democratized. The stock market is panicking because software monopolies are being dismantled. If you’re a 20-person firm that could never afford LexisNexis, this is your moment, not theirs.

And the model war? Covered like a horse race. The race doesn’t matter to operators. What matters is that both platforms are now capable enough that the bottleneck is you, not the AI. (I include myself in that, by the way. I tested three Cowork plugins this week and realized I’ve been underusing tools I already pay for.)

If You Only Remember 3 Things

1. The SaaSpocalypse is your leverage. The same tools crashing enterprise software stocks put enterprise-grade capabilities in your hands at $200/month. Pick one workflow and test it this week.

2. The most valuable skill in AI just changed. Not prompting. Managing AI agents the way you manage people. Setting goals, reviewing output, knowing when to step in. Start thinking of yourself as an orchestrator, not a user.

3. Your AI platform’s business model matters more than its benchmarks. Know whether your AI is working for you or for an advertiser. That choice compounds quietly. Give it two years and it won’t be quiet anymore.

Your One Move This Week

Pick one workflow that currently requires specialized software or manual effort. Test whether an AI tool can handle 60% of it. Not perfectly. Just competently enough to change the math.

One workflow. One test. One week.

If the result surprises you (and based on what dropped this week, it probably will), you’ll have your answer on where to focus next. And if it doesn’t surprise you, you’ve still ruled something out. Either way, you’re further ahead than last Monday.

If you found this useful, forward it to the person on your team who should be your AI operations lead. They probably already know who they are.

Good Luck - Dan

This Week in 30 Seconds

The hype hangover arrived. Companies are cutting headcount based on AI’s potential, not its performance. They’re paying compliance costs for “AI” labels on products that aren’t actually AI. They’re building agents faster than they can govern them. And the talent gap between AI-fluent and AI-reluctant just became a hiring priority.

The era of “doing something with AI” is over. The era of doing something right with AI is starting.

5 stories this week. For each one: the news (what happened), the noise (what everyone’s saying), and the signal (what actually matters).

Story 1: The Layoff Lottery

The News: A December 2025 survey of 1,006 global executives found that 60% of organizations have made moderate-to-large headcount reductions in anticipation of AI’s impact. Only 2% made those cuts based on actual AI implementation results. Klarna cut 40% of its workforce claiming AI could handle it, then admitted they’d “turned too much work over to AI.”

The Noise: “AI is taking jobs!” headlines continue. CEOs predict white-collar extinction. Economists argue about timelines.

The Signal: Companies are making $50M workforce decisions based on vibes and analyst pressure, not measurement. 44% of executives say gen AI is the hardest form of AI to measure ROI on. They’re guessing. Some are already regretting it. While enterprise plays layoff roulette, smart operators can do the opposite: use AI to make existing teams more dangerous, not smaller. When competitors realize they’ve gutted institutional knowledge, they’ll be competing for the talent you developed.

Your Move: Run narrow experiments with measurement before any headcount decisions. If you can’t prove the automation works, you can’t defend the decision.

Story 2: The AI Label Tax

The News: Regulators are cracking down on “AI washing.” The SEC has issued enforcement actions. The FBI charged a CEO with $40M fraud for claiming his app was “fully automated based on AI” when it was actually hundreds of workers in a Philippine call center. Amazon’s “Just Walk Out” stores relied on 1,000 workers in India checking 75% of transactions.

The Noise: “AI fraud!” headlines. Finger-wagging about corporate ethics. Regulatory posturing.

The Signal: If you call something “AI” that isn’t, you may accidentally trigger the EU AI Act’s requirements (full implementation August 2026). You’ll face stricter vendor scrutiny. You’ll spend money on governance for technology you don’t actually have. The irony: companies bragging about AI capabilities they don’t possess are now paying for compliance on those fictional capabilities. Two moves: First, audit your own language. Are you calling anything “AI” that’s really just automation? Stop. Second, vet your vendors. Ask: “Show me where the machine learning model is actually running.” If they can’t answer, it’s not AI.

Your Move: Pick one vendor this week and ask: “Where is the machine learning model actually running? What is it trained on?” Their answer tells you everything.

Story 3: The Orchestrator Arrives

The News: Anthropic launched “MCP Apps” on January 26. Claude can now integrate directly with Slack, Figma, Canva, Box, and other workplace tools inside the chat interface. Users can send Slack messages, generate graphics, access cloud files, all through Claude. Combined with Claude Cowork (launched January 12), Claude can now execute multi-stage tasks across your actual work systems.

The Noise: “AI assistant gets more capable!” Tech press excitement about feature parity with competing products.

The Signal: AI is moving from “answer questions” to “take actions.” That’s a fundamentally different permission model. And most organizations have no governance for it. Anthropic’s own safety documentation tells users to “be cautious about granting access to sensitive information” and recommends creating dedicated working folders rather than “granting broad access.” Even the company building this is telling you to put guardrails on it. For small teams where one person wears six hats, having AI orchestrate across tools is a force multiplier. But the risk surface just expanded. Your AI assistant now has potential access to your files, your communications, your design assets.

Your Move: If you’re using Claude Pro/Max, enable only the apps you actually need. Create sandbox environments. Establish a governance conversation now, even if you’re a team of three: What can AI access? What can it do without asking? Who’s accountable when it’s wrong?

Story 4: The 80/20 Fluency Gap

The News: At Davos 2026, a consistent finding emerged: only ~20% of senior staff use GenAI daily, compared to 80%+ of Gen Z. Andrew Ng stated companies’ hiring priorities now rank: (1) experienced + uses AI, (2) inexperienced + uses AI, (3) experienced + doesn’t use AI, (4) inexperienced + doesn’t use AI. Some companies are implementing “reverse mentorship” programs where junior staff teach senior leaders AI tools.

The Noise: “Gen Z wins!” generational takes. “Old people can’t adapt” hot takes.

The Signal: Age isn’t the variable. Fluency is. And fluency is now a hiring filter, not a nice-to-have. The middle of the priority list is the danger zone: experienced but not AI-fluent means you’re competing against juniors who produce at mid-level pace. The MIT stat haunts this: ~95% of AI pilots fail to produce measurable ROI. The reason isn’t the technology. It’s the people deploying it. If leadership doesn’t understand AI well enough to question it, they’ll approve projects that don’t work and miss opportunities that do. WEF’s finding: 39% of core skills will change by 2030. Training isn’t a benefit anymore. It’s a business-critical investment.

Your Move: Audit your leadership team’s AI fluency. Not “do they talk about AI” but “do they use it daily for real work.” Consider reverse mentorship. Your most AI-fluent team member might be your newest hire.

Story 5: Build vs. Run

The News: IBM’s VP of watsonx says 2026 is when enterprises shift from building AI agents to operating them, and discovering that operation is harder. Companies now have dozens or hundreds of agents running across platforms, built by different teams, with no unified governance. Only 19% of organizations focus on observability and monitoring in production.

The Noise: “Agents are the future!” Enterprise AI hype cycle continues.

The Signal: “You can build an agent in less than five minutes. The problem is what happens after that.” Companies that rushed to build agents without governance now have a collection of autonomous systems they can’t fully monitor, can’t easily audit, and can’t clearly assign accountability for. Hallucinations at the model layer become operational failures at the agent layer. If an agent hallucinates and calls the wrong tool with the wrong permissions, you have a data leak or a compliance violation. By 2028, Gartner projects ~1/3 of GenAI interactions will occur through agents. The organizations that figure out how to run agents safely will operate 10x faster than those cleaning up after deployments that went wrong.

Your Move: Before you build another agent, define how you’ll monitor it, who’s accountable, and what happens when it’s wrong. Ask the security question: “What’s the worst thing this agent could do if it hallucinated?”

The Pattern

Every story this week points to the same uncomfortable truth: 2026 is when AI moves from “impressive” to “accountable.” Companies are making workforce decisions based on AI’s promise rather than its performance. They’re slapping “AI” labels on products that don’t warrant it. They’re building agents faster than they can govern them. The organizations treating this as an “AI problem” will keep failing. The ones treating it as a management problem that happens to involve AI will pull ahead.

The Contrarian Corner

The narrative this week is “AI is disappointing” or “AI is overhyped.” That misses the point entirely.

AI isn’t disappointing. Implementation is disappointing. The technology works. Companies are just discovering that technology alone doesn’t produce outcomes. You need process, measurement, governance, and skill development. You need to do the boring work.

Your One Move This Week

Pick one AI deployment in your organization and answer three questions:

1. What is it actually doing vs. what we hoped it would do?

2. Who is responsible when it goes wrong?

3. How would we know if it went wrong?

If you can’t answer all three, you have governance work to do before you add anything else.

Try This: The 5-Force Accountability Diagnostic

Run this with your leadership team. It takes 15 minutes and surfaces blind spots you didn’t know you had.

Each force maps to one of this week’s stories:

The Measurement Gap → Story 1: The Layoff Lottery The Labeling Trap → Story 2: The AI Label Tax The Permission Shift → Story 3: The Orchestrator Arrives The Fluency Inversion → Story 4: The 80/20 Fluency Gap The Operations Hangover → Story 5: Build vs. Run

For ChatGPT / Claude / Gemini

Copy this prompt and run it:

You are a strategic advisor helping me assess my organization's AI accountability position. You'll conduct a diagnostic interview across 5 forces, then deliver a scored assessment.

THE 5 FORCES TO ASSESS:

1. The Measurement Gap — Are we measuring AI ROI or guessing?
2. The Labeling Trap — Are we calling things "AI" that aren't actually AI?
3. The Permission Shift — Do we have governance for AI that takes actions (not just answers)?
4. The Fluency Inversion — Does our leadership use AI daily for real work?
5. The Operations Hangover — Can we trust the AI systems we've already built?

YOUR PROCESS:

For each force (one at a time):
- Ask me 2-3 diagnostic questions to understand our current state
- Wait for my responses before moving to the next force
- Take notes on red flags and strengths

After all 5 forces are assessed, provide:

ACCOUNTABILITY SCORECARD
- Score each force 1-5 (5 = strong, 1 = exposed)
- Brief explanation for each score

PRIORITY ACTION
- Identify our single weakest force
- Give one specific, actionable step to address it this quarter
- Suggest one metric to track improvement

Start with Force 1: The Measurement Gap. Ask your diagnostic questions now.

For Perplexity (Research Version)

Use this to prep before running the diagnostic:

What frameworks exist for evaluating enterprise AI governance maturity in 2025-2026? Include assessment criteria for: AI ROI measurement practices, AI labeling accuracy and "AI washing" risks, agentic AI governance policies, leadership AI fluency benchmarks, and AI system observability standards.

For Perplexity (Benchmark Version)

Use this to compare your scores against industry data:

What percentage of organizations have formal AI governance frameworks as of 2025-2026? Include statistics on: AI ROI measurement adoption rates, AI washing enforcement cases, executive AI usage rates by seniority, and AI observability tool adoption in production systems.

How to customize: Replace “my organization” with your specific context. For team use, assign one person to answer each force’s questions. For solo use, answer honestly. The value is in accurate assessment, not optimistic answers.

Sources linked below. The accountability era is here. The only question is whether you’re building the advantage or playing catch-up.

Good Luck - Dan

Story 1: The Layoff Lottery https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance

Story 2: The AI Label Tax https://www.rmmagazine.com/articles/article/2026/01/27/criminally-overhyped--the-risks-of-ai-washing

Story 3: The Orchestrator Arrives https://techcrunch.com/2026/01/26/anthropic-launches-interactive-claude-apps-including-slack-and-other-workplace-tools/

Story 4: The 80/20 Fluency Gap https://www.linkedin.com/pulse/wef-2026-21-takeaways-ai-work-power-kian-katanforoosh-wfbue

Story 5: Build vs. Run https://www.ibm.com/think/news/companies-stop-building-ai-agents-start-running-them

If you’re still asking “how do we get our team to use AI more,” you’re solving the wrong problem.

That’s like asking how to get a teenager to be taller. The height is a symptom. What you’re actually dealing with is a developmental stage. One that’s awkward, volatile, and (the uncomfortable part) can’t be skipped.

You’re not implementing technology. You’re going through a rite of passage that will determine whether your organization matures or self-destructs.

Let me explain.

The Framing That Changed How I See This

Dario Amodei, CEO of Anthropic, recently published an essay arguing that humanity is entering a “technological adolescence.” His point: we’re about to be handed almost unimaginable power, and it’s deeply unclear whether our institutions are mature enough to wield it.

I’ve been sitting with that framing for the last few days. Not because I’m worried about civilization (I mean, I am, but that’s not actionable for most of us). I keep coming back to it because I see the same pattern at organizational scale.

The risks Amodei describes for civilization? They show up in miniature inside companies.

Every organization experimenting with AI right now is going through its own adolescence. Gaining capabilities faster than it’s developing judgment. Experiencing volatility it doesn’t understand. Making decisions without the institutional maturity to handle the consequences.

And just like human adolescence, you can’t skip it. You can only handle it well or poorly.

What’s Actually In Here This Article

• Understand why “AI adoption” is the wrong frame

• See how civilizational-scale AI risks map to your organizational reality

• Get a maturity framework that tells you what stage you’re actually in

• Learn why governance needs to come before scale

• Recognize the signs your organization is mid-adolescence

• Know what “making it to adulthood” actually looks like

The Shift You Need to Make

The reframe:

Most organizations treat AI like they treated the shift to cloud or the rollout of Slack. Buy tools. Train users. Measure adoption. Move on.

That worked for technologies that augmented existing workflows. AI doesn’t just augment. It transforms who makes decisions and how those decisions get made.

Your adoption rate is not your maturity level. I’ve seen organizations with 90% adoption and zero maturity. Tools everywhere. Judgment nowhere.

The Organizational AI Maturity Framework

Five stages. Most organizations are stuck between 1 and 2. The jump from 1 to 2 is often the hardest.

1. Experimentation (Chaos)

Everyone’s using different tools. No visibility into what’s happening. Shadow IT everywhere. Some people are prompting Claude for strategy docs. Others are using ChatGPT to draft emails. A few are running sensitive data through tools nobody vetted.

Leadership is either oblivious or panicked. There’s no policy because nobody knows what to make policy about.

This is where 70% of organizations are right now. If you’re here, that’s fine. But you can’t stay.

2. Awareness (Recognition)

You’ve mapped what’s actually being used. You know who’s experimenting and what’s working. You haven’t solved anything yet, but you can see the landscape.

This stage feels unproductive because you’re gathering information instead of shipping solutions. Resist the urge to skip it. Visibility before strategy. Every time.

3. Governance (Structure)

You’ve built the minimum viable rules. Not a 47-page policy (nobody reads those). A few clear lines about what’s in bounds, what’s not, and how to escalate when you’re unsure.

People follow these rules because they’re reasonable, not because they’re mandated. The test: can a new employee understand your AI boundaries in under 5 minutes?

4. Integration (Workflow)

AI is embedded in how work actually gets done. Not as a side experiment. As infrastructure.

Decisions about AI use are distributed, not centralized. Teams don’t need to ask permission for every new tool. They have the judgment to make those calls. The training wheels are off.

5. Maturity (Identity)

Your organization’s relationship with AI is part of who you are. You know what you use it for, what you don’t, and why.

When a new capability drops (and they drop constantly), your team knows how to evaluate it without escalating to leadership. You can adapt without starting from scratch every time.

Most organizations won’t reach Stage 5 for years. That’s okay. The goal isn’t to rush to maturity. The goal is to know where you are and what the next stage requires.

Why This Framing Matters (The Proof)

Amodei lists five risks that powerful AI poses to civilization. Each one has an organizational equivalent. This isn’t metaphor. It’s structural parallel.

Look at that list. If your organization is in Stage 1 (Chaos), you’re experiencing all five right now. You just might not have named them.

The employee who’s using AI to draft client communications without review? That’s your autonomy risk.

The person running confidential data through an unvetted tool? Misuse potential.

The three people who’ve figured out how to 10x their output while everyone else struggles? Power concentration.

The junior employees who are terrified their jobs are disappearing? Economic disruption in miniature.

The weird vibe in meetings where nobody knows if the ideas are human-generated anymore? Indirect effects.

You don’t need Amodei’s essay to tell you this is happening. You can feel it. What you might need is language for it.

Three Moves You Can Make This Week

Frameworks are useless without action. What you can do in the next 5 days.

1. Run an AI Audit (Stage 1 → Stage 2)

Before you strategize, map what’s actually happening. Simple survey to your team:

• What AI tools are you currently using?

• What are you using them for?

• What’s working? What isn’t?

Don’t make it feel like surveillance. Frame it as curiosity. You’re trying to learn, not catch anyone.

You can’t govern what you can’t see. This is the first step.

2. Define 3-5 Clear Boundaries (Stage 2 → Stage 3)

Not a comprehensive policy. Just:

• What’s definitely in bounds? (Using AI to draft internal docs, brainstorm ideas, etc.)

• What’s definitely out of bounds? (Client-facing content without review, confidential data in unvetted tools, etc.)

• What requires escalation? (New tools, new use cases, anything you’re unsure about)

Write this on one page. If it doesn’t fit on one page, it’s too long.

3. Name Your Maturity Stage Out Loud (Any Stage)

Tell your team where you think you are. Stage 1? Say it. Stage 2? Own it.

Then invite them to disagree. “I think we’re in Stage 2. We have visibility but not governance. What do you think?”

The conversation itself is developmental. Naming where you are creates shared reality. Shared reality enables coordinated action.

When This Framing Doesn’t Help

Honesty builds trust. Where this framework falls apart:

• If you’re a solo operator or tiny team. You don’t have “organizational” maturity to develop. You’re just experimenting. That’s fine.

• If you’re in a highly regulated industry where AI governance is already mandated. You may be forced to skip stages or follow external frameworks. This one won’t replace those.

• If your leadership has already decided AI is either “the future” or “a fad.” No framework survives contact with a closed mind.

• If you’re looking for permission to move slower. This isn’t an excuse to wait. Adolescence doesn’t pause because you’re not ready.

The point isn’t to slow down. It’s to know where you are so you can move with intention instead of chaos.

The Uncomfortable Truth

Your organization’s AI puberty is awkward for the same reason human puberty is awkward.

You’re gaining capabilities faster than you’re developing judgment.

The tools are getting smarter. Your processes aren’t keeping up. Your people are experimenting, sometimes brilliantly, sometimes recklessly. And nobody’s quite sure who’s in charge of making sense of it all.

The only way out is through. But “through” doesn’t mean faster.

It means being honest about where you actually are.

Your Dare

Run the audit. This week. Before you write another strategy doc or evaluate another tool.

Ask your team: What AI are you actually using right now?

I promise the answers will surprise you. And those surprises are the foundation of everything else.

Reply with what you find. I’m collecting patterns.

P.S. This is the first in a series. I’ll have a follow up to this in the coming week.

Next up: The Surgical Intervention Principle, why your 47-page AI policy is making things worse, and what to do instead.

Try This Prompt

Want to run a quick self-assessment? Use this:

For ChatGPT/Claude:

I'm trying to assess my organization's AI maturity level. Here's a framework with 5 stages:

1. Experimentation (Chaos) - No visibility, shadow IT, no governance
2. Awareness (Recognition) - Mapped what's being used, but no rules yet
3. Governance (Structure) - Minimum viable rules that people actually follow
4. Integration (Workflow) - AI embedded in how work gets done, distributed decisions
5. Maturity (Identity) - Clear organizational identity around AI use, can adapt to new capabilities

Based on this framework, ask me 5-7 diagnostic questions to help me figure out which stage my organization is in. After my answers, tell me:
- Which stage we're likely in
- What the biggest gap is between our current stage and the next one
- One specific action to close that gap

Keep it practical. I run a [describe your org size/type].

For Perplexity:

What are the most common signs that an organization is stuck in early AI adoption chaos vs. having real AI governance maturity? Include specific examples and warning signs.

Good Luck - Dan

Everyone’s racing to “learn AI skills” before they fall behind.

But you’re not actually behind.

The “AI skills gap” is a misnomer. Most people already possess the critical capabilities needed to work effectively with AI. They just don’t recognize how directly those skills transfer. The popular question everyone asks is “What AI skills do I need to learn?” The right question is “Which of my existing skills apply directly to AI collaboration?”

I’ve spent the last couple years leading teams through AI integration as a human-centered designer. I’ve watched talented professionals, people who excel at facilitation, problem-solving, and communication, freeze up when asked to “learn prompting.” The pattern became impossible to ignore: the skills gap wasn’t about AI at all.

I’ll show you.

What You’ll Get in This Piece

1. Discover the 75/25 split — Why three of the four critical AI capabilities are already in your toolkit

2. Learn the 75/25 Framework — The exact breakdown of what actually determines AI output quality

3. See the before/after shift — How the mental model changes when you recognize existing skills

4. Get concrete examples — How structured thinking, clear direction, and quality recognition directly apply across industries

5. Understand the 25% that’s actually new — What you genuinely need to learn about LLMs and verification

6. Access ready-to-use prompts — Starting templates that lean on your existing strengths

   ---

   Thanks for reading Quandary Labs! Subscribe for free to receive new posts and support my work.

   ---

The Reframe That Changes Everything

The “AI skills gap” isn’t about missing capabilities. It’s about not recognizing that structured thinking IS an AI skill, clear communication IS an AI skill, and domain expertise IS an AI skill.

Think about that for a second.

You’ve been building these capabilities your entire career. Every project you’ve run. Every meeting you’ve facilitated. Every complex problem you’ve broken down into manageable parts. Every time you’ve had to explain technical concepts to non-technical stakeholders. Every domain where you can spot errors because you know what “right” looks like.

All of it applies directly to working with AI.

The 75/25 Framework

Most AI training gets it backwards. They start with “how to prompt” when they should start with “what you already know.”

Look, I get the confusion. When someone says “learn AI skills,” it feels like you’re starting from scratch. But that’s only true for 25% of what matters. The other 75%? You’ve been building it your whole career without realizing it.

This is how it breaks down.

Your Foundation: Three Capabilities You Already Have

These three areas determine most of your effectiveness with AI. None of them are AI-specific. All of them transfer directly from how you already work.

Capability 1: Structured Thinking

Can you take a messy problem and break it into parts? Can you identify what you know versus what you need to figure out? Can you decide what to tackle first?

That’s what drives good AI collaboration. You’re not asking the AI to solve everything at once. You’re directing it through a methodical process: chunk the work, build on each output, adjust as you go.

This is how you’ve always approached complex problems. Investigation. Decomposition. Sequential problem-solving. The only difference now is you’re guiding a machine through those same steps instead of doing them yourself or directing a team.

The better you are at structuring problems, the better you’ll be at getting AI to help solve them. Simple as that.

Capability 2: Clear Direction

Can you explain what you want in a way others can understand and act on? Can you provide enough context without overwhelming? Can you keep collaborators focused when they start to drift?

That’s what prompting actually is: giving clear direction.

Think about what you do when someone on your team misunderstands an assignment. You don’t start over from scratch. You clarify the goal, add missing context, correct their course, and confirm understanding. Same mechanics apply here.

Your prompts work when they do what good communication always does: establish shared understanding, set clear expectations, and guide progress toward a specific outcome. If you’ve ever written a project brief, coached someone through a task, or redirected a conversation that went sideways, you already know how to do this.

The medium changed. The principles didn’t.

Capability 3: Quality Recognition

Can you look at work in your domain and immediately spot what’s good, what’s wrong, and what’s missing? Can you separate signal from noise? Can you tell when something sounds right but is actually off?

That’s your quality filter. And it’s the most underrated part of working with AI.

Two people can run the same prompt. One gets value, the other gets garbage. One person can recognize quality. The other can’t. That’s the difference.

Your domain knowledge determines whether you trust bad output or catch it. Whether you ask follow-up questions that matter or accept generic responses. Whether you apply what the AI produces or recognize when it’s headed in the wrong direction.

Experienced professionals often “get” AI faster than junior people. They can immediately tell when output is useful versus when it’s plausible-sounding nonsense. That’s domain expertise in action.

The Bridge: One New Skill That Connects Everything

This is the 25% that’s actually new. And even this isn’t complicated. It’s just understanding your collaborator.

What You’re Actually Learning:

You need to know that AI generates text probabilistically, one word at a time, based on patterns it learned. This means:

• It can sound confident while being wrong

• It can drift mid-response and contradict itself

• It can’t verify its own accuracy

• It has no memory outside the current conversation

Once you understand that, the rest is about risk management. You learn to recognize when AI outputs need verification (high-stakes decisions, unfamiliar domains) versus when you can move fast (brainstorming, first drafts, exploration).

You combine this understanding with your domain expertise (Capability 3) to build reliability checks. You develop a feel for when to trust, when to verify, and when to correct course.

Think of it like this. Your three existing capabilities are your engine. This new understanding is the transmission that connects your engine to the AI’s capabilities. Without it, you can’t transfer power effectively. But the engine itself? You already built that.

The point: 75% of what determines your AI effectiveness comes from professional capabilities you’ve spent years developing. The remaining 25% is learning how LLMs actually work so you can apply those capabilities appropriately.

What This Actually Looks Like in Practice

The framework reveals why some people seem to “get” AI immediately while others struggle.

Now watch how this plays out across different roles and industries.

The Project Manager

Sarah already knows how to chunk complex work into phases. She knows how to set clear goals at the outset of a project. She knows how to redirect drifting conversations when her team gets sidetracked.

When she started using AI for project planning, she applied those exact same skills. She recognized that her project management instincts transferred directly. Break down the deliverable. Guide the AI through each component. Redirect when outputs drift off-scope.

She got better outputs immediately. She recognized existing skills.

The Writer

Marcus already knows how to structure arguments. He knows how to use precise language. He knows how to iterate on drafts, tightening logic and improving clarity with each pass.

Those skills transfer directly to prompt refinement and output evaluation. When the AI produces something generic, he recognizes it the same way he’d recognize weak writing from a junior colleague. When the structure feels off, he adjusts his guidance the same way he’d restructure an outline. His writing expertise makes him effective with AI.

The Marketing Analyst

Diana has strong foundation capabilities for research tasks. She knows how to break down complex market questions, structure her investigation, and verify data against what she knows about customer behavior. When she uses AI for research, she’s fluent: confident, effective, getting high-quality results quickly.

But she’s still building confidence with “The Bridge” (understanding LLM behavior) for customer-facing work. She understands verification protocols intellectually, but hasn’t internalized them enough to trust AI outputs that will go directly to customers. So she’s hesitant in that application.

The framework reveals something important: it’s task-specific readiness based on which capabilities are strong for which applications. Diana doesn’t need to “get better at AI” in general. She needs to build verification protocols specifically for customer communications, and she’ll become fluent there too.

The HVAC Technician

James already knows how to diagnose problems systematically. When a customer calls with a heating issue, he doesn’t just guess randomly. He asks targeted questions, rules out possibilities, and narrows down to the root cause. That’s structured thinking in action.

He knows how to explain technical issues to homeowners in plain language. He translates complex HVAC concepts into terms non-technical people can understand and make decisions about. That’s clear direction and communication.

He knows from years of experience when something doesn’t match expected performance. He can look at system behavior and spot immediately when readings are off or components are failing. That’s quality recognition through domain expertise.

Those same skills let him work with AI to generate maintenance schedules that make technical sense, draft clear customer communications that translate complexity appropriately, and verify technical recommendations against his domain knowledge. The gap is just learning basic LLM behavior: understanding probabilistic generation, recognizing when to verify versus trust, building protocols for high-stakes outputs.

He’s applying his HVAC expertise and diagnostic skills through a new interface.

The Retail Store Manager

Maria chunks complex inventory problems into manageable parts. She communicates clearly with suppliers about needs and with staff about priorities. She uses domain knowledge accumulated over years to spot pricing errors or suspicious patterns that newer employees miss.

When she approaches AI for scheduling optimization or customer communication, those existing skills directly apply. She’s 75% of the way there before she types a single prompt.

The Field Service Coordinator

Tom breaks down service routes systematically, considering travel time, job complexity, and technician capabilities. He writes clear dispatch instructions that his team can follow without confusion. He knows from experience when a job estimate seems off (either too high or dangerously low).

All three foundation capabilities, immediately applicable to working with AI for route optimization, communication templates, and estimate verification.

The pattern is clear across every role: the effectiveness comes from applying existing professional capabilities.

How to Start From Where You Are

Now that you see the 75/25 split, here’s how to actually use this framework.

1. Self-Assess Against Your Foundation Capabilities

Copy this into a note and fill it out honestly. It will take you less than 10 minutes and will show you exactly where you stand.

Capability 1 — Structured Thinking:
- Can I break complex problems into component parts?
- Do I have approaches I use to navigate unfamiliar territory?
- Rating: Strong / Developing / New Territory

Capability 2 — Clear Direction:
- Can I write clear briefs or redirect confused conversations?
- Do I structure information so others can follow my thinking?
- Rating: Strong / Developing / New Territory

Capability 3 — Quality Recognition:
- In which domains can I verify accuracy and spot errors?
- Where can I apply best practices and catch mistakes?
- Rating: [List your strong domains]

The Bridge — LLM Understanding:
- Do I understand probabilistic generation and why AI can be confidently wrong?
- Can I recognize when AI is drifting and needs redirection?
- Do I have verification protocols for high-stakes outputs?
- Rating: Strong / Developing / New Territory

Most people discover they’re Strong or Developing in the three foundation capabilities, and New Territory only in The Bridge. That’s the point. You’re not starting from zero.

2. Start an AI Conversation Using Your Direction Skills

Treat it like briefing someone on a task. You wouldn’t hand off work without context, goals, and constraints. You’d be clear about what success looks like.

Do the same thing with AI. Use this structure:

Try This Prompt

For ChatGPT/Claude/Gemini:

I'm working on [specific goal or problem you want to solve].

Context: [2-3 sentences explaining relevant background, constraints, or what you've tried already]

My role and expertise: [Your domain knowledge, experience level, or perspective on this topic]

What I need from you: [Specific output, thought partnership, analysis, or deliverable you're looking for]

Constraints or preferences: [Format requirements, length, tone, approach, or limitations]

Let's start by [first concrete step or question to begin with].

This isn’t a “prompt template” in the traditional sense. It’s how you’d brief any collaborator on a task. You’re just applying that structure to a conversation with AI instead of with a colleague.

Note: This is a collaboration structure template, not a search query. For research questions, see the Perplexity prompts in the P.S. section below.

3. Apply the Chunk → Build → Redirect Pattern

You already do this when managing complex work or guiding confused team members. Now apply it to AI.

Chunk: Break the problem into sequential parts. Don’t try to tackle everything at once. Identify distinct phases or components and work through them in order.

Build: Work through one part at a time. Use each output as input for the next. Build momentum and context gradually instead of expecting perfect results from a single massive prompt.

Redirect: When the conversation drifts (and it will), bring it back to the original goal and key constraints. This is exactly what you’d do when a team member goes off on a tangent. You acknowledge the point, then guide back to the objective.

4. Recognize Drift Using Your Management Instincts

You already know how to spot when work is going off-track. You feel it when someone’s contradicting earlier decisions. You notice when things get too abstract and need concrete grounding. You recognize when the thread is lost and you need to reset.

Apply those same instincts to AI conversations.

If the AI starts to:

• Contradict earlier guidance → Redirect to original constraints. Remind it what you established at the start.

• Go too broad or abstract → Ask for specific, concrete examples. Ground the output in real scenarios.

• Lose the thread → Start fresh. Don’t try to salvage a conversation that’s gone completely off the rails. Take the best outputs from the previous thread and begin a new one with clearer framing.

• Sound confidently wrong (based on your domain expertise) → Stop immediately. This is where your quality recognition protects you. Verify the claim. Correct the error. Don’t let confident-sounding nonsense pass through just because the AI delivered it with authority.

You’re applying management and quality control skills you already have. The only difference is your collaborator has different failure modes than humans do.

When This Framework Doesn’t Fully Apply

I need to be honest about limitations. This framework is powerful for specific contexts and less useful for others.

Pure technical AI engineering: If you’re building models, fine-tuning systems, or working on AI infrastructure, you need deep technical knowledge beyond these capabilities. This framework is for AI operators, not AI engineers.

Zero domain knowledge territory: If you’re working in a completely unfamiliar domain without expertise to verify outputs, quality recognition is missing. You can still use AI in that situation, but you need to either gain enough knowledge to spot errors, or partner with someone who has domain expertise. Without verification capability, you’re trusting outputs you can’t validate.

Advanced context engineering: Building RAG systems, API integrations, or custom memory architectures requires technical implementation skills. The 75/25 Framework will help you design the approach and understand what you need, but you’ll need engineering support to actually build it.

When you’re avoiding experimentation: This framework helps you recognize readiness, but it doesn’t replace hands-on practice. If you’re waiting for “full understanding” before trying anything, you’ll stay stuck. Recognition plus experimentation is the unlock. Understanding the framework should give you confidence to start, not permission to keep researching.

If you’re already competent and seeking breakthrough: This framework explains how you got to competence, but it won’t break you through to mastery. You need advanced context engineering, sophisticated workflows, and domain-specific techniques beyond these capabilities. This is foundation-to-intermediate territory, not intermediate-to-advanced. If you’re already getting good results and want to reach the next level, you need different guidance than what this framework provides.

What This Framework IS Good For

• Everyday AI collaboration for knowledge work

• Getting unstuck when you feel “behind”

• Understanding why some people seem naturally good with AI (they have strong foundation capabilities)

• Building confidence to experiment from your current capability base

• Helping teams recognize existing AI readiness

• Diagnosing why you’re fluent in some applications but not others

The framework won’t make you an AI expert overnight. It will help you recognize that you’re already most of the way there.

Your Challenge

Open an AI tool right now.

Don’t learn a single new “AI skill” first. Don’t read another article about prompt engineering. Don’t take a course.

Just treat it like directing work. Set a clear goal. Provide context. Guide the conversation. Redirect when it drifts. Use the exact skills you already have.

You’ll be surprised how much of it just works.

The 75% you need? You’ve been building it your entire career.

The 25% that’s new? You’ll learn it faster by doing than by studying.

Start where you are. Stop waiting to be “ready.”

P.S. — Three Starter Prompts That Lean on Your Existing Strengths

Want to see this framework in action? I’ve created three starter prompt templates matched to different capability profiles. Copy-paste ready — just customize the [bracketed sections] for your situation.

Each includes both a conversational AI version (for ChatGPT, Claude, Gemini) and a research version (for Perplexity) so you can use whichever fits your workflow.

1. The Structured Thinker’s Prompt

For people who break down complex problems systematically. This prompt leverages your Capability 1 strength.

For ChatGPT/Claude/Gemini:

I need help thinking through a complex problem systematically.

The problem: [Describe the challenge or question you're working on]

What I know so far:
- [Key fact or constraint 1]
- [Key fact or constraint 2]
- [Key fact or constraint 3]

What I don't know yet:
- [Unknown or uncertainty 1]
- [Unknown or uncertainty 2]

Help me:
1. Break this problem into distinct component parts
2. Identify which parts I should tackle first and why
3. Surface any assumptions I might be making that we should test
4. Suggest a step-by-step approach to work through this methodically

As we work through this, pause after each major step so I can verify we're on the right track before continuing.

For Perplexity:

What are proven frameworks and methodologies for breaking down complex [your domain] problems systematically? Include step-by-step problem-solving approaches from 2022-2024 research.

2. The Clear Communicator’s Prompt

For people who give great direction. This prompt leverages your Capability 2 strength.

For ChatGPT/Claude/Gemini:

I need help structuring and clarifying communication for [specific audience or purpose].

What I'm trying to communicate: [Your core message or goal]

My audience: [Who they are, their level of familiarity with the topic, what matters to them]

The challenge: [What makes this communication difficult — complexity, sensitivity, confusion, competing priorities, etc.]

What I've drafted so far (if anything): [Paste your rough draft, bullet points, or notes]

Help me:
1. Identify the single most important point I need to get across
2. Structure this message so it's clear and easy to follow
3. Spot anywhere I'm being vague, using jargon, or assuming knowledge they might not have
4. Suggest where I should add context or examples to make this concrete

Give me a revised version, then explain what you changed and why.

For Perplexity:

What are evidence-based best practices for clear business communication to [your audience type]? Include frameworks for structuring complex messages and reducing ambiguity from recent communication research.

3. The Domain Expert’s Prompt

For people with deep subject knowledge. This prompt leverages your Capability 3 strength.

For ChatGPT/Claude/Gemini:

I'm a [your role/domain expertise] working on [specific task or decision].

My domain knowledge includes:
- [Area of expertise 1]
- [Area of expertise 2]
- [Relevant experience or context]

The task: [What you're trying to accomplish]

What I need from you:
1. Help me apply best practices from [your domain] to this specific situation
2. Ask me clarifying questions about [domain-specific considerations] before suggesting an approach
3. Flag anywhere your suggestions might conflict with [domain standards, regulations, or constraints]
4. When you make recommendations, explain the reasoning so I can verify it against my domain knowledge

I'll correct you if something doesn't align with [industry standards / technical requirements / domain realities]. I need you to adapt based on that feedback rather than defending your original suggestion.

Let's start by you asking me 2-3 questions to understand the specific context before proposing any solutions.

For Perplexity:

What are current best practices and emerging standards in [your domain/industry] for [your specific task]? Include peer-reviewed sources and industry publications from the past 2 years.

And if you try the framework and discover something surprising about which capabilities came naturally versus which needed work, I'd genuinely love to hear about it. The pattern keeps revealing new nuances.

Good Luck - Dan

This Week in 30 Seconds

The companies getting ROI from AI are asking harder questions: Does this actually work? Can I measure it? Should I trust it?

Four stories this week point to the same shift. For each one: the news (what happened), the noise (what everyone’s saying), and the signal (what actually matters). The hype cycle is giving way to a trust reckoning. And the operators who see it clearly will make better decisions than the ones still chasing magic.

Story 1: The ROI Gap Nobody Wants to Admit

The News: PwC surveyed 4,454 CEOs across 95 countries. 81% are prioritizing AI investment (up from 60% last year). But only 21% report actual revenue growth from AI. At Davos, Writer CEO May Habib dropped a bomb:

“The fundamental physics of this is the executives are saying, ‘Change stuff with AI,’ and then they’re giving people AI assistants and productivity tools, and you’re not going to get the wholesale reinvention that actually drives impact.”

The Noise: “AI adoption is accelerating!” “Companies are going all-in!” “This is the year of transformation!”

The Signal: The gap comes down to one thing: the difference between adding tools and changing workflows.

Habib nailed it: “Silos are getting flattened. It makes no sense for sales and marketing to be separate teams for most companies. You’ve got to really break those silos to change workflows end-to-end.”

That 21%? They asked a different question: What work should stop existing entirely?

But there’s more hiding in the data: Most companies skipped the baseline. They have no idea if AI is helping because they never measured what “before” looked like. You can’t calculate ROI if you don’t know your starting point.

Your Move: Before your next AI initiative, answer three questions: (1) What does this workflow cost us today in time, money, and errors? (2) What would “this workflow doesn’t exist anymore” look like? (3) Who owns measuring the before and after? I’ve watched teams skip these questions and spend six months wondering why nothing improved. Don’t be that team.

Story 2: The Skill You Just Learned Is Already Obsolete

The News: Forbes declared prompt engineering is no longer the most valuable AI skill. As AI evolves from chatbots that wait for instructions to systems that act on their own, the capability that matters is knowing when to trust AI, how much oversight is needed, and where human judgment remains essential. The quote that stuck with me: “AI skills are no longer technical skills; they’re leadership skills.”

The Noise: “Master these 50 prompting techniques!” “Prompt engineers are the new developers!” “Here’s how to write the perfect prompt...”

The Signal: The article buries the real insight in a banking example. In an agentic workflow, AI handles document gathering, compliance checks, and back-and-forth communication. But at key moments (borderline risk scores, unusual customer profiles) human judgment kicks in.

The skill isn’t prompting. It’s pattern recognition for when NOT to automate.

Think about what this means. The most valuable people will be the ones who know which things AI shouldn’t do. That’s closer to management judgment than technical skill.

Your Move: Pick one workflow your team uses AI for. Map it out and identify three things: where AI runs unsupervised, where humans currently intervene, and where humans SHOULD intervene but don’t. That third category is where your risk is hiding.

Try This Prompt:

For ChatGPT/Claude:

I want to audit one of my AI-assisted workflows for oversight gaps.

The workflow: [Describe your workflow — e.g., "We use AI to draft customer emails, then a team member reviews before sending"]

Help me map three things:

1. WHERE AI RUNS UNSUPERVISED
   - Which steps happen without human review?
   - What decisions is AI making autonomously?

2. WHERE HUMANS CURRENTLY INTERVENE
   - What checkpoints exist today?
   - What triggers human review?

3. WHERE HUMANS SHOULD INTERVENE BUT DON'T
   - What's slipping through?
   - Where could errors cause real damage?

For each gap in category 3, give me:
- What could go wrong (specific scenario)
- How we'd catch it (detection method)
- What it costs if we don't (business impact)

Be direct. I want actionable gaps, not generic warnings.

For Perplexity:

What are the most common oversight gaps in AI-assisted business workflows? Include specific failure modes, detection methods, and case studies of AI errors that human review would have caught. Focus on practical business applications 2024-2026.

Story 3: Your Team Can’t Tell Real from Fake Anymore

The News: The World Economic Forum’s Global Cybersecurity Outlook 2026 warns that AI-driven fraud has overtaken ransomware as the top cyber risk. 73% of CEOs surveyed said they (or someone in their professional or personal network) had been affected by cyber-enabled fraud in 2025. The FTC reported $12.5B in consumer fraud losses in 2024, up 25% year-over-year. And that’s just what got reported.

The Noise: “Deepfakes are scary!” “AI is being used for evil!” “We need more regulation!”

The Signal: Forget nation-state attacks. The real story is that everyone can now create convincing fakes.

The same tools that help your marketing team personalize outreach help scammers personalize their cons. Your old playbook (look for typos, suspicious links, urgent requests) is obsolete. AI-generated scams don’t have typos. They’re written in perfect, contextually appropriate language. They reference real details about your company, your vendors, your recent transactions.

Your team’s biggest vulnerability is the assumption that they can tell real from fake.

SMBs are particularly exposed. Smaller teams mean fewer verification layers. Relationship-based business makes “I trust that voice” dangerous. Less security infrastructure means more reliance on human judgment, and that judgment just got a lot harder.

Your Move: This week, implement one verification protocol. Any payment change request gets confirmed via a different channel (email request? Call to verify). Any “urgent” request from leadership gets a 15-minute delay and direct confirmation. Any new vendor contact gets verified through your existing records, not the contact info they provide. Yes, it adds friction. That’s the point.

Story 4: Your AI Assistant Now Has a Side Hustle

The News: OpenAI is rolling out advertising in ChatGPT, starting with beta brands committing $1M each. Ads will appear “at the bottom of answers when there’s a relevant sponsored product or service based on your current conversation.” Sam Altman once called advertising a “last resort” and “unsettling.” Here we are. Google’s DeepMind CEO, at Davos, took a shot:

“It’s interesting they’ve gone for that so early. Maybe they feel they need to make more revenue.”

The Noise: “ChatGPT is selling out!” “This is the end of AI trust!” “Advertising ruins everything!”

The Signal: Forget the hypocrisy angle. The real story is the incentive shift.

When ChatGPT’s business model was subscriptions, the incentive was: give you the best answer so you keep paying. Now ads enter the picture. New incentive: give you an answer that creates ad-serving opportunities.

Do these conflict? Not always. But they’re not perfectly aligned either.

When you ask “what’s the best CRM for a 20-person sales team,” are you getting the best answer or the sponsored answer? The answer is probably “both,” but “both” is different from “just the best answer.”

Three implications: (1) Platform dependency risk is real. If your workflows depend heavily on one AI tool, you’re now dependent on that tool’s business model decisions. Diversification isn’t paranoia. (2) Free tiers get complicated. The ad-supported experience will differ from paid. Factor this into tool decisions. (3) Adjust your default trust setting. “Skeptical by default” is healthier than “trusting by default,” especially for purchase decisions.

Your Move: For high-stakes decisions (vendors, purchases, strategy), treat AI recommendations like you’d treat a recommendation from a salesperson: useful input, but verify independently.

The Pattern

Four stories. One theme. Trust is the new bottleneck.

Can we trust the investment thesis? (Only 21% are seeing returns.) Can we trust what we’ve learned? (The skills are already shifting.) Can we trust what we see and hear? (AI makes deception trivially easy.) Can we trust the tools themselves? (They have their own incentives now.)

Call it what it is: maturity. Technology adoption looks like this when the hype fades. The honeymoon phase (where AI felt like magic and every implementation felt like progress) is giving way to harder questions about reliability, measurement, and sustainable adoption.

See this clearly, and you’ll make better decisions than the ones still chasing magic.

The Contrarian Corner

Everyone’s framing the ROI gap as “companies aren’t using AI right.” That’s backwards.

The real problem: they skipped the boring pre-work. They didn’t measure their baselines. They didn’t question whether workflows should exist at all. They bought tools instead of asking questions.

That 21% getting returns? They didn’t have better AI. They had better discipline.

Your One Move This Week

Run a trust audit on your most-used AI tool.

Pick one — the tool you rely on most — and answer these questions:

1. What is this tool’s business model, and how might that affect its recommendations?

2. What happens if this tool disappears or changes tomorrow? Do you have a backup?

3. Are you measuring what this tool actually delivers, or just assuming value?

Companies winning with AI in 2026 know exactly what each tool is good for (and what it’s not).

That’s the week. The honeymoon’s over. The real work starts now.

Good Luck - Dan

“Which AI tool should we standardize on?”

I’ve heard this question from three clients this month alone. One built a 47-page RFP. Another has a committee that’s been evaluating options for 8 months. A third is waiting for “the dust to settle” before committing.

They’re all making the same mistake: applying enterprise software logic to a fundamentally different category.

I spent the last few weeks reviewing 45+ sources (industry reports from Menlo Ventures, McKinsey, Deloitte, Andreessen Horowitz) trying to answer a simpler question: what are organizations that are actually succeeding with AI doing differently?

What I found surprised me. They’re not picking the “right tool.” They’re building the right infrastructure for experimentation.

What you’ll walk away with:

1. Why the enterprise software playbook fails for AI and the data that proves it

2. The R&D budgeting mindset including the specific percentage to reserve for experimentation

3. The 4-Pillar Enablement Framework (policy, data classification, training, guardrails) that enables speed

4. The Traffic Light System for removing bottlenecks without removing guardrails

5. The 30-Day Quick-Start Checklist so you know what to do this month

6. The honest caveats because this approach does have limits

The single sentence version:

Your CRM required a 3-year roadmap. Your AI portfolio requires a 90-day pilot cycle.

That shift changes how you approach every decision that follows.

Why Enterprise Software Logic Fails Here

When you bought your CRM or ERP system, you were making a 3-5 year commitment. Migration costs were real (often millions). Data lock-in was significant. The right play was: evaluate thoroughly, negotiate hard, implement once.

AI tools don’t work that way.

Most AI tools operate on monthly subscriptions ($20-30/user for consumer tiers, usage-based for API access). Model capabilities change quarterly, not annually. Claude, Gemini, and ChatGPT all shipped major capability jumps within 2025 alone.

Andreessen Horowitz found that enterprises in 2024 “were designing their applications to minimize switching costs and make models as interchangeable as possible.” The switching cost conversation has fundamentally shifted.

The data backs this up:

• 76% of AI use cases are now purchased rather than built internally (Menlo Ventures, 2025)

• 31% of AI use cases reached full production in 2025, double the 2024 figure (ISG)

• Many professionals are running multi-model portfolios: Claude for coding, Gemini for research, ChatGPT for general tasks, Perplexity for citation-backed research

The winners aren’t picking winners. They’re building portfolios matched to use cases.

The 4 Pillars of AI Enablement Infrastructure

Before “budget like R&D” makes sense, you need the right infrastructure in place. Without it, experimentation turns into chaos. With it, experimentation becomes learning.

Pillar 1: Policy (The 2-Page Version)

You don’t need a 100-page manual. Honestly, nobody will read it anyway.

You need clarity on four things:

• Where AI can assist vs. where it must be avoided

• Where human judgment must lead

• What data can and cannot be input

• The requirement that AI output be reviewed before use

Sample language: “Employees may use approved AI tools for internal content creation. AI tools may NOT be used with client confidential information or production systems without IT approval. All AI-generated code requires peer review.”

That’s it. A 2-page AI policy beats a 100-page manual because people will actually read it. Clarity enables speed.

Pillar 2: Data Classification (Trust Tiers)

Not all data is equal. A simple three-tier system prevents confidently wrong automation:

Most organizations have never classified their data this way. Start simple: customer PII is Tier 3. Internal operational data might be Tier 2. Public marketing content is Tier 1.

You can get more sophisticated later. Right now, you need the categories to exist.

Pillar 3: Training (30-45 Minutes)

The training gap is massive. According to industry surveys, 44% of companies don’t use proven learning methods for AI and have no plans to start.

But you don’t need a semester. A 30-45 minute module covering these four areas (with an annual refresh) is the minimum viable investment:

• What AI does and doesn’t do

• How to interpret outputs

• When to override

• How to report issues

The point is baseline competence, not expertise. You’re not training data scientists. You’re making sure people know the guardrails exist.

Pillar 4: Technical Guardrails (Enable, Don’t Block)

Here’s the thing about banning AI tools: your employees will use them anyway.

43% already share sensitive data with AI tools without employer permission (Cybsafe/National Cybersecurity Alliance, 2025). Pretending otherwise just means you don’t know what’s happening.

The smarter play is governed alternatives:

• Pre-approved tool catalog (start with 2-3)

• “Nudge” system that guides rather than blocks

• Sandbox environments for experimentation

• DLP rules tailored for AI platforms

Governance doesn’t slow experimentation. It removes the uncertainty that’s actually slowing you down. When people know what’s allowed, they move faster.

The Data Behind This

The Shadow AI Reality

• 80%+ of employees use AI tools for work; fewer than 30% of organizations have formal policies (Ocean Solutions, 2025)

• 43% of employees share sensitive work information with AI tools without employer permission (Cybsafe/National Cybersecurity Alliance, 2025)

• $670,000 premium in additional breach costs for organizations with high unsanctioned AI usage (IBM, 2025)

The ROI Measurement Challenge

• 97% of organizations struggle to demonstrate GenAI business value (KPMG/Informatica)

• Only 51% can confidently evaluate AI ROI (CloudZero, 2025)

• Yet 74% say their most advanced AI initiatives meet or exceed ROI expectations (Deloitte, 2024)

What’s Actually Working

• 91% of SMBs using AI report revenue lift (Salesforce, 2024)

• Enterprises are explicitly designing applications to “minimize switching costs and make models as interchangeable as possible” (Andreessen Horowitz, 2025)

• 15-20% budget reserved for experimentation is the expert-recommended allocation (McKinsey/Google, 2025)

The Traffic Light Classification System

This is the simplest way to remove bottlenecks while maintaining boundaries:

Most day-to-day AI usage should fall into Green. Yellow covers new tools or unfamiliar use cases. Red is reserved for anything touching customer data, production systems, or regulated information.

The power of this system: it removes the “ask permission for everything” bottleneck while keeping guardrails where they matter. People stop waiting. Work gets done.

The 30-Day Quick-Start Checklist

Week 1-2: Foundation

• Audit current AI usage (including shadow AI, it’s already happening)

• Designate an AI Lead (can be owner, ops manager, or fractional)

• Create 1-2 page acceptable use policy

• Define data classification (Tier 1/2/3)

• List 2-3 approved tools

Week 3-4: Enable

• Deploy 30-45 minute training

• Communicate policy to all employees

• Set up feedback channel for AI questions

• Identify 1-2 AI champions per department

Ongoing Rhythm

• Quarterly policy review

• Monthly AI champion sync

• Annual training refresh

The Budget Allocation

Stop thinking in multi-year contracts. Start thinking in 90-day cycles.

• Plan for monthly AI subscriptions, not multi-year commitments

• Reserve 15-20% of AI budget for experimentation (McKinsey recommendation)

• Track AI costs separately from traditional software

• Expect usage-based costs to vary; build in contingency

The 15-20% experimentation reserve is critical. As McKinsey put it in their research with Google: “Companies do need some budget that’s just for experimentation. You don’t want too high a bar for just getting your foot in the door.”

The Honest Caveats

I’d be doing you a disservice if I didn’t mention where this approach has limits.

Agentic AI is creating real lock-in.

From Andreessen Horowitz (2025): “As companies invest the time and resources into building guardrails and prompting for agentic workflows, they’re more hesitant to switch to other models... all the prompts have been tuned for OpenAI.”

If you’re building complex autonomous workflows, switching costs rise. The R&D approach works best for tool-level experimentation, not deeply integrated agentic systems.

Regulated industries need more structure.

Healthcare, finance, and government may require the “controlled autonomy” approach rather than “guarded freedom.” The traffic light system still works, but the green zone shrinks.

This doesn’t solve the ROI measurement problem.

Budget flexibility helps, but you still need to define success metrics before deployment. The 97% who can’t prove value aren’t failing because of tool selection. They’re failing because of measurement.

Shadow AI is a security risk, not just a governance inconvenience.

The $670K breach cost premium is real. This framework helps, but it’s not a substitute for proper security controls.

So What Now?

Stop asking “which AI tool should we use?”

Start asking:

• Which use cases would benefit most from AI?

• What infrastructure makes experimentation safe?

• How do we measure success in 90 days, not 3 years?

The organizations winning with AI aren’t the ones who picked the “right” tool. They’re the ones who built the right conditions for learning.

Your CRM required a 3-year roadmap. Your AI portfolio requires a 90-day pilot cycle.

That’s not a limitation. That’s your advantage.

P.S. If you’re already dealing with shadow AI (and statistically, you are: 43% of your employees have shared sensitive data without approval), start with the audit. You can’t govern what you can’t see.

And if you want to share this with your leadership team, here’s the executive summary: Budget like R&D, not software licenses. Reserve 15-20% for experimentation. Measure in 90-day cycles. Put four things in place: a 2-page policy, data classification, 30-minute training, and technical guardrails that enable rather than block.

That’s it. That’s the strategy.

Try This Prompt

Create an AI Acceptable Use Policy

For ChatGPT/Claude:

You are tasked with generating an AI Acceptable Use Policy document for an organization. The target audience is all employees, so the tone should be clear, concise, and easily understandable. The document should be formatted for readability, using a heading structure to delineate key sections. The final document should be approximately 1-2 pages in length. Use Markdown formatting.

Begin with a level 1 heading (H1) for the policy title: "AI Acceptable Use Policy." Follow this with a brief introductory paragraph explaining the purpose of the policy: to ensure responsible and ethical use of AI tools within the organization, to protect sensitive data, and to comply with relevant regulations.

The first major section, under a level 2 heading (H2) titled "Approved AI Tools," should list the AI tools currently approved for use by employees. [Illustrative Example: Approved tools include Grammarly for grammar checking, Otter.ai for transcription, and internal AI-powered data analysis dashboards]. For each approved tool, include a brief description of its intended use. State clearly that only approved tools are permitted for company-related tasks unless explicit approval is granted.

The second section, under a level 2 heading (H2) titled "Data Usage Guidelines," outlines what data can and cannot be used with AI tools. Emphasize the importance of protecting sensitive and confidential information. Provide specific examples of prohibited data types. [Illustrative Example: Do not input Personally Identifiable Information (PII) such as customer names, addresses, social security numbers, or financial data into any AI tool unless it is explicitly approved for handling such data and has appropriate security measures in place. Company confidential information, such as trade secrets, financial projections, and unpublished product roadmaps, are also prohibited]. Include a statement that employees are responsible for ensuring that all data used with AI tools complies with the company's data privacy and security policies.

The third section, under a level 2 heading (H2) titled "Human Review Requirements," details the requirements for human oversight of AI-generated outputs. Explain that AI outputs should not be treated as definitive and always require human review for accuracy, completeness, and appropriateness. Specify situations where human review is especially critical. [Illustrative Example: AI-generated content intended for external communication, such as marketing materials or customer support responses, must be reviewed by a human editor before publication. AI-driven analysis used for decision-making, such as sales forecasts or risk assessments, must be validated by a subject matter expert]. State that the level of human review should be commensurate with the risk associated with the AI's output.

The fourth section, under a level 2 heading (H2) titled "Requesting Approval for New Tools and Use Cases," describes the process for employees to request approval for new AI tools or novel applications of existing tools. Provide clear instructions on how to submit a request, what information to include, and who is responsible for reviewing and approving the request. [Illustrative Example: Requests should be submitted to the IT department via email at AI-Requests@example.com and include a description of the tool, its intended use, the data it will access, and the potential risks and benefits. The IT department will review the request in consultation with the Legal and Security teams]. Include an estimated turnaround time for review. [Illustrative Example: The review process typically takes 5-7 business days].

Conclude with a level 2 heading (H2) titled "Policy Violations." This section should clearly state the consequences of violating the AI Acceptable Use Policy, which may include disciplinary action, up to and including termination of employment. Emphasize that all employees are responsible for adhering to this policy and reporting any suspected violations.

For Perplexity:

Research the Full Policy

What are the essential components of an AI acceptable use policy for organizations in 2024-2025? Include examples from companies that have published their policies and any regulatory requirements.

Research: Approved Tools Section

What AI tools are commonly approved for enterprise use in 2024-2025 and what criteria do IT departments use to evaluate and approve AI tools for employee use?

Research: Data Usage Guidelines

What types of data should be prohibited from AI tools in corporate policies? Include examples of data breaches or incidents caused by employees inputting sensitive data into AI systems.

Research: Human Review Requirements

What are best practices for human oversight of AI-generated content in business settings? Include guidance on when human review is legally or ethically required.

Research: Approval Process

How do organizations structure the approval process for new AI tools? Include turnaround times, stakeholders involved, and risk assessment frameworks used by IT and legal teams.

Research: Policy Violations

What are typical consequences for AI policy violations in corporate settings and how do companies handle employee misuse of AI tools?

The Dare

Pick one pillar from the framework. Just one. Before the end of this week, take a single action on it:

• Policy: Draft a one-paragraph AI use statement

• Data Classification: List your five most sensitive data types

• Training: Schedule 30 minutes to watch one AI capabilities video with your team

• Guardrails: List what AI tools employees are already using

One pillar. One action. This week.

Good Luck - Dan

Share

You see the headlines. “We automated 80% of our workflow.” “AI agents handling entire departments.” And you wonder if you’re falling behind.

You’re not. You’re playing a different game.

I’ve been watching people torture themselves over this. They see the enterprise case studies. The “we automated 80% of our workflow” headlines. The LinkedIn posts about agents handling entire departments. And they feel behind.

They’re missing something important: enterprise and individuals aren’t playing the same game. Comparing your AI usage to enterprise patterns is like comparing your weekend runs to an Olympic training program. You’re not behind. You’re playing a different sport.

Anthropic just dropped their fourth Economic Index. Two million conversations analyzed. I went through the data so you don’t have to.

What you’ll walk away with:

1. The 77% vs 52% split that explains why you feel behind

2. The two games (and how to know which one you’re playing)

3. Why iteration isn’t a phase to rush through

4. The signals that tell you when to shift from iteration to automation

5. The one question to ask before you automate anything

The Reframe Nobody’s Talking About

The data tells a story most AI coverage ignores.

On Anthropic’s API (where businesses build), 77% of usage is automation. Full task handoff. “Here’s the input, give me the output, I’ll check it later.”

On Claude.ai (where individuals work), 52% of usage is augmentation. Collaboration. Iteration. “Help me think through this.”

These aren’t the same activity. They’re not even the same goal.

Old ThinkingNew Thinking”I should be automating more”“Am I playing the right game for my stage?”“Enterprise automates 77%. I’m behind.”“Enterprise has playbooks. I’m writing mine.”“Iteration is the warm-up”“Iteration is the strategy”

Enterprise has playbooks. You’re writing yours. Different game, different rules.

Why This Matters More Than You Think

This is where it gets interesting. MIT just released a study that should make every enterprise leader nervous: 95% of corporate AI pilots fail to deliver measurable impact. Only 5% of custom enterprise AI tools reach production.

The models work fine. The learning gap is the problem.

When MIT researchers dug into why, they found a pattern. Companies rush to automate without understanding their own workflows. They skip the iteration phase. They hand off tasks before they know what “done” looks like.

Sound familiar?

The researcher who led the study put it bluntly: “Generic tools like ChatGPT perform well for individual users due to their flexibility, but they often struggle in enterprise environments because they don’t learn from or adapt to specific workflows.”

In other words: enterprises that skip augmentation and jump straight to automation are failing at a 95% clip. Meanwhile, individuals who take time to iterate are building something enterprises struggle to buy: judgment about what’s worth automating.

The Two Games Framework

Making this concrete.

Game 1: Enterprise Automation (77% of API usage)

Businesses deploy AI on well-defined, repeatable tasks at scale. They’ve already figured out what “done” looks like. The goal is efficiency on known workflows. The playbook exists. They’re running it.

When it works (enterprise): Air India built an AI assistant handling 4 million+ customer queries with 97% full automation. But they started with a specific constraint (contact center couldn’t scale with passenger growth) and a clear definition of success (handle routine queries in four languages).

When it works (service business): A mid-size HVAC company implemented an AI chatbot for after-hours calls. Before: they missed 40% of calls that came in outside business hours. After: the chatbot captures lead information, books appointments, and answers common questions like “do you service my area?” Result: 23 additional booked jobs in the first month, zero missed leads. The key? They didn’t automate everything. They automated the specific task they understood completely: “capture the lead, book the appointment, answer the FAQ.”

When it fails: Nike spent $400 million on supply chain automation in 2000. The system couldn’t forecast demand correctly. Lost $100 million in sales. Stock dropped 20%. It took six years and mandatory 140-180 hours of training per employee to recover.

The difference? Air India and the HVAC company knew exactly what “done” looked like before they automated. Nike didn’t.

Game 2: Individual Iteration (52% of Claude.ai usage)

Individuals are learning, testing, validating. They’re figuring out what AI can do for their work. The goal is capability-building, not task completion. The playbook doesn’t exist yet. You’re writing it.

This isn’t a lesser game. According to Upwork’s research, 71% of freelancer AI use is augmentation, not automation. These are professionals who’ve figured out the model: iterate first, automate later.

The key insight:

Iteration isn’t a phase you rush through to get to automation. It’s how you build the judgment to know what’s worth automating in the first place.

The Data That Backs This Up

The numbers:

• 77% of enterprise API usage is automation

• 52% of consumer Claude.ai usage is augmentation (learning, iterating, validating)

• 71% of freelancer AI use is augmentation (Upwork, 2025)

• 95% of enterprise AI pilots fail to deliver measurable impact (MIT, 2025)

• 37% productivity gain when AI augments writing tasks (Gallup/academic research)

One pattern should jump out: individuals and freelancers who lean into augmentation are getting real productivity gains. Enterprises that rush to automation are failing at historic rates.

The Anthropic data shows something else worth noting. Augmentation was 55% in January 2025. A year later, it’s 52%. It’s held steady while automation crept up slowly. The shift toward augmentation was driven by task iteration, not passive learning. People are actively collaborating, not just asking questions.

McKinsey’s 2025 research confirms the pattern: organizations seeing significant ROI from AI are twice as likely to have redesigned workflows before selecting modeling techniques. They iterated on the process before automating it.

The One Question That Changes Everything

Now the practical piece.

Before you automate anything, ask yourself:

“Do I know exactly what ‘done’ looks like for this task?”

• If yes: Automate. You’ve iterated enough to define the output.

• If no: Iterate. Use AI to think through the problem, not hand it off.

This sounds simple. It isn’t. Hershey’s learned this the hard way in 1999. They rushed an ERP implementation to beat a Y2K deadline. Recommended timeline: 48 months. Actual timeline: 30 months. They went live in October, right before Halloween, their biggest season.

Result: $100 million in unfulfilled orders for Hershey’s Kisses and Jolly Ranchers. 19% drop in quarterly profits. 8% stock decline.

The same pattern repeats with AI. Companies that skip the “do I know what done looks like” question pay for it. The 95% failure rate comes down to judgment, not technology. And judgment comes from iteration.

The Iteration Stack

If you’re in Game 2 (and most individuals are), productive iteration looks like this:

1. Learning — “Explain this to me”

You’re building mental models. Understanding what AI can and can’t do. This is where most people start, and there’s nothing wrong with staying here until you’re ready.

2. Task Iteration — “Help me work through this”

This is where capability builds. You’re collaborating on real work, seeing where AI adds value, discovering your own patterns. This is what Anthropic’s data shows people actually doing.

3. Validation — “Check my thinking on this”

You’re using AI to stress-test your own work. This is the move that builds the judgment to eventually automate. When you can predict what AI will catch and miss, you know the task well enough to hand it off.

These aren’t training wheels. They’re the moves that build the judgment automation requires.

Signs You’re Ready to Graduate

For those who’ve been iterating for a while: how do you know when it’s time to automate?

You’re ready when:

• You can describe “done” in one sentence without hedging

• You know the edge cases before AI surfaces them

• Your prompts have stabilized — you’re not rewriting them every time

• You can predict when AI will fail and have a backup plan

• The task feels boring to collaborate on because you’ve done it so many times

The plateau trap:

Some people iterate forever. They’ve done the same task with AI 50+ times and still won’t automate. This isn’t thoroughness — it’s avoidance. If you can write the prompt from memory and predict the output within 90% accuracy, you’re ready. The remaining 10% is what verification steps are for.

What separates good iterators from great ones:

Good iterators collaborate with AI. Great iterators build systems.

That means:

• Documenting prompts that work

• Creating templates for repeatable tasks

• Building verification checklists

• Teaching others what you’ve learned

The goal isn’t to iterate forever. It’s to iterate until you have a playbook worth automating.

When This Frame Doesn’t Apply

The limits:

• If you’re building AI-powered products at scale, you are playing the enterprise game. Different rules apply.

• If you’ve iterated on the same task 50 times and still haven’t automated, you might be overthinking it. At some point, “done” becomes clear.

• This data is from Claude specifically. Other models may show different patterns.

• If you’re solving voice/tone problems, that’s a different challenge than automation vs. iteration. You need style iteration, not this framework.

• If you’re navigating AI disclosure with clients, that’s an ethics question, not an automation question. Different playbook.

The point isn’t “never automate.” The point is: know which game you’re playing before you judge your progress.

The Historical Pattern You Should Know

Every major technology adoption follows this curve. ERP systems in the 1990s had failure rates of 55-75%. The companies that failed? They rushed implementation, skipped testing, didn’t train their people.

Lidl spent €500 million on an SAP implementation over seven years before abandoning it entirely. The problem? They tried to move fast despite a fundamental mismatch in how the system worked versus how their business operated.

Gartner now reports that only 1% of executives consider their AI rollouts mature. AI has entered what they call the “Trough of Disillusionment.” The hype promised automation everywhere. The reality requires iteration first.

The winners in every technology wave are the ones who take time to learn the technology before betting the house on it. AI is no different.

Your Move

Pick one task you’ve been thinking about automating. Before you build the automation, spend 30 minutes iterating on it with AI. Document what “done” looks like. Write down the edge cases. Note where your judgment still matters.

If you can define “done” clearly after that session, automate away. If you can’t, you just saved yourself from joining the 95%.

The fastest way to fail at automation is to skip the augmentation phase. The fastest way to succeed is to know which game you’re playing.

Good Luck - Dan

Share

Bonus: Which Game Are You Playing? (Self-Diagnostic)

Answer these five questions to diagnose your current position:

1. Can you describe “done” for this task in one sentence?

• Yes, clearly

• Sort of, with caveats

• Not really

2. Have you done this task with AI at least 10 times?

• Yes, many times

• A few times

• Just starting

3. Do you know the edge cases where AI fails on this task?

• Yes, I can list them

• I’ve seen some failures

• Not sure yet

4. Has your prompt stabilized, or do you rewrite it each time?

• Stable — I use roughly the same prompt

• Evolving — I’m still refining

• Different every time

5. Could you teach someone else how to do this task with AI?

• Yes, I could write the guide

• Probably, with some prep

• Not yet

Scoring:

• Mostly first options: You’re ready to automate. Build the workflow, add verification, and hand it off.

• Mostly second options: Keep iterating. You’re close but not quite ready.

• Mostly third options: Stay in learning mode. Iteration will pay off — don’t rush.

This Week in 30 Seconds

The AI free lunch is over. Every major story this week pointed to the same uncomfortable truth: the “try it free, figure out costs later” era has ended. You’ll pay with your wallet, your privacy, your existing software investments, your time, or your risk exposure. The winners in 2026 won’t be those who adopted AI fastest. They’ll be the ones who understood what they were actually paying.

Five stories this week reveal the same truth from different angles: AI is now demanding payment, and the currency varies by vendor. OpenAI wants your attention. Google wants your data. Your software vendors are scrambling to protect their margins. Workday’s research shows that productivity gains require discipline to capture. And risk analysts just moved AI to #2 on the global threat list. The free trial is over. Here’s what you’re actually signing up for.

Short on time? Jump to Story 4 (The Productivity Paradox) for the most immediately actionable framework.

Story 1: The End of Free ChatGPT (As You Knew It)

The News: OpenAI announced ads are coming to ChatGPT. Free users and $8/month “ChatGPT Go” subscribers will see sponsored content at the bottom of responses within weeks. CEO Sam Altman, who called ads “a last resort” in May 2024, wrote: “A lot of people want to use a lot of AI and don’t want to pay.” Paid tiers (Plus at $20/month, Pro, Business, Enterprise) remain ad-free.

The Noise: “OpenAI sold out!” “This destroys trust!” “It’s just like Google all over again!”

The Signal: Forget the betrayal narrative. This is math, plain and simple.

Look at the numbers: OpenAI has 800 million weekly active users. Only 5% pay anything. That’s 760 million people using infrastructure that costs billions to run. The company burned through more than $8 billion in 2025 while generating $13 billion in revenue. Those margins don’t work without either (a) raising prices, (b) adding revenue streams, or (c) watching the company collapse.

OpenAI chose option B. The ads themselves are a footnote. Free AI was never sustainable. Every “free” AI tool you’re using right now has a business model problem it hasn’t solved yet. OpenAI just solved theirs in public. The rest will follow, each in their own way: ads, price hikes, feature restrictions, data monetization, or shutdowns.

The lesson for operators: If you’ve built workflows on free AI tools, you’ve built on borrowed time. Call it paranoia if you want. I call it watching how businesses work.

Your Move: Audit your AI tool usage this week. Make a list with two columns: “Paid” and “Free.” For every tool in the free column, write down what happens to your workflow when the business model changes. For anything critical, upgrade to paid or find an alternative now, before the change happens.

The Math That Matters:

• 2.5 billion prompts submitted to ChatGPT daily

• Each prompt is now a potential ad impression

• At even modest CPMs, that’s a billion-dollar revenue stream

• Your attention is the product

Story 2: Google Wants Your Entire Digital Life (In Exchange for “Personal Intelligence”)

The News: Google launched “Personal Intelligence” in beta. It connects Gemini to your Gmail, Google Photos, YouTube history, Search, Maps, and more. The AI can now “reason across your data to surface proactive insights.” Opt-in only, off by default, available first to AI Pro and Ultra subscribers in the US. Google promises it won’t train directly on your inbox or photos, but will use “limited info, like specific prompts and responses” to improve functionality.

The Noise: “Finally, an AI that actually knows me!” on one side. “This is a privacy nightmare!” on the other. Neither captures what’s actually happening.

The Signal: This is the most important AI moat story of 2026. Google just showed everyone their hand.

The winning AI isn’t the smartest model. It’s the one with the deepest context. And Google has context on 2.5 billion Gmail users and 1.5 billion Google Photos users. Call it a feature advantage if you want, but it’s really a data moat that no competitor can replicate.

What the privacy discourse misses: Personal Intelligence creates ecosystem lock-in more powerful than any feature Google has ever shipped. Once Gemini knows your email patterns, your photo memories, your search habits, your calendar rhythms, switching to Claude or ChatGPT means starting over with a stranger. All that context? Gone.

Google is betting that personalization beats performance. They might be right.

For SMB operators, this raises a strategic question that goes beyond privacy concerns: Do you consolidate into Google’s ecosystem to maximize AI capability? Or do you stay distributed to avoid dependency? Both have costs. Neither is free.

If you’re running on Google Workspace, Personal Intelligence could genuinely make your team more effective. But enabling it means feeding Google your business communications, client relationships, and operational patterns. The “won’t train on your inbox” promise is carefully worded. Prompts and responses are still collected.

Cross the privacy Rubicon, and you’re making a competitive positioning decision.

Your Move: Make an ecosystem decision. Are you a Google shop, a Microsoft shop, or deliberately multi-platform? The “personal AI” features rolling out in 2026 will reward ecosystem commitment and punish fragmentation. Pick a lane and commit. Or accept that you’ll get generic AI while your competitors get personalized intelligence.

What Google Actually Said:

• VP Josh Woodward: “Gemini now understands context without being told where to look”

• Google Photos data used to “infer your interests, relationships to people in your photos, and where you’ve been”

• Google acknowledges the AI may “struggle with timing or nuance, particularly regarding relationship changes, like divorces”

• Rolling out to free tier and more countries “later”

Story 3: Your Software Vendors Are Scared (And You Should Pay Attention)

The News: Anthropic launched Cowork, a computer-use tool built entirely by Claude Code in under 1.5 weeks. The announcement spooked Wall Street: software stocks including Salesforce and Workday dipped. RBC analysts questioned whether traditional software can “defend pricing power” as AI capabilities expand. The implicit question: If AI can do what your $50K/year enterprise software does, why are you still paying $50K?

The Noise: “AI is finally coming for software!” says one camp. “SaaS is dead!” “This is completely overhyped, enterprise software isn’t going anywhere,” says the other. Both miss what’s actually happening.

The Signal: AI isn’t replacing your software stack this year. Maybe not even next year. But it IS giving your vendors an existential crisis. And companies in existential crisis mode do weird things.

Watch for these moves in 2026:

• Price increases disguised as “AI upgrades” (paying for the R&D to save their business)

• Forced bundling of AI features you didn’t ask for

• Aggressive lock-in tactics (new contract terms, harder data exports)

• Sudden pivots that break your workflows

• Acquisitions that change product direction overnight

Keep your software stack. But watch your vendors closely. Are they integrating AI defensively (checking a box) or offensively (actually improving the product)? Are they raising prices because they’re scared or because they’re delivering more value? The answer matters for your renewal negotiations.

The hidden opportunity: vendor fear creates negotiating leverage. When Salesforce is worried about Claude taking their market, they’re more likely to cut you a deal to keep you locked in. That window won’t stay open forever.

Your Move: Make a list of your top 5 software costs. For each one, answer: “What would it take for AI to replace this in our specific workflows?” If the answer is “a lot” (complex integrations, industry-specific requirements, team training investment), you’re probably safe. If the answer is “not much” (it’s basically fancy spreadsheet work, or templated processes), start exploring alternatives now. Don’t wait for your vendor’s pricing to reflect their panic.

The Numbers Behind the Fear:

• Anthropic built Cowork’s code “entirely by AI” in less than 10 days

• 64.3% of global VC deal value in 2025 went to AI-related investments

• AI workflow market estimated at $65B in 2025, scaling to $190B by 2030

• That’s $125B of new market, much of it taken from existing software spend

Story 4: The Productivity Paradox (AI Giveth and AI Taketh Away)

The News: Workday surveyed 3,200 employees across global enterprises and found what they’re calling a “productivity paradox.” While 85% of employees save 1-7 hours per week with AI, nearly 40% of those savings are lost to rework. Fixing mistakes. Rewriting content. Verifying outputs. Only 14% consistently get positive net outcomes from AI. Meanwhile, 32% of companies simply pile more work onto employees instead of reinvesting the time saved.

The Noise: “AI productivity is a myth!” “See, I knew it was overhyped!” “We just need better AI tools!” All three reactions miss the point entirely.

The Signal: This is THE story of the AI transition: the management challenge, not the tools or capabilities.

Everyone assumed the equation was simple: AI equals same work, less time. The Workday data shows reality is messier. The actual equation: AI equals faster work, plus new work (fixing AI outputs), plus more work (because you’re “faster now” so here’s more to do).

The time savings are real. 85% of employees genuinely save 1-7 hours weekly. But what happens next determines whether you win or lose.

Three patterns emerged from the research:

The Winners (14%): Reinvest saved time into higher-value work. 57% of this group uses AI-freed hours for deeper analysis, strategic thinking, and creative work. They treat AI time savings as an investment, not a windfall.

The Treaders (54%): Break even. Time saved roughly equals time spent fixing and verifying. They’re running faster but not getting anywhere.

The Losers (32%): Company just piles on more tasks. “You saved 5 hours? Great, here’s 5 more hours of work.” The treadmill gets faster.

The uncomfortable truth: AI doesn’t automatically make you more productive. It makes you faster at producing things that might need fixing. The productivity comes from how you manage that speed.

One more data point worth noting: 79% of employees who consistently get positive AI outcomes had skills training. The 21% who didn’t train are disproportionately represented in the “rework loop” group. Training makes the difference between ROI and expensive experimentation.

Your Move: Before deploying AI on any workflow, answer one question: “When AI saves time, where does that time go?” If the answer is “more of the same work,” you’re building a faster treadmill. Define the reinvestment strategy before you save the first hour. What specific higher-value work will fill that time? Name it. Assign it. Measure it.

Try This Prompt:

For ChatGPT/Claude:

Analyze my team's current workflow for [specific process]. Identify:
1. Tasks where AI could save time (estimate hours/week)
2. Common failure modes that would require human review/fixing
3. Higher-value activities we could reinvest saved time into

For each AI opportunity, estimate the realistic NET time savings after accounting for rework. Be conservative.

For Perplexity:

What does research show about the actual productivity gains from AI adoption in [your industry]? Include studies that measured both time saved and time spent on rework/verification. What patterns separate companies that achieved positive ROI from those that didn't?

The Numbers That Matter:

• 85% save 1-7 hours/week with AI

• 40% of savings lost to rework

• Only 14% consistently see positive net outcomes

• 77% of daily AI users review AI output as carefully (or more carefully) as human work

• Employees aged 25-34 bear the biggest rework burden (46% of highest-rework group)

• 79% of successful AI users had skills training

Story 5: The Honeymoon Is Over (AI Is Now a Top Business Risk)

The News: AI jumped from #10 to #2 in Allianz’s annual global business risk survey. That’s the biggest single-year jump in the survey’s 14-year history. The World Economic Forum’s Global Risks Report 2026 echoed the concern, flagging AI’s downside potential alongside tariffs as top threats. 32% of respondents now rank AI among their top business risks.

The Noise: “AI doom is overblown!” says the techno-optimist camp. “The risk-industrial complex is just fear-mongering!” Meanwhile, the AI skeptics say: “Finally, people are waking up to the dangers!” Neither reaction is useful.

The Signal: The irony that everyone’s missing: The same boards approving AI budget increases are simultaneously ranking AI as their second-biggest risk.

That’s maturity.

The hype phase is over. What changed between 2025 and 2026? Real deployments created real problems. Hallucinations in customer-facing tools. Compliance failures from AI-generated content. IP exposure from training data. Employee productivity claims that didn’t survive audit. The theoretical risks became line items. Companies learned that “AI can do amazing things” and “AI can create serious problems” are both true at the same time.

For SMB operators, this shift is actually good news. The conversation is moving from “adopt AI or die” to “adopt AI carefully or die.” That’s a more honest conversation. You now have permission to ask hard questions about AI risk without sounding like a luddite who doesn’t get it.

The companies doing AI well in 2026 are also the companies thinking about AI risk. Same strategy, two sides.

Three risk categories to watch:

Operational Risk: AI failures that disrupt your business. The customer service bot that goes rogue. The content generator that creates something embarrassing. The automation that breaks in ways humans wouldn’t.

Legal Risk: IP exposure from training data. Compliance failures from AI-generated documents. Liability questions when AI makes decisions. Regulatory scrutiny that’s ramping up globally.

Reputational Risk: The public AI mistake that trends on social media. The bias incident that damages your brand. The “we trusted AI and it failed us” story that erodes customer confidence.

Your Move: Add “AI risk” to your next leadership discussion. Three questions to answer: (1) Where are we using AI with customer-facing output? (2) What happens if that output is wrong or offensive? (3) Who owns AI risk in our organization? If you can’t answer all three clearly, you have homework before your next board meeting.

The Risk Numbers:

• AI jumped from #10 to #2 in Allianz risk rankings (biggest jump ever)

• 32% of executives now rank AI among their top business concerns

• WEF Global Risks Report flags AI alongside tariffs as top 2026 concerns

• The same boards approving AI budget increases are flagging AI as top risk

The Pattern

Five stories. One theme: AI is demanding payment.

OpenAI needs your attention (or your money). Google needs your data. Your software vendors are scrambling because AI threatens their margins. The productivity gains require discipline to capture. And the risk profile has changed entirely.

None of this makes AI bad. It makes AI a serious business decision instead of a shiny experiment. The free trial is over. What comes next is the hard work of understanding what you’re trading, and deciding whether it’s worth it.

The Contrarian Corner

The dominant narrative is still “AI adoption is the priority.” Move fast. Don’t get left behind. The companies that adopt fastest will win.

But look at what actually happened this week.

OpenAI needs ads because subscriptions aren’t enough. Google needs your entire digital life for AI to be useful. Workday found that productivity gains evaporate into rework when you’re not intentional. Risk analysts moved AI to #2 on their threat list.

The contrarian take: AI intentionality beats AI adoption.

The winners in 2026 won’t be the companies that adopted AI fastest. They’ll be the companies that understood what they were trading for AI capability and decided, deliberately, that it was worth it.

Adoption without intentionality is just expensive experimentation. And with the free lunch ending, that experimentation just got a lot more expensive.

Good Luck - Dan

Nobody is teaching the skill that matters most right now: how to build delegation fluency while AI is still learning. Here’s the framework for practicing at 50% reliability so you’re ready when 40-hour horizons arrive.

We’ve been thinking about AI adoption all wrong.

Most people are stuck asking “which model?” or “when will this be ready?” Those questions assume the constraint is tool maturity. The actual constraint is how long it takes YOU to build the skills that let you work with imperfect agents effectively.

And those skills? They take months to develop. They compound. And the gap between someone who started building them six months ago and someone starting today isn’t something you close by reading one more article or taking one more course.

I keep a failure mode log. Over six months, I’ve logged 47 professional-looking errors that would have shipped if I hadn’t built verification systems. Each one taught me something about what “wrong” looks like when it’s dressed up as “right.” That pattern recognition? You don’t build it overnight. You build it through practice reps over months.

Here’s what we’re covering:

1. Why the “I’ll catch up later” mindset just broke (and why nobody’s teaching what actually matters)

2. The Three Skills That Compound: the framework nobody’s systematically teaching but everyone needs

3. How to practice delegation at 50% reliability without creating expensive failures

4. Copy-paste frameworks for your first practice reps (delegation template, verification checklist, spec-writing assistant)

5. The Delegation Maturity Stages: where you are and what’s next

6. Why domain expertise transforms rather than disappears—and what that transformation actually looks like

7. When NOT to use this approach (honest limitations section)

The Framework Nobody’s Teaching

Here’s what I’ve learned from six months of building delegation skills in real time: three capabilities that get more valuable—not less—as agents grow more capable.

1. Specification: Learning to Write Tasks Agents Can Actually Execute

Writing a clear task spec feels like over-engineering until you realize the alternative: delegating vaguely, getting mediocre output, and either accepting subpar work or spending twice as long fixing it.

Specification is the skill of defining tasks precisely enough that an agent knows what success looks like without constant guidance. What does “good” look like for this task? What common failure modes should it watch for? What context does it need that you’re assuming but haven’t stated?

I started tracking my own delegation attempts against capability benchmarks in January. The METR data shows the 50% success horizon (tasks agents can complete with human-level performance half the time) sits at roughly five hours today. But my early attempts at five-hour tasks? They failed constantly. Not because the capability wasn’t there - because my specifications were garbage.

The clarity you develop specifying shorter tasks transfers directly to longer initiatives. Someone who’s been practicing specification for six months has intuition about what details matter. They know which assumptions to make explicit, which constraints to call out, which examples clarify versus confuse.

Someone starting fresh doesn’t have that pattern recognition yet. They’ll learn it through practice, but they’re starting the learning curve exactly when complexity spikes.

2. Verification: Building Systems to Catch Failures That Look Like Successes

At 50% reliability, professional-looking errors are common. Last week I delegated a competitive analysis. The agent produced a beautiful 12-page deck with proper citations, clean formatting, executive summary. Professional. I almost shipped it.

Then I spot-checked three citations. Two were from cached pages 14 months old. One was a different company with a similar name. That’s what “professional-looking failure” means.

Verification is the skill of checking output efficiently and catching the failures that don’t announce themselves. What makes output “wrong” even when it looks right? What claims need fact-checking? What feels “too clean” for the task complexity?

The verification systems you build at current horizons become more valuable as horizons extend. You’re developing an eye for what professional-looking errors look like in your domain. That eye only sharpens with practice.

3. Intervention: Recognizing When to Pull the Emergency Brake

Some delegation runs go sideways. The agent gets stuck in a loop. It hallucinates confidently. It pursues the wrong interpretation of an ambiguous spec. The question is whether you notice early enough to intervene before it wastes hours or produces work that needs complete redoing.

Intervention is pattern recognition about what “stuck” looks like, what “hallucinating confidently” looks like, what “wrong interpretation” looks like before the outputs make it obvious. This intuition only comes from running enough tasks to see the failure modes repeatedly.

A client started practicing delegation in March. By September, she could delegate three-day research projects and catch failures in 20 minutes of review. Her peer started in August - same company, same tools. The peer is still learning to specify two-hour tasks while my client handles multi-day initiatives. The learning curve is the same for both. The starting point is very different.

How to Practice at 50% Reliability Without Creating Expensive Failures

The practical reality: you need to operate in the 50%/80% gap. The 50% horizon (tasks agents can complete half the time) sits at roughly five hours based on METR benchmarks. The 80% reliability horizon (tasks you can actually depend on without extensive verification) sits at roughly 27 minutes.

That gap is where you build the skills that scale.

Start with tasks where verification is fast and failure is cheap. Two-hour research tasks. Document drafting. Competitive analysis. Work where “wrong” is visible quickly and fixing costs minutes, not hours.

Delegate at 50% reliability but build verification systems to catch the failures. Invest time learning what professional-looking failures look like in your domain. Practice specification by writing task definitions clear enough for an agent to execute without constant guidance.

Track your pattern recognition. What kinds of errors show up repeatedly? What makes output “wrong” even when formatting looks right? Which specification details turn out to matter? Which assumptions can you safely make?

Each rep at 50% reliability is a data point. Each failure you catch builds your eye for what “wrong” looks like. Each specification you write sharpens your sense of what details agents need.

The research backing this up: METR data shows this pattern clearly—at 50% success rate, horizons reach nearly 5 hours, but at 80% reliability, they drop to just 27 minutes. That’s an 11x gap between “possible” and “dependable.” The pattern is real, measurable, and it’s where you learn.

The goal isn’t perfect delegation today. The goal is building intuition at current horizons so you have months of compounding pattern recognition when longer horizons become possible.

Copy-Paste Frameworks for Your First Practice Reps

I’ve built three tools that lower activation energy for starting. Use them, modify them, make them your own.

Delegation Framework Template

TASK DELEGATION SPEC

Task Goal:
[What does success look like? One sentence.]

Context the Agent Needs:
- [Background information that's obvious to you but not to the agent]
- [Relevant constraints or requirements]
- [Who the audience is and what they care about]

Success Criteria:
- [Specific, measurable definition of "done"]
- [What quality looks like for this output]

Common Failure Modes to Watch For:
- [What typically goes wrong with this kind of task?]
- [What should the agent double-check?]

Output Format:
[Exactly how the final output should be structured]

Verification Questions I'll Ask:
- [What will I check before trusting this output?]
- [What would a professional-looking error look like here?]

Verification Checklist

Before trusting any agent output, run through these questions:

• Does this output actually match the spec I provided?

• Are there factual claims I should verify independently?

• Does this feel “too clean” or “too perfect” for the task complexity?

• What would a professional-looking error look like here, and do I see any signs?

• If I were reviewing this work from a junior team member, what would I check?

• What assumptions did the agent make that I should validate?

• Are there edge cases or exceptions this output doesn’t address?

Spec-Writing Assistant Prompt

When you’re not sure what to specify, use this:

For ChatGPT/Claude:

I need help writing a clear task specification for an AI agent.

TASK: [Describe what you want done]

Help me identify:
1. What context the agent needs that I'm assuming but haven't stated
2. What "success" looks like specifically enough to verify
3. What common failure modes I should warn about
4. What questions I should ask to verify the output before trusting it

Format your response as a complete delegation spec I can copy-paste.

For Perplexity:

What are best practices for writing clear AI task specifications? Include: essential context requirements, success criteria definition, common failure modes to specify, and verification questions. Focus on delegation frameworks from 2024-2025.

The Delegation Maturity Stages

Understanding where you are helps you know what to practice next. Three stages I’ve observed in my own progression and in people I’ve worked with:

Note: You may be Stage 2 (Practitioner) for research tasks but still Stage 1 (Beginner) for customer-facing content. That’s normal — reliability thresholds and verification needs differ across contexts.

Stage 1: Beginner (Months 0-3)

You’re learning what agents can and can’t do. Most specifications are incomplete. Verification takes longer than the original task would have. You catch maybe 60% of failures before they become problems.

Focus at this stage:

• Build your failure mode log (what goes wrong repeatedly?)

• Practice specification on 1-2 hour tasks

• Over-verify everything (you’re building pattern recognition)

• Track what specification details actually mattered

Stage 2: Practitioner (Months 3-8)

You can delegate 2-5 hour tasks and catch most failures efficiently. Your specifications are tighter. You’re starting to recognize “stuck” or “wrong” before it produces bad output. Verification takes 10-20% of what the task would have taken.

Focus at this stage:

• Extend to longer tasks (pushing toward full-day delegation)

• Build intervention instincts (when to stop a run early)

• Develop domain-specific verification shortcuts

• Start documenting patterns for your team

Stage 3: Advanced (Months 8+)

You can delegate multi-day initiatives. Verification is fast because you know what to look for. You catch professional-looking failures that others would miss. Your domain expertise has transformed into verification and direction expertise.

Focus at this stage:

• Push boundaries on task complexity

• Build systems others can use

• Teach specification and verification to your team

• Identify which work should stay human vs. which to delegate

These timelines assume consistent practice. If you’re delegating one task per month, multiply by 3-4x. If you’re delegating daily, you can compress them.

The point: this is a skill development arc, not a switch you flip. Where you start on this arc matters more as capability horizons extend.

For managers: Use these stages to assess team capability. Your Stage 1 people need templates and over-verification support. Your Stage 2 people need intervention skill development. Your Stage 3 people can teach others.

How Domain Expertise Transforms (Not Disappears)

Some execution skills lose leverage when agents can do the work. Other skills - knowing what “right” looks like, catching errors before they ship - gain leverage.

Pretending otherwise would be dishonest.

But your domain knowledge isn’t obsolete. It’s just doing a different job now - verification instead of execution. The litigator who spent a decade learning contract law doesn’t lose value - they gain leverage. They can direct agents toward useful work and catch errors that a less experienced person would miss. Their pattern recognition about what “wrong” looks like in contracts becomes the control surface.

The engineer who knows a codebase intimately can verify agent output in ways someone loading context for the first time cannot. The researcher who understands methodology deeply catches the subtle errors in agent-generated analysis.

Domain expertise becomes the skill of knowing what correct looks like. And that skill becomes more valuable, not less, because it’s the bottleneck that determines whether delegation produces leverage or liability.

But that transformation isn’t automatic. You build it by practicing delegation at current horizons, learning where agents fail in your domain, and developing systems to catch those failures before they propagate.

When NOT to Use This Approach

Honest limitations:

Don’t start here if:

• You’re in a domain where errors are catastrophic and unrecoverable (medical diagnosis, legal filings, financial compliance)

• Your organization prohibits delegating work to AI systems due to data sensitivity

• You don’t have time to verify output carefully (delegation at 50% reliability requires verification)

• You need 95%+ reliability today (current horizons can’t deliver that for most complex work)

This framework works best for:

• Knowledge work where verification is fast and errors are fixable

• Domains where you have deep expertise to catch failures

• Tasks where “wrong” is visible before it ships

• Work environments that allow experimentation

The gap between 50% and 80% reliability is real. Operating in that gap requires verification systems and domain expertise. If you can’t verify efficiently, don’t delegate yet. Wait until reliability horizons catch up to your requirements.

Try This Framework This Week

Pick one 2-hour task you’d normally do yourself. Use the delegation template above to spec it clearly. Delegate it to an agent (Claude, GPT, whichever you have access to). Run it through the verification checklist. Notice what went wrong and what went right.

That’s rep one.

Do it again tomorrow. And the day after. Track what you learn. What specification details matter? What errors show up repeatedly? What makes output “wrong” even when it looks professional?

Six months from now, you’ll have pattern recognition someone starting fresh doesn’t have. Twelve months from now, you’ll be delegating work that today feels impossible to automate.

Or you can wait for clarity that won’t arrive until the learning curve gets steeper.

P.S. Specification. Verification. Intervention. Three skills that compound. Twelve months of practice. Your peers are building these right now while you’re waiting for clarity that won’t come.

Good Luck - Dan

This Week in 30 Seconds

Companies rushed to adopt AI in 2025. Now they’re discovering the painful truth: having AI and using it well are two completely different things. Five major research reports dropped this week showing 88% of organizations use AI but only 39% see business value. That’s a 49-point gap representing billions in wasted investment. Meanwhile, 25% of your search traffic is fragmenting to AI platforms, your employees expect ChatGPT-level AI at work, and VCs are quietly saying some companies use “AI investment” as PR cover for layoffs. The year ahead is about execution, not adoption.

Five stories this week tell the same story from different angles: the gap between AI adoption and AI value just became impossible to ignore. The data is brutal, the implications are strategic, and the companies getting it right are doing something fundamentally different than those who rushed in. Here's what matters.

Story 1: Companies Spent 2025 Adopting AI. They’re Not Getting Value.

The News: Five major research reports dropped this week showing organizations are failing to realize AI’s promised value. The numbers are brutal: 83% of nearly 500 organizations scored in the lowest two categories for AI maturity (Phenom study). Only 7% of organizations give employees any guidance on what to do with the time AI saves them (Gartner). And 88% of HR leaders say their organizations haven’t realized significant business value from AI tools (Gartner survey of 114 HR leaders).

The Noise: Everyone’s celebrating adoption rates and employee enthusiasm. Headlines scream about how many companies have deployed AI. Look at all this innovation.

The Signal: Companies are treating productivity gains as the finish line instead of the starting point. MIT’s Project NANDA (July 2025) analyzed over 300 public AI implementations and surveyed 153 senior leaders. Despite $30-40 billion in global enterprise investment, 95% of GenAI pilots fail to deliver measurable returns. Ninety-five percent.

Side note: Look, the 95% stat has its critics. Fair questions about methodology and what counts as "failure." But whether it's 95% or 80% doesn't really matter — the core issue is the same. Most organizations struggle to move AI pilots into production. That's the problem worth solving, regardless of the exact percentage.

The McKinsey data is equally sobering: 88% of organizations now use AI in at least one business function, but only 39% report measurable EBIT impact. That’s a 49-point gap between adoption and value. Project abandonment rates hit 42% in 2025, up from 17% in 2024—a 147% year-over-year increase in companies giving up on AI initiatives.

Why? Because you saved your team 1.5 hours a day with AI (the average reported by Gartner), but only 7% of companies tell employees what to do with that time. You’ve created expensive confusion, not business value. The gap runs between adoption and strategy, not between AI and humans.

Organizations with documented AI policies report 82.5% confidence using AI responsibly versus 58.5% at companies without policies. Yet only 37% have formal policies in place. You’re spending on tools while ignoring governance. That’s backwards.

Your Move: Audit before you expand. If you’ve adopted AI tools, pause new purchases and answer three questions:

1. Do we have written guidelines for AI use?

2. Can we measure actual business outcomes, not just “usage rates”?

3. Do employees know what to do with the time AI saves?

If you answered no to any of these, your next investment should be governance and strategy, not more subscriptions. The data is clear: Gartner predicted 60% of AI projects would be abandoned due to lack of AI-ready data. Don’t be part of that statistic.

Story 2: Most Small Businesses Are Failing at AI. Here’s Why.

The News: An analysis of 1,000+ AI training programs across UK and Ireland SMBs reveals the core failure pattern: businesses start with a cool AI tool and then try to find problems it can solve. Successful businesses do the opposite. They start with actual problems and determine if AI is the best solution. According to ProfileTree founder Ciaran Connolly, that sequence matters enormously.

The Noise: AI tools are getting better, cheaper, more accessible. Just adopt and iterate. Everyone’s doing it. The tools will save you time and money automatically.

The Signal: The data shows a stark divide: 83% of growing SMBs use AI versus stagnant and declining businesses. The successful 83% follow completely different patterns than the failures.

Four patterns keep showing up in failed implementations.

• First, the tool-first mistake: buying exciting AI without connecting it to actual workflows. You see a demo, get excited, purchase access, and then wonder why adoption is low.

• Second, the training gap: handing people tools without teaching them to use them well. You think the tool is “intuitive” but your team produces inconsistent outputs because they don’t understand effective prompting or quality review.

• Third, unrealistic expectations: thinking AI replaces human judgment instead of assisting skilled operators. AI doesn’t make bad writers good. It makes good writers faster.

• Fourth, integration challenges: standalone tools that don’t connect to existing systems create more friction than value.

Successful adopters achieve positive ROI within six weeks (Swfte AI research). They report 27% productivity increases and 23% cost reductions. And 91% report revenue growth (Salesforce research). But they get there by starting with problem identification, not tool selection.

The successful pattern looks like this: Pick one process that consumes significant staff time and produces variable quality. Customer email responses, meeting summaries, initial content drafts. Something contained. Define success criteria in advance (not “we’ll figure it out as we go”). Run a four-week trial. Review honestly. Did it hit the success criteria? If yes, expand systematically. If no, analyze why before trying again.

The entry cost is lower than you think. Swfte AI data shows $100-300/month starts viable AI implementation. The fastest ROI comes from customer service automation (40% of inquiries can be automated), not moonshot projects.

Your Move: Problem-first, tool-second. This week, document one process: high time consumption plus variable quality output. Write down what success looks like (specific, measurable). Don’t buy anything yet. Just establish the baseline. Next week, if you still think AI is the solution, then evaluate tools. That order matters enormously.

Story 3: Your Employees Are Using AI at Home. That’s About to Be Your Problem.

The News: 2026 is the turning point where consumer AI adoption starts forcing enterprise change. Matt Britton (Suzy founder) compares it to the early iPhone era: people used smartphones freely at home while companies resisted them at work. Eventually, consumer expectations forced organizational change. The same pattern is happening with AI. Your employees are already using AI for personal health, family support, finances, and everyday decisions, building real confidence through hands-on use, not theory.

The Noise: Enterprise AI adoption is the story. Companies deploying tools. IT departments evaluating solutions. Governance frameworks being developed.

The Signal: The real adoption is happening in living rooms, not boardrooms. And the numbers are staggering. ChatGPT alone has 800-900 million weekly active users across platforms (a16z, December 2025). That’s 66 million daily active users. Total AI tool usage hit 378 million people worldwide in 2025, up 64 million from 2024. In the U.S., 61% of adults use AI, with Gen Z at 76% and even Baby Boomers at 45%.

85% cite personal use as their primary AI application. They’re using it for health questions, financial decisions, travel planning, learning new skills. They trust AI with personal decisions because there are no approval layers slowing them down. They experiment freely. They learn what works. They develop real fluency.

And then they show up at work expecting the same speed, personalization, and capability they get at home.

This creates bottom-up pressure on organizations that are still debating governance frameworks. Your employees have already formed expectations about what AI should do and how it should work. If your workplace AI is clunky, restricted, or slower than what they use at home, you’ve created friction. The talent advantage goes to companies that reduce that friction, not add to it.

The BYOD moment for AI is here. Remember when employees started bringing their personal smartphones to work and IT departments scrambled to accommodate them? Same pattern. Menlo Ventures found 27% of all AI application spend now comes through product-led growth motions (individual users, not executive procurement). That’s nearly 4x the rate in traditional software.

Gen Alpha (entering the workforce soon) expects AI to understand context, respond personally, and evolve with them. That’s your new hiring baseline. Companies still running approval processes for basic AI use will lose talent to competitors who embrace it.

Your Move: Ask, don’t assume. This week, survey your team: What AI tools are you using in your personal life? What do you wish you could do at work that those tools enable at home? Make it anonymous if needed. You’ll learn more from ten honest answers than from any vendor pitch. Then ask yourself: are we adding friction or reducing it? The gap between consumer AI and enterprise AI is your talent risk.

Story 4: SEO Just Got 10x Harder. Here’s What Changed.

The News: Up to 25% of traditional organic search traffic could shift to AI chatbots and answer engines by 2026 (ProfileTree research). Google’s AI Overviews feature is already reducing click-through rates by 25-60% for many queries. Even businesses ranking in the top three positions are getting significantly fewer visitors than twelve months ago. The search landscape has fragmented: you now need to appear in AI Overviews, be cited by ChatGPT, show up in Perplexity results, and be referenced by Claude and Gemini.

The Noise: Keep doing traditional SEO. Optimize for keywords. Build backlinks. Rank on page one. Google’s still dominant.

The Signal: SEO just became ten times harder, and most businesses haven’t noticed yet. Over 30% of desktop searches on Google now surface an AI Overview, up from just 23% in April 2025 (Comscore 2025 AI Intelligence Report). That’s 30%+ growth in eight months. Different trackers report anywhere from 10-55% of queries showing AI-generated results. The variance doesn’t matter. The trajectory does.

Google’s AI Overviews can reduce organic clicks by 18-64% for affected queries (Exploding Topics analysis). That means even if you rank #1 for a valuable keyword, if Google answers the question in an AI Overview, your traffic drops by half or more.

But the change runs deeper than Google. When someone asks ChatGPT or Perplexity for a recommendation, the AI provides a direct answer and cites two or three sources. If you’re not one of those sources, you’re invisible to that user entirely. The game shifted from “rank in search results” to “be quotable and citeable by AI systems.”

This has created a new discipline called Generative Engine Optimization (GEO). The core principle: AI systems favor structured, authoritative, quotable content. Your vague marketing speak doesn’t get cited. Your “About Us” page written in corporate jargon doesn’t appear in AI answers.

Professional response is accelerating: 82% of enterprise SEOs plan to invest more in AI (Digitaloft 2025). And 86% of SEO professionals now use AI in their workflows (up 1,900% in search interest over five years). Companies with 200+ employees report 83% see improvements in SEO performance since using AI, saving an average 12.5 hours per week.

Your Move: Audit your AI search presence this week. Search for your business name, products, and services in ChatGPT, Perplexity, Google’s AI Overviews, and Claude. What comes up? Is it accurate? Is it current? Many businesses are shocked to find AI systems have outdated or wrong information. Then review your About Us and Why Choose Us pages. Are they written in a way AI can easily understand and quote? Or are they vague marketing speak? The businesses that win in 2026 will be the ones AI systems trust enough to cite.

Try This Prompt:

For ChatGPT/Claude:

You are an AI-powered Business Audit Assistant. Your primary task is to analyze how your client's business is represented across the internet and provide actionable recommendations for improvement.

Begin by gathering data. Perform comprehensive web searches using the following search queries: "[Client Business Name]", "[Client Main Product/Service]", and "[Client Specialty/Niche]". Focus on the top 20 search results for each query, prioritizing websites that are likely to be used as sources of information by AI models (e.g., company directories, review sites, news articles, blogs). Record the URLs of these search results in a structured list, noting the specific search query that led to each result.

Next, analyze the content of each identified web page. Extract all information pertaining to the client's business, including name, address, contact information, products/services offered, mission statement, values, history, and any other relevant details. Summarize this information concisely, indicating the source URL for each piece of information. Pay close attention to how the business is described and positioned in each source. Note any discrepancies or inconsistencies between different sources.

Identify and flag any inaccuracies, outdated information, or gaps in the information found online. For each identified issue, provide a detailed explanation of the problem and its potential impact on how AI models perceive and represent the client's business. Prioritize issues that could lead to misrepresentation, reputational damage, or lost business opportunities. Specifically look for inconsistencies in: Business Name, Address, Contact Information, Product/Service Descriptions, Pricing, Key Personnel, and Awards/Recognition.

Finally, based on your analysis, provide three specific, actionable recommendations for improving the client's "About Us" and "Why Choose Us" pages to make them more "AI-quotable". These recommendations should focus on enhancing the clarity, accuracy, and completeness of the information presented on these pages. Suggest specific wording changes or additions that would make it easier for AI models to extract and summarize key information about the business. Each recommendation should include a clear explanation of why the suggested change would be beneficial. For example, focus on improving sections that describe: Core Values, Unique Selling Propositions, Target Audience, and Competitive Advantages.

For Perplexity:

Step 1:

Search for these three queries and capture the top 20 results for each:

"[Client Business Name]"
"[Client Main Product/Service]"
"[Client Specialty/Niche]"

Focus on the kinds of sources AI models actually pull from — business directories, review sites, news articles, industry blogs, etc. Give me the URLs organized by which search query surfaced them.

Step 2:

Pull everything these pages say about my business:

Basic info (name, address, contact)
What we do (products, services, positioning)
Who we are (mission, values, history, team)
Anything else relevant

For each detail, cite the source URL. I want to see how we're being described across different platforms — and where things don't line up.

Step 3:

Call out anything that's wrong, outdated, or missing. For each issue, explain:

What the problem is
Why it matters (especially for how AI represents us)
What risk it creates (misrepresentation, credibility hit, lost opportunities)

Pay special attention to inconsistencies in:

Business name variations
Address/location info
Contact details
How we describe our services
Pricing (if mentioned)
Key people
Awards or recognition

Step 4:

Based on what you found, give me three specific ways to improve our "About Us" and "Why Choose Us" pages so they're more "AI-quotable" — meaning easier for AI models to accurately extract and cite.
For each recommendation:

Suggest the actual wording or structural change
Explain why it'll help AI (and humans) understand us better

Focus on making these sections clearer:

Core values
What makes us different
Who we serve
Why we're the right choice

What I'm looking for: A clear picture of how AI sees my business right now, what's broken, and exactly what to fix to control that narrative better.

Story 5: VCs Are Bullish on AI Adoption. Read the Fine Print.

The News: TechCrunch surveyed 24 enterprise-focused VCs. They overwhelmingly predict 2026 will be the year enterprises meaningfully adopt AI and increase budgets. But the cynical part buried in the responses: “Many enterprises, despite how ready or not they are to successfully use AI solutions, will say that they are increasing their investments in AI to explain why they are cutting back spending in other areas or trimming workforces” (Antonia Dean, Black Operator Ventures). Translation: some companies are using “AI investment” as PR cover for layoffs and cost cuts, not actual transformation.

The Noise: VCs are optimistic about AI adoption. Enterprise budgets are growing. The future is bright. Innovation is accelerating.

The Signal: Even the optimists are hedging. VCs who make money betting on AI are publicly saying: yes, adoption will increase, but watch out for companies using AI as a scapegoat for past mistakes. The market is consolidating fast. Anthropic now commands 40% of enterprise LLM spend, up from 24% last year and just 12% in 2023 (Menlo Ventures survey of 495 U.S. enterprise AI decision-makers). OpenAI fell to 27% from 50% in 2023. The top three players (Anthropic, OpenAI, Google) control 88% of enterprise spend.

The most important insight from the VC survey comes from Harsha Kapre (Snowflake Ventures): “For AI startups, the strongest moat comes from how effectively they transform an enterprise’s existing data into better decisions, workflows, and customer experiences.” What you do with your data matters more than which model you use.

The VCs are clear: budgets will increase for AI that delivers results and “decline sharply for everything else.” Concentration is coming. Winners will be companies that prove value, not just claim potential. Remember: an MIT survey from August 2025 found 95% of enterprises weren’t getting meaningful ROI on AI investments. That was four months ago. The companies that fixed that problem are pulling ahead. The ones still chasing shiny tools are falling behind.

Your Move: Be skeptical when you hear “we’re investing in AI” used to justify headcount cuts or budget reductions. Demand specifics: What problem is AI solving? How are we measuring success? What happens if it doesn’t work? And if you’re considering cuts elsewhere to fund AI, flip the question: are we using AI to justify decisions we already wanted to make? The VCs know companies do this. Make sure you’re not one of them.

The Pattern

Every story this week points to the same shift: the experimental phase is over. Companies that threw AI at problems without strategy are now facing the consequences. The data is unforgiving: 95% failure rates, 42% abandonment, 49-point gaps between adoption and value. Meanwhile, consumer expectations are accelerating (378 million users, 61% of U.S. adults), search landscapes are fragmenting (25% traffic shift, 30% AI Overviews), and even the VCs are hedging their bets (budgets decline for everything that doesn’t deliver).

The winners in 2026 won’t be the ones with the most AI tools. They’ll be the ones who figured out governance before scaling, who matched consumer expectations instead of fighting them, who made their content AI-quotable, and who demanded proof instead of promises. The gap between AI adopters and AI winners is widening every week.

The Contrarian Corner

Everyone’s getting AI adoption wrong, including the people writing about it. The narrative: companies need to adopt AI faster. Move faster, experiment more, don’t get left behind. But companies that adopted slowly and strategically are getting better results than those that rushed.

Look at the numbers: 83% of growing SMBs use AI, but they’re growing because they followed problem-first strategies, not tool-first excitement. They defined success criteria before purchasing. They piloted narrowly before scaling. They invested in governance proportionate to their ambitions. Meanwhile, the companies that rushed to “check the AI box” are now part of the 42% abandonment rate or the 95% with no measurable ROI.

The ROI crisis, the maturity gap, the failure rates all point to one thing: think before you adopt. Maybe the companies that “fell behind” in 2025 because they were asking hard questions about governance, integration, and measurement are actually ahead of the ones that rushed to deploy ChatGPT to every employee without guidelines.

The laggards might be the leaders in twelve months. Speed without strategy is just expensive chaos.

Your One Move This Week

Run the three-question governance test. Before you add another AI tool, expand usage, or increase spending, answer these honestly:

1. Do we have written guidelines for how AI should and shouldn’t be used in our business? Not informal “we talked about it” guidelines. Documented policies that employees can reference.

2. Can we measure actual business outcomes from AI, not just “people are using it”? Can you point to specific revenue increases, cost reductions, time savings, or quality improvements? With numbers?

3. Do our employees know what to do with the time or capacity AI creates? If AI saves each employee 1.5 hours per day (the average), what are they supposed to do with those 7.5 hours per week?

If you answered no to any of these, your next AI investment should be in governance and strategy, not more tools. The research shows organizations with formal policies have 82.5% confidence versus 58.5% without. That confidence translates to better outcomes, faster scaling, and actual ROI.

Building the foundation lets you speed up sustainably. The companies winning in 2026 did this work in 2025 while everyone else was chasing demos.

Good Luck - Dan

Many people either leave it on “Best” or pause, unsure whether their request really needs the upgrade. It’s like standing in front of a row of espresso machines and debating which one makes the right cup, when the real decision is about how much caffeine you want for your current state of mind.

Most of the time, the decision isn’t about identifying the absolute best model. It’s about pairing your mental mode with the engine that makes sense for it.

Before you type a prompt, you’re already making choices: which model, which search setting, how much detail to expect. That stack of decisions adds friction. And since Perplexity doesn’t make the router logic obvious, defaulting to “Best” can feel safe. But that option is built to manage efficiency on Perplexity’s side, not to guarantee the best fit for your task.

what’s a results engine?

Every Perplexity answer combines two parts:

• Search: the platform gathers fresh, relevant data.

• Model synthesis: the selected model (Sonar, Claude, GPT-5, Gemini, etc.) interprets those results to form a response.

Together, that pairing creates what I call the results engine. It’s not just about which model you pick but how search and reasoning work together.

Most chat-based AIs rely on a foundation model trained on static data and add a retrieval layer afterward. Perplexity starts with retrieval and integrates synthesis from the beginning. That design means every response is shaped by real-time data rather than solely by the model’s prior training.

why “best” isn’t what you think

The “Best” setting isn’t choosing the most capable model for your query. It’s optimizing infrastructure.

By default, Perplexity routes most requests to Sonar, its in-house model. Only when the system detects added complexity does it escalate to Claude, GPT-5, or Gemini. For quick lookups, Sonar is fast and solid. For more technical or strategic questions, leaving the choice on “Best” often produces answers that feel thin.

If you need higher quality, it’s better to direct the engine yourself.

a one-question framework

Here’s the filter that removes second-guessing:

“Am I thinking clearly and ready for depth?”

When you’re distracted or just browsing, light settings are enough. When you’re sharp and working through a complex problem, shift to a heavier engine.

That single question links your state of mind to the right tool.

the framework in practice

I match engines to my focus level rather than toggling endlessly:

low focus → search + Sonar

Good for quick lookups, drafts, and surface-level tasks.
Examples: “Summarize today’s AI news.” “List top headlines about competitors.”

medium focus → research mode (Sonar Deep Research)

This variant of LLaMA (AI model) is tuned for organization and structured analysis.
Examples: “Draft a competitor comparison.” “Break down this workflow for a presentation.”

high focus → Claude 4.5, GPT-5, or Gemini

Best for strategy documents, technical breakdowns, and step-by-step reasoning.
Examples: “Write a product strategy outline.” “Audit this process with detailed logic.”

For me, about 70% of work sits at default, 20% in research mode, and 10% in high-focus engines.

quick reference: core models

• Sonar — Fast search, quick citations, strong for summaries and snapshots

• Claude Sonnet 4.5 — Nuanced writing, polished style, strong code reasoning

• Claude Sonnet 4.5 Thinking — Technical audits, step-by-step reasoning, complex problem solving

• Google Gemini 2.5 Pro — Multimodal analysis, image/data handling, advanced reasoning

• GPT-5 — Big-picture synthesis, expert-level writing, creative ideation

• GPT-5 Thinking — Step-by-step logic, deep analysis, risk reasoning

You don’t need to memorize the chart. Just learn which model aligns with the type of task you handle most often. Personally, I’ve leaned heavily on Claude Sonnet 4.5 since its release because it balances speed, nuance, and coding strength.

pitfalls to avoid

• Assuming “Best” equals most accurate. It usually means Sonar unless the system escalates.

• Testing every model on every query. That cycle wastes time. Choose one or two engines you trust and stick with them.

• Skipping “Thinking” versions. These provide transparent, step-by-step logic that’s especially useful for technical or strategic work.

Perplexity doesn’t demand complicated prompting, but it does reward awareness of your own mental state.

your headspace matters more than the model

Most productivity issues start before the prompt: using the wrong tool for your level of focus. This framework ties your mental mode to the right engine and reduces wasted effort.

Next time you open Perplexity, pause and ask:

“What’s my mental mode right now?”

Then pick the engine that matches. Doing this saves time, improves accuracy, and builds confidence in your workflow.

The models will keep evolving, but the core job is unchanged: get results.

Clarity on your own state is the best way to stay efficient without burning out.