The framework
How I think about AI.
AI should amplify how you think, not replace it. The people and companies that win with AI stay in the driver's seat, using it to think more rigorously, not to avoid thinking at all. Everything I build rests on that, and on the belief that the goal was never to think less. It was to think better.
The goal of AI is not to offload your thinking. It's to amplify it.
What's at risk
What it costs to get this wrong.
Deploy AI without redesigning the human systems underneath it, and the gains don't compound, they leak. And the failures aren't separate. They cascade: fear at the top becomes dysfunction in the middle becomes collapse at the individual level.
Enterprise
The fear arrives before the tools, and leaders stay quiet.
People are processing layoffs and headlines before a single tool is introduced. By rollout day they aren't starting from neutral. They're starting from fear, and leaders who treat launch as day one have already missed the moment that mattered.
Only 13% of workers feel rewarded for reinventing their work with AI, and 65% fear falling behind. Microsoft 2026 Work Trend Index, 20,000 workers
Leaders issue transformation mandates while hiding their own uncertainty. Without that signal from the top, no one below feels safe to learn in the open.
52% of CXOs never communicate AI's role internally; 72% of those who stay silent cite fear of reputational damage. Capgemini / Workday, 2025
Nobody owns the redesign. IT deploys the tools, HR manages the change, and no one owns how the work actually gets done in between, so the default is lift-and-shift.
“The last mile is not a technology problem. It is an operating-model problem.” Harvard Business Review, March 2026
Unaddressed, that fear cascades straight down into the team.
Team
Without safety from the top, the gains never circulate.
Psychological safety is the missing infrastructure. People won't experiment where failure is visible and the stakes feel high, and every other team practice fails before it starts.
83% of executives say a psychologically safe culture measurably improves AI success; psychological barriers now outweigh technical ones. MIT Technology Review, December 2025
Gains stay trapped in pockets. One person figures something out and the learning never moves. It's a circulation problem, not an adoption problem.
Only 19% of AI users reach the high-performing “Frontier” where capability and readiness reinforce each other. Microsoft Work Trend Index
AI removes the forcing function for collaboration. People stop looping each other in, and the weak ties that carry new ideas across a company quietly atrophy.
74% of knowledge workers now use AI for support once provided by colleagues. Harvard Business Review, 2026
Isolated and unsupported, individuals are left to use AI without a foundation.
Individual
Without a foundation, more usage makes it worse, not better.
People using AI without training show measurable declines in confidence, competence, and motivation. The tool compounds the very gap it was supposed to close.
“Diminished competence” is one of six components of the psychological debt AI creates, and it feeds a vicious cycle. Harvard Business Review, May 2026
The heaviest AI users burn out the fastest. The time AI frees gets reloaded with heavier work, not returned as breathing room.
55% of heavy AI users report burnout versus 32% overall, and 51% plan to quit within three to six months. McKinsey, 13,000 workers
AI now handles the formative tasks that used to build judgment. Deskilling is at least reversible. Never-skilling, where the foundational skill is never built at all, may not be.
Entry-level roles are down 35%, collapsing the apprenticeships that used to build expertise. WEF / Harvard, 2026
And individual loss becomes collective: the fewer people who can still think rigorously, the less new knowledge enters the system at all.
Fear cascades down. Capability erodes up. That is the tax you pay for deploying AI without redesigning the work underneath it, and it's exactly what the framework below is built to prevent.
Drawn from work-practices research developed with my collaborators Jen Rambur and Stephanie, and grounded in the 2025–2026 research on AI and the workforce.
The practice
AI Organizational Design.
The technology question is mostly settled. AI works. The real question is who should do what work, and how you design that on purpose, so people stay engaged and capable while the organization gets the full benefit. It is the conversation most companies haven't started, and it's the one that actually matters. Four questions sit under it.
Not just can AI do it, but should it? What's lost if AI does it, and what's gained if the human does? That question comes before any tool.
The right balance of cognitive load, automation, and human collaboration for this role, this person, this team. Designed, not left to drift.
As agents run more tasks in parallel and the pace accelerates, how does a person stay sharp and engaged instead of overwhelmed? We're not machines.
The work that keeps you learning and sharpens your expertise. That work gets protected, not automated. Not all value is productivity value.
How I train for AI
Fluency isn't prompting harder. It's building a system.
Most people think using ChatGPT is using AI. It isn't, not in the full sense. Real capability comes from context, specialization, and trained systems, and from learning how to navigate constant change rather than mastering any one tool.
The single most impactful thing anyone can do is give AI more and better context: your role, your values, your constraints, your history with the problem. A vague prompt gets vague output. Your prompt can only go as far as the context and the persona behind it.
Real capability rests on six connected skills, not clever one-offs: prompting and context, persona design, iteration, quality control, task breakdown, and human-AI collaboration. Each one supports the others.
General AI has no memory of your business, standards, or voice. A trained agent is built on your data, so the output isn't plausible, it's specific. I prepare a reusable data model once, then every agent draws from it. One agent can't do everything well, so I build one per tightly-scoped job and connect them.
Experts use AI to move faster and catch what's wrong. Non-experts need the opposite: slow it down, learn first, build judgment before generating output. The same tool, two completely different modes, and knowing which is which is the whole game.
What I protect
The parts of work that keep people sharp, and human.
This is the dimension of AI adoption that doesn't fit on a slide, which is exactly why most transformations skip it, and exactly why they stall.
When AI can generate insight in seconds, knowing the answer is no longer what makes you valuable. Judgment is. Expertise is what catches what's wrong, and it's what sustains your value as AI does more of the generating.
Used to avoid thinking, AI erodes the very capability that made people valuable. Used to handle the low-leverage work, it frees you for judgment. We are at AI's social-media moment: the early window, before the costs compound, when we still get to design for them instead of cleaning up the fallout later.
As AI agents learn how you work, they become a portable record of how you think, and the question of who owns that, you or your employer, is genuinely unresolved. When your judgment trains a company's model with no say and no stake, your expertise is being extracted, not developed. Protecting people means protecting their cognitive sovereignty, not just their headcount.
Most people aren't afraid of AI as a tool. They're afraid of losing what work gives them: the ability to provide, their place, their identity. And a person in fear brings only a fraction of themselves to the work. You cannot transform an organization with people holding most of themselves back, which is why moving through the fear, not around it, is the whole job, and the part most transformations skip.
Where it comes from
This isn't theory. It's a decade of doing the work.
This framework is drawn from teaching hundreds of professionals to build with AI across the Maven and Phoenix Formula cohorts, from leading Future of Work transformation at enterprise scale, and from writing about all of it every week.
Every proposal I write and every engagement I run is structured around it: assess honestly, model the return, deploy with change management, measure whether it worked, and keep the capability in the room after I leave. The thinking and the doing are the same thing.
Let's talk
This is the thinking behind every engagement.
If it sounds like the way you want to bring AI into your organization, let's talk about where you're headed.