Before leaders reorganize their companies around AI, they may need to rethink what parts of leadership are uniquely human.
getty
Sunday night. Two weeks of the calendar are open on my laptop. I am trying to write down what made a particular client call successful. Over the past several months, I’ve been building what I call a Wisdom Engine, an AI system trained on decades of my writing, conversations, research, and operating experience to better understand which parts of leadership judgment are truly durable in an AI-native economy. The fifty-three-minute call had ended with a leadership team rethinking their year, and I wanted to understand why. What exactly had changed the conversation? Which part was pattern recognition, which part was experience, and which part was something harder to define?
I sit there for twenty minutes. I cannot do it.
That is the moment I realized most of us are running the wrong audit.
Every leader I talk to is asking some version of “what’s our AI strategy.” It is the wrong first question. The first audit is not the company’s. It is yours. Because before you can decide what AI does inside your organization, you have to know what is left of your own job once AI has taken the rest. And almost nobody has looked.
What’s Actually in Your Last Two Weeks
There is a wave of personal AI audits circulating among knowledge workers right now. Nate Jones has one of the sharpest, asking you to sort the last two weeks of your work into theatre, commodity, on-the-line, and durable. Run on a senior individual contributor, the audit is sobering. Run on a leader, it is something else.
A meaningful share of what shows up on a leader’s calendar is not even on the spectrum the audit measures. It is not work, exactly. It is presence. Showing up to the meeting because the meeting needs you to be in it. Holding open a question against a room full of people who want it closed. Sitting with a number that does not quite add up and refusing to round.
Most of that work is invisible to the systems that measure leadership. It is also invisible to the AI you are about to deploy. The two facts are connected.
Jobs Are Already Unbundling. Yours, Too.
In a piece I wrote with Sangeet Paul Choudary in Harvard Business Review last fall, we argued that AI’s deepest impact is not automation but coordination. AI unbundles jobs into their component tasks, then rebundles those tasks into new configurations around new constraints. Most leaders read that argument and apply it to the org chart. What should HR redesign, what should the COO automate? What should the new operating model look like?
Almost none of them apply it to themselves.
But of course leadership is a job, made of tasks, and the tasks unbundle. The status review unbundles. The synthesis of three decks into one document unbundles. The first draft of the strategy memo unbundles. A surprising amount of the calibration work that used to require a senior human, reading a deck, pattern-matching across deals, summarizing across functions, is now good enough from a model running for pennies. If the only thing you brought to your role was the ability to produce those outputs faster and cleaner than your team, you have a problem you have not yet named.
What is left after the unbundling is the rebundling. And for leaders, the rebundle is happening around something narrower than most people admit: judgment. Not analysis. Not synthesis. Judgment. The part that decides which framing to hold open, which constraint to accept, which call to make when the data will not tell you. The part that does not reduce to a process document, because the moment you reduce it, someone else can run it.
The Explorer’s Audit
I think about this through what I call the Explorer’s Mindset. Operators run known terrain better. Explorers read terrain that has not been mapped yet. AI is now a better operator than most operators. It is not yet a better explorer than the best ones. The work that compounds for a leader in 2026 is not the work AI can do faster. It is the work AI cannot frame at all. It’s the work where the question itself is contested, where nobody has even agreed on what the real inputs are yet, where you sit in a conversation and read what is actually going on rather than what is on the agenda.
The Wisdom Engine experiment I have been running on myself is partly a forcing function for this. Trying to articulate which of my inputs become judgment teaches me, week by week, which fraction of my own work is genuinely mine and which fraction is operator work I have been mistaking for leadership. The answer has been uncomfortable. A larger share of what I have been doing for years than I would like to admit could now be done by something that costs less than my coffee.
Three Things To Do Before Next Monday
If I were giving a leader one weekend of homework, it would be this.
Run the audit on yourself before you run it on anyone else. Pull two weeks of your calendar. For each item, ask one question: would the meaningful outcome have changed if a well-trained model had been in the room instead of me? Be honest. The fraction that comes back “yes” is the fraction of your role you are about to lose, whether you redesign it deliberately or not.
Start building your own Wisdom Engine. Not because you need the technology, but because the act of trying to articulate your judgment teaches you what you actually do. The exercise is the value. You will discover that the most valuable parts of your thinking are the parts you cannot fully write down. That is the insight, not the failure.
Redesign the calendar. Move your durable work. Move the question-holding, the contested framings, the conversations where you are reading rather than reporting, out of the margin and into the center. Most senior leaders spend roughly ten percent of their week on what only they can do, and the rest on what a model could now do for them. Flip those numbers and you will be unrecognizable to your competitors in eighteen months.
The AI strategy your company needs starts with knowing what to build around. That answer does not live in the org chart. It lives in your last two weeks.
The audit takes ninety minutes. Most leaders will not run it. The ones who do will spend the next decade competing against an unusually small field.