Before leaders restructure their companies around AI, they may need to rethink what parts of leadership are uniquely human.
getty
Sunday night. I have two weeks’ worth of calendars open on my laptop. I’m trying to write down why certain client calls were successful. For the past few months, I’ve been building what I call the Wisdom Engine. This is an AI system trained on my decades of writing, speaking, research, and operational experience to better understand what parts of leadership decisions are truly durable in an AI-native economy. The 53-minute conference call ended with the leaders reflecting on the year, but I wanted to know why. What exactly changed the conversation? Which part is pattern recognition, which part is experience, and which part is hard to define?
I sit there for 20 minutes. I can’t do that.
That’s when I realized that most of us are doing audits wrong.
Every leader I talk to has asked in some way, “What is our AI strategy?” Your first question is incorrect. The first audit is not of the company. It’s yours. Because before you can decide what AI will do in your organization, you need to know what will remain in your job after the AI takes time off. And almost no one is watching.
What actually happened in the last two weeks
There is currently a wave of personal AI audits among knowledge workers. One of the most astute comments comes from Nate Jones, who calls for the last two weeks of work to be categorized into theater, merchandise, online, and long-term. It is run on the basis of senior individual contributors and audits are rigorous. Running on a leader, that’s something else.
An important part of what appears on a leader’s calendar is not even in the scope of what an audit measures. Honestly speaking, it’s not a job. It’s presence. I attend meetings because I have to. Leave the question open to a room full of people who want closure. Refuse to round off by leaving numbers that don’t add up.
Most of that work doesn’t look like a system for measuring leadership. It is also invisible to the AI you are deploying. The two facts are connected.
Employment is already falling apart. you too.
In an article I wrote with Sangeet Paul Chaudhary, harvard business review Last fall, we argued that AI’s deepest impact would be coordination, not automation. AI splits the job into component tasks and rebundles those tasks into new configurations based on new constraints. Most leaders read that discussion and apply it to their org chart. What should HR redesign and what should the COO automate? What should the new operating model look like?
Few people apply that to themselves.
But, of course, leadership is a job, it consists of tasks, and tasks are separated. Status reviews are not bundled. Combining three decks into one document unbundles them. The first draft of the strategy memo will be unbundled. What used to be an incredible amount of reconciliation work that required a senior person to read materials, pattern match across trades, and summarize entire features can now be done with a model that costs pennies. If the only thing you brought to your role was the ability to produce faster, cleaner output than your team, you have a problem you haven’t named yet.
What remains after unbundling is rebundling. And for leaders, the backlash is about something narrower than many acknowledge: judgment. It’s not an analysis. It’s not synthetic. judgement. This is the part that decides which frames to keep open, which constraints to accept, and which calls to make if the data doesn’t tell us. Because the parts that don’t get reduced to a process document can be executed by someone else the moment you do.
explorer audit
I think about this through what I call the “explorer mindset.” Operators perform better on known terrain. Explorers read terrain that has not yet been mapped. AI is now a better operator than most. It’s still not a better explorer than the best one. The jobs that matter to leaders in 2026 are not the ones that AI can do faster. This is a work that AI cannot frame at all. It’s the job of sitting in a conversation and reading what’s actually going on, not what’s on the agenda, when the question itself is contested and no one even agrees yet on what the actual input is.
The Wisdom Engine experiments I’m running myself are part of the functionality that forces this. When you try to clarify which of your inputs become judgmental, you find out every week which parts of your work are truly yours and which parts are the work of the operator you mistakenly thought was leadership. The answer was unpleasant. A greater portion of what I’ve done over the years than I’d like to admit could now be done with something cheaper than coffee.
3 things to do by next Monday
If I were to give leaders a homework assignment for the weekend, this would be it.
Perform an audit on yourself before performing an audit on others. Let’s pull out two weeks on the calendar. Please ask one question for each item. Would the meaningful results have changed if a well-trained model had been in the room instead of me? Be honest. The parts that get a “yes” are the parts of the role you’re going to lose, whether you intentionally redesigned them or not.
Start building your own wisdom engine. Not because we need technology, but because the act of clarifying our judgments teaches us what we actually do. Practice is what makes it worth it. You will find that the most valuable parts of your thoughts are those that cannot be completely written down. It’s not a failure, it’s an insight.
Redesign your calendar. Move durable pieces. Holding questions, framing discussions, and moving conversations that are reading rather than reporting from the margins to the center. Most senior leaders spend about 10% of their week doing things that only they can do, and the rest of the week doing things their models can do for them. If you flip those numbers over, you’ll be unrecognizable to your competitors in 18 months.
The AI strategy your company needs starts with knowing what to build around. The answer doesn’t exist in the org chart. It lives in your last two weeks.
The audit will take 90 minutes. Most leaders don’t do that. Those who do so will spend the next decade competing against an unusually small field.
