A few months ago I was running a session where AI was going to solve everything. Not some things. Everything. It was not my story, I was there to facilitate, but the tone had been set before I could say a word. The slides were clean, the arrows pointed up, and someone had said the word “exponential” three times before the coffee went cold.
The room was a mix. The people who decide, and the people who actually do the work, at the same table for once. And as the picture built toward a future without friction, I watched the practitioners do something that stayed with me. They looked away. At their screens, at the window, at nothing in particular. Not out of boredom. Out of experience. They already knew that the demo and the Tuesday morning are two different worlds.
At some point I stopped the presentation. I asked the people who were looking away what they saw. The room went quiet, and then came the plain truth the slides had left out.
That gap is what this piece is about.
What one CEO said about other CEOs
Aaron Levie, the founder of Box, posted something recently that stuck with me. Tech CEOs, he wrote, are “uniquely prone to AI psychosis.” Not because they are foolish or naive, but because in his view they stand too far from “the last mile of work that still has to happen to generate most value with AI.”
Read that line again. The last mile of work that still has to happen before AI delivers anything real.
What makes it land is that Levie is no AI skeptic. He sells software, he believes in the technology, he uses it every day. His point is quieter and sharper for it: if you do not pick up the tools yourself, you cannot judge them. You do not get to call the game from the stands.
“Psychosis” is a metaphor here, not a diagnosis. It describes a state where your conviction comes loose from what is actually happening. A story in your head that has gotten so smooth that reality can no longer get through. And when that story rolls down through an organization from the top, it spreads.
The last mile is where the work lives
In innovation work I keep hearing that the idea is the hard part. It rarely is. The idea is the pitch. The idea fits on a slide.
Value shows up further along, on the last mile, where things get messy. That is where the customer responds a little differently than the model predicted. That is where the legacy system refuses to connect. That is where the colleague has to trust the new way of working before she will use it. The last mile is not a detail you handle later. The last mile is the work.
And distance is exactly where this breaks. The higher you climb in an organization, the more the world starts to look the way it looks on slides. Tidy. Linear. Stripped of the friction that every practitioner knows by heart. You start steering by the summary of the summary, and one day you mistake that summary for the real thing.
People who make things tend to see this sooner. Not because they are smarter, but because they feel the friction. A photographer knows the best frame is rarely the planned one. A strategist who has once run an implementation herself knows how many promises die between the kickoff and the handover. That knowledge lives in your hands, not in your deck.
What I heard from the room that afternoon was not resistance to AI. They were not afraid of the technology. They were tired of the distance. Tired of plans written by people who had never once opened the tool.
The backlash is a signal, not noise
You can see that same distance now at the big platforms. AI is being pushed in everywhere, whether you asked for it or not. And people are starting to push back.
One small example that stuck with me: DuckDuckGo reported that installs were up thirty percent, partly because users are done with the AI search results Google keeps forcing on them. DuckDuckGo is tiny next to Google, so this is no earthquake. But it is a reading. There is a serious group of people who do not want the current direction.
It is tempting to wave such numbers away as nostalgia, or as resistance to change. I think that is a mistake. It is the same pattern as in that meeting room, only at the scale of millions. A product dreamed up from the top meets a user standing on the last mile who thinks: this does not solve my problem, this is your problem that you are handing to me.
Distance from the user catches up with you. Quietly at first, then in the numbers.
What this means for whoever decides
This is not a case against AI. I work with it every day and I would not go back. It is a case against judging from a distance.
If you make decisions about AI, you earn your judgment by doing. Not by reading a report about the tool, but by using the tool yourself for a week on your own real work. Not the polished demo, your messy Tuesday. Only then do you feel where the promise stops and the last mile begins.
Three things I carry into every engagement:
Go to the last mile yourself. Sit beside the people doing the work, not across from them. The truth about your AI strategy lives there, not in the steering committee.
Distrust your own fluent story. If a plan sounds too smooth, there is probably friction that got polished out, friction that is still there in real life. Go find it before the market finds it for you.
Measure the distance, not just the ambition. How far does the person deciding stand from the person executing? That number often predicts more than any roadmap.
The people in that room who looked away were not wrong. They knew something that had been lost at the front of the presentation. Not that AI does not work, but that working is something you do, not something you present.
You walk the last mile yourself. Otherwise you do not even know it is there.
This piece belongs to an ongoing line on AI and execution. See also why AI strategy is always human work.