LISTEN
Listen to this perspective
Written by Oleg Cohen. Companion narration uses the Brian voice.
Something strange is happening inside business. Almost every leader has heard about AI, and most of their employees have tried it. Companies are buying subscriptions, boards are asking questions, consultants are presenting roadmaps, and vendors announce a new kind of agent every week. There is no shortage of activity.
I increasingly suspect the activity is hiding something. A large share of business leaders don’t actually know what to do. Some know they don’t know. Some don’t yet know enough to realize it. Some have decided that giving employees ChatGPT, Claude or Copilot means the company has dealt with AI. Others launched pilots, formed committees and hired consultants, and somewhere between the demonstration and operating reality, progress stalled.
Then there is a smaller group that understands enough to be genuinely concerned. They aren’t frightened of AI itself. They’re concerned about something larger: the sense that the future their company is entering may not resemble the one in which it succeeded. They can feel the ground moving, but they can’t yet see where it’s moving to.
That combination produces a peculiar executive condition. The future matters enormously, and it is harder than ever to predict. Doing nothing feels dangerous, and so does doing the wrong thing. And the organization may have neither the understanding nor the means to steer between the two. I call this the Executive Reality Gap: the distance between what a leader senses is changing and what their organization is equipped to do about it.
Three realities
The first reality is that the future won’t resemble the past. That doesn’t mean everything changes. Customers will still have needs, trust and economics will still matter, and execution will still decide outcomes. But intelligence is becoming far more available. Work that once required scarce specialists can be accelerated, supported or done by machines. Small teams can exercise capabilities that once took large organizations, and competitors can appear with very different cost structures. Nobody knows exactly what the future enterprise will look like. Assuming it will look like today’s enterprise with better software is getting harder to defend.
The second reality is that the future is a fog. Executives hear that AI will change everything and that most AI projects fail. They see agents that impress in a demonstration and make elementary mistakes in production. They hear about enormous productivity gains and can’t find them in their own financial statements. McKinsey’s latest survey found 80 percent of respondents reporting personal productivity gains and only 37 percent reporting any EBIT impact. Leaders are told to move immediately, and also to worry about governance, security, privacy and lock-in. Somewhere inside all of that, they are expected to form a strategy.
The third reality is the uncomfortable one. Many organizations aren’t equipped to move through the fog. They may have excellent people, strong businesses and capable technology teams. But this transition asks for something other than buying software. It asks a company to understand what is changing and what it means for this particular business. It has to decide which assumptions still hold, imagine what has become possible, experiment, redesign work, keep human judgment where it belongs, and adjust as it learns. That isn’t an AI procurement problem. It’s the ability to change, which I described in What to Commit To When You Can’t Predict as institutional maneuverability.
The adoption illusion
AI adoption is not the same thing as a company becoming more capable. A company can have thousands of employees using AI and remain essentially unchanged. Usage looks respectable, a presentation goes to the board, and the organization is declared to be making excellent progress. It may have created thousands of individually augmented employees. That’s useful, but it isn’t an augmented institution. I’ve written about the mechanics in Your AI Is Getting Smarter. Your Company Isn’t. and Busy With AI.
The danger here isn’t failure. It’s premature satisfaction. The organization succeeds at adoption and mistakes that success for transformation. Visible activity produces comfort, and comfort reduces urgency, while somewhere a competitor may be redesigning the operating model the company depends on.
Where leaders actually stand
To understand why so little changes, it helps to stop treating executives as one group. They sit in very different places.
Some are in denial. Their industry is special, their relationships protect them, their expertise can’t be copied. Sometimes they’re right about the near term. The more useful question for them isn’t whether AI can replace their business. It’s whether someone else can redesign the economics around it.
Some are unaware. They aren’t resisting. They’ve heard the vocabulary and perhaps attended a presentation, but they can’t yet connect the technology to capability, competition or economics. They don’t yet know which questions to ask.
Some are curious. They read, attend conferences and ask employees to explore. Curiosity is valuable, but it can become a comfortable place to stay. Enjoying the ideas isn’t the same as making a consequential decision.
Some are productivity adopters. Employees have tools, there are pilots, perhaps a formal AI strategy. These leaders already believe AI matters. Their risk is believing they’re further along than they are.
Some are stalled experimenters. They tried. Something worked in a demonstration, and then the data was fragmented, integration got hard, governance intervened, the economics were unclear, and the pilot couldn’t scale. They have scar tissue, and scar tissue is valuable. They’ve learned that the missing piece usually isn’t another model. It’s design: how the business, its knowledge and the technology fit together.
Some are strategically concerned. They’re asking where margins will move, which parts of their value proposition become commodities, what expertise should become institutional, and how much time they have. They aren’t looking for a demonstration. They’re looking for orientation.
And a few have crossed a threshold. They’ve stopped asking how to adopt AI and started asking what the company should become.
Most leadership teams contain several of these at once. The CEO may be strategically concerned while the CFO is a productivity adopter and the board is curious. A company tends to move at the pace of the position its leaders can agree on, which is rarely the most advanced one in the room.
Seeing it isn’t enough
A leader can understand all of this and still be unable to act. Three things decide whether recognition turns into change.
The first is recognition itself: does this person see that the old assumptions may not hold, for this company specifically? The second is agency: can they commit capital, change priorities and alter how the company operates? The third, and the one most often overlooked, is consequence ownership: do they personally bear the downside if the company gets this wrong, and capture part of the upside if it gets it right?
Those three are distributed unevenly. A deeply informed manager with no authority can see the problem and still not move it. A CIO may understand the technology well and lack the authority to redesign the business. An owner may understand half as much and have everything at stake. Behind those three sit practical conditions: money to experiment, tolerance for acting before certainty, an organization able to absorb change, and a time horizon long enough to care.
Consequence ownership matters more than title. Picture two executives with similar budgets and similar understanding. One is rewarded for running a successful AI program: deployed on time, adopted, milestones met, board satisfied. The other is accountable for what the company will be worth in five years. The first asks whether the program worked. The second asks whether the company is becoming the company it will need to be. That isn’t a judgment about character. It’s what incentives and time horizons produce. An executive can build a good career by appearing current and moving on before the long-term consequences arrive. An owner can’t leave the consequences behind so easily.
That makes certain people unusually important here: founder CEOs, family owners, private company principals, board members with real exposure, and private equity sponsors.
Private equity
Private equity concentrates these conditions in one place: economic exposure, a defined holding period, a value-creation mandate, visibility across a portfolio and the authority to push for change. For a sponsor, the question isn’t whether AI cuts costs. It’s whether the portfolio company becomes more valuable and more transferable. Can expertise scale? Can acquisitions integrate faster? Can institutional knowledge survive turnover? Can a roll-up behave like one company? I took up the valuation side in EBITDA Up, Value Down and the dependency side in Underwriting the Heroics.
The useful question for an operating partner is where each portfolio company’s leadership sits, both on the spectrum and on recognition, agency and consequence ownership. A CEO with high agency and low recognition needs a different conversation from one who sees the change clearly but lacks the authority or the slack to act on it.
Why the middle may move first
Size matters, though not in the obvious way. Large enterprises have money, technology organizations, consultants and governance. They also have politics, committees, legacy systems and competing agendas. The sophistication is real, and so is the swamp. A large company can spend $50 million on AI and remain surprisingly hard to change.
The lower middle market and middle market sit somewhere more interesting. These companies are complex enough for their knowledge to be valuable and for improvement to be worth millions. They are small enough that the operating model is still comprehensible and a CEO or owner can actually change it. The distance between insight and action is short. As I argued in The Forgotten Business of America, they are also the companies least served by the platforms and programs built for the Fortune 500.
The productive anxiety zone
There is a version of this argument that turns into fear-selling, and I want to avoid it. Nobody needs fear manufactured for them. Reality supplies enough uncertainty already.
The useful state sits in a narrow band. Too little concern and nothing happens. Too much and people freeze. Understanding without authority produces analysis. Authority without concern produces no urgency. Concern without the capacity to execute produces frustration.
The leaders who do best seem to occupy what I’d call a productive anxiety zone. They can say something like this: I know the future will be different. I don’t pretend to know exactly how. Doing nothing is itself a decision. I have something real at stake, I have the authority to act, and I’m prepared to learn my way forward.
That leader isn’t confident in a prediction. False confidence may be the more dangerous position. What they have is a willingness to act under uncertainty without betting the company on one fashionable technology: bounded experiments, real evidence and repeated adjustment. Strategy then becomes less about choosing a destination and more about building the capability to navigate when the destination moves. The advantage may lie less in how much AI a company has than in how quickly it can become appropriately different.
The real dividing line
Perhaps that is the real divide. It isn’t between companies that use AI and companies that don’t, or between those with an AI strategy and those without one. It’s between enterprises waiting for the future to become clear and enterprises learning to move while it’s still unclear.
Where I sit: I advise through Kainora and am building technology for companies trying to make that move, so I have an interest in how leaders answer these questions.
Two questions for you. Where does your leadership team actually sit, and where would you need it to be? And who among you owns the consequence of getting this wrong?
The future is a fog. The advantage belongs to those who learn how to move through it.
Sources and further reading
- McKinsey, The State of AI in 2026, August 2026.
- EBITDA Up, Value Down, What to Commit To When You Can’t Predict, Your AI Is Getting Smarter. Your Company Isn’t., Busy With AI, The Forgotten Business of America and Underwriting the Heroics, earlier essays in this series.