The meeting was the last one before lunch, and everyone in the room could feel it. Fourth-quarter review, a long agenda, and near the bottom, an item three letters wide: AI.
The CEO stood when it was his turn. He was good at this part. He had run companies for twenty years, and he knew how to walk a board through a story. He told them where the company had been a year ago, nervous and behind, and where it stood now. Then he brought up the slide.
It was a good slide. One logo for every function. Engineering on AI coding assistants. Marketing generating in an afternoon what used to take a week. Support deflecting a third of its tickets before a human ever saw them. Sales with a model scoring the pipeline every night. He walked it left to right, function by function, and by the time he reached the far edge the company was, by any fair reading, an AI company now.
Heads nodded. A few directors asked the questions directors ask. What did it cost. Who owns it. Are we ahead of our competitors or behind. The CEO had an answer for each. He was, I remember thinking, completely in command of the material.
Then a voice from the far end of the table.
She was one of the newer directors, added the previous spring, quieter than the rest. The kind who reads the deck the night before and comes in with the pages already marked. She had not said much all morning. She waited until he finished.
“Where do we have AI,” she asked, “relative to where we actually make decisions?”
The CEO had a slide for everything except that one.
She didn’t ask it to trap him. You could tell from her tone. She asked it the way you ask a question you already suspect the shape of, and hope you’re wrong.
So she drew it out for the room. Forget the tools for a second, she said. Take the handful of decisions this company makes in a year that actually move the enterprise. Not the thousand small ones. The few big ones. What we charge. Where we point the next two quarters of engineering. Which customer segment we chase, and which we quietly stop serving. The decisions that, a year from now, we will trace everything back to.
Now put the AI on top of that. Where does it sit?
The room did the exercise. You could watch it happen. And the answer, once everyone saw it, was almost funny. The AI sat everywhere except there.
Every tool on that slide made execution faster. Work the company had already decided to do, done quicker and cheaper. None of it went anywhere near the decisions that determined whether the work was worth doing at all. They had spent real money speeding up the cheap calls and left the expensive ones running on the same instincts as five years ago.
Here is the honest defense of what the CEO built, because it deserves one.
He was right to start where he started. You do not hand the biggest decisions in the company to a new tool on day one. You build trust in low-stakes places first. Let it draft the campaign, score the lead, close the easy ticket. Watch where it fails. Learn its texture before you let it near anything that matters. Every sensible operator I know did exactly this, and so did I. Adoption at the edges is the right way to begin. It is how you earn the right to move inward.
The trap is stopping there and calling it a strategy. The edge is where you begin, not where you settle. What the director caught was a company that had mistaken the on-ramp for the destination. A year of real progress, genuine and worth celebrating, that had quietly stalled at exactly the layer where the leverage wasn’t.
I know the shape of that stall because I built one myself.
A few years back I was advising a company through its own version of this, and I was proud of the map we made. We had gone function by function, found every workflow AI could accelerate, and put real numbers next to each. Hours saved. Cost taken out. It was clean work, and the client loved it.
It took me longer than I would like to admit to notice that not one item on my map touched a decision the executive team actually made. I had optimized the whole floor of the building and never gone upstairs. On a consulting engagement, I had done exactly what the CEO did with his logo slide. When I finally saw it, I was a little embarrassed. The map wasn’t wrong. It was aimed at the cheap part of the problem, because the cheap part is the part that is easy to measure.
I have sat in a version of that boardroom more than once now, and there is a pattern to what happens after.
The CEO walks in the first time with a story. Here is what we adopted, here is what it cost, here is the momentum. It is a narrative, and usually a good one. Then a board starts asking the director’s question, meeting after meeting, and something shifts in how he prepares. The next time he comes back, he is not telling a story. He has brought something closer to a model. Here are the six decisions that move this company. Here is where judgment still runs on gut. Here is the one we are going to put real intelligence behind first, and here is how we will know if it worked.
The board notices the difference immediately. The questions get sharper on both sides of the table. Nobody is admiring logos anymore.
I asked her once, after a different meeting, why she kept pressing when everyone else seemed satisfied.
She said something I have kept. Adoption is easy to show and easy to fake, she said. Any CEO can fill a slide with logos in an afternoon. Her job on the board was not to confirm the company was adopting. It was to ask where the intelligence sat relative to where the leverage was, and to keep asking until the two lined up. That was the whole job. Everything else was theater she could see through.
What I did not expect was what the question did to the rest of the company.
Once the top of the house started asking where AI sat relative to the real decisions, the question rolled downhill. A VP who had been proud of her team’s ticket-deflection numbers started asking which of her own recurring decisions ran on habit. An engineering lead stopped bragging about coding speed and started asking whether the roadmap that speed served had been chosen well. The director had aimed one question at the CEO. It ended up rearranging how three levels below him thought about their own work. That was the part no one planned.
The CEO didn’t have an answer that day. He had something more useful. A question he couldn’t unhear.
If you sit on a board, the next agenda already has an AI update on it. You can let it be a tour of tools, logo by logo, everyone nodding. Or you can ask where the AI is relative to where the decisions get made, and watch how long the room goes quiet.
The layer that sits above the tools and decides what they are for is the one almost no company is building on purpose. That is the layer Kennesys builds.