I keep having the same conversation in boardrooms, on Zooms. Over cocktails.
Smart executives, well-resourced teams, trusted colleagues, treating AI like a software rollout. Pilot it. License it. Train people on it. Move on.
That isn’t what’s happening.
So I went back.
Back through Stanford’s lectures on large language models. Back through research from OpenAI, Anthropic, DeepMind. Back through decades of helping technology move from novelty to necessity. To my own work on Intel Inside.
What I learned working on Intel Inside is that real transformations can happen from the inside out. What I’m seeing now is that Intelligence Inside is that pattern at a different scale. Cognition is inside the work itself now — not a tool added to it, but the substrate the work is being made of.
The job then was making a complicated, accelerating technology legible to the people who had to live with it. The job now is the same. But this curve makes Moore’s Law look gentle.
Everyone is working with one or more models now. Language has become programmable, and language sits at the center of most knowledge work. The question isn’t which model you use. It’s what you build around it — workflow, context, judgment.
Pulling people’s heads out of the sand — or out of the frontier models — isn’t easy. Asking an accomplished agency to allow us to observe and document their workflows so we can redesign them can be uncomfortable. Delivering strategic work in hours that used to take months changes more than the conversation. It changes the underlying business model. And as execution and optimization become commoditized and automated, judgment becomes leverage. Taste becomes leverage. The rest follows. But how then will the money flow?
A few weeks ago, my fifteen-year-old son, Henri — about to finish ninth grade, thinking about a summer job — asked if he could take over Makeable when he graduates college.
I told him:
I’m not really sure what Makeable is going to be when you graduate college. But if you can help me figure it out, sure.
That answer wasn’t humility. It was math.
The companies that adopt an AI model the fastest — or at the largest scale — may not be the ones that win. It’s the ones who understand where intelligence actually belongs. Not as a feature. Not simply a tool. But inside the work itself.
More Signal. Less Noise.