Signals from the Curve — Issue 003 (OID)
Wisdom that outlasts the algorithm.
THE CURVE
Most organizations don't have an AI problem. They have a map problem.
Researchers at the Karolinska Institute just ran more than 1,200 people through an innovation game. One group understood how things connect, what fits with what. The other got the same task with that meaning stripped out. The first group innovated. The second, even when they could watch and copy everyone around them, performed no better than a random bot. (PNAS, May 2026.)
Motivation didn't rescue the second group. Watching the room and copying what worked didn't rescue them either. Without a map of how things connect, more effort just searched the dark faster.
Now swap "people" for "your company." Everyone you compete with rents the same models you do. The intelligence is commoditized. What isn't commoditized is whether your organization has a map: a working picture of how the place actually thinks.
Here's the part most leaders miss. Every org runs on two operating systems:
The visible one is the org chart, the documented processes, the manuals.
The invisible one is the tacit layer, the unwritten rules, the "ask Dave, he'll know" knowledge that quietly keeps everything running.
The real intelligence lives in the second system. Strip away the map of how things connect, and even capable people innovate at random. That is what the study actually measured.
This is exactly the layer companies skip when they deploy AI. They point a powerful model at the documented system, the visible one, and they get the corporate version of that second group. Lots of capability. No map. Motion without innovation.
> When everyone has the same engine, the map is the edge.
The discipline of building that map on purpose - of making an organization's invisible intelligence visible before you hand work to a machine - is what I call Organizational Intelligence Design. Not org design, which is boxes and lines. Not "AI adoption," which is renting the engine. OID is the layer in between: the connective intelligence that decides whether a model's raw horsepower becomes your company's specific breakthroughs or just faster noise.
It reframes the whole agent conversation, too. We keep asking how to build better agents. The sharper question is what has to be true around an agent before it can be trusted with real work. An agent dropped into an organization whose real intelligence is invisible is the second group in the study, every time.
The research even names the trap on the far side. Too rigid a map, and you start missing the counterintuitive combinations that become the breakthroughs. So the goal was never a frozen map. It's a living one. Offload the routine connections to the machine, and protect the human room to chase the unreasonable idea.
Some orgs design for that compound. Others bolt AI onto the org chart and call it strategy.
THE SIGNALS
→ Half-Baked. [ FIRST WAVE ] The 2026 language shift from "AI adoption" to "organizational readiness." A handful of CEOs and the big-house reports (IBM, Deloitte) started framing this year's AI work as a management problem, not a deployment problem. Watch which leaders go one layer deeper and start treating the invisible operating system, not the org chart, as the actual project. That's the tell that they get it.
→ Hot Take. Most "AI readiness" assessments measure the wrong layer. They audit the documented system, the processes you could already find in a binder. They never touch the tacit layer where the real intelligence lives. A readiness score that only sees the org chart is grading the map of the building, not the people who actually know where the exits are.
→ Confession. I keep wanting to fully map the invisible system, to get it all written down and machine-ready. The Karolinska paper just reminded me why I shouldn't. Over-codify, and you freeze the priors that produce the next breakthrough. The map has to stay alive. That's harder than finishing it.
THE NEXUS
> What does your organization know that it has never written down? And what happens to that knowledge the first day you hand the work to an agent?
If you can't answer the second question, you haven't found your AI strategy yet. You've found your risk.
Reply if you see this differently.
THE MONDAY MOVE
Pick one recurring decision your team makes well. A renewal call, a triage, a go/no-go.
Don't document the process. Everyone documents the process, and the process is the visible layer, the part that was never the hard part. Instead, document the signals: what does your best person actually check before they decide? What do they notice that they've never been asked to explain?
That short list, the signals behind one good decision, is the smallest unit of organizational intelligence an AI can actually learn from. It's also the part that walks out the door when that person leaves.
One decision. One list of signals. Monday. That's the whole move.
THE COMPOUNDING ASSET
I wrote up the other half of this argument: you can't extract a company's intelligence into a database, you have to design the system that holds it. It's the framework sitting underneath today's issue.
The grounding
This newsletter is called Signals from the Curve because there are two kinds of forecasting: the curve and the cliff. The cliff says everything changes at once. The curve says it's already changing, you just have to know where to look.
I write here every Wednesday morning about AI strategy, leadership, and the parts of work that compound. If something here changed how you're thinking, hit reply. I read every response.
— Chris
Chief of AI & Strategy at Essential Innovations · Founder, Attainable AI · Adjunct Faculty, Columbia University