A regional insurance company I sat with recently was doing most things right. One of their claims teams had built a genuinely good AI workflow, the kind that took a multi-day process down to an afternoon and actually held up under scrutiny. The team lead walked me through it and you could hear the pride in her voice. It was excellent work.
Then I asked what felt like a simple question. Had anyone else in the company picked it up? She thought about it and said she wasn’t sure. As it turned out, another group two floors away had spent the previous quarter solving nearly the same problem from scratch. Neither team knew the other existed. The win was real, but orphaned.
I’ve come to believe this is the most expensive pattern in enterprise AI right now, but it remains hidden. Companies are generating real wins. The wins just have nowhere to go.
We Keep Expecting AI to Install Itself
Here’s the part that blows my mind.
We seem to have decided that AI, the most powerful technology most of us will adopt in our careers, is somehow going to install itself. Spread itself. Get learned on people’s own time, in the margins of their real jobs. That we can reshape the organization while asking for almost no new human infrastructure in return.
It won’t. AI doesn’t install itself.
We actually know this, because we’ve done it before and done it better. Think back to the 1990s, when we put a PC on every desk. We didn’t just hand people a beige box and tell them to go play. We stood up IT departments. We built help desks. We created whole roles whose entire purpose was to help humans absorb a new capability into how they worked. The technology was the easy part. The human infrastructure around it was the actual project.
With AI, a lot of organizations have skipped that step. We’ve done the equivalent of dropping the PCs on the desks and walking away, and then we’re surprised when the value stays trapped in the few teams motivated enough to figure it out alone.
The Layer Nobody Drew on the Org Chart
If you sketch how most AI transformations are actually built, you’ll usually find two layers. At the top there’s strategy, the North Star, the executive sponsor, the budget. At the bottom there are the teams doing the real work, running pilots, finding what actually helps. Both of those layers get a lot of attention.
What’s almost always missing is the layer in between, the one that runs sideways across the organization and connects those teams to each other. Without it, every win is vertical. It goes up to leadership as a nice slide, and it stays down in the team that built it. It never travels across to the six other teams who could use it.
Toyota has a word for the thing that’s missing here. They call it yokoten, which translates roughly as “across everywhere.” It’s the deliberate, horizontal, peer-to-peer spread of an improvement from the team that discovered it to every other team that could use it. And the part I love most is that yokoten doesn’t just copy the result. It copies the thinking that produced the result, because the reasoning is what lets the next team adapt it instead of blindly pasting it in. That distinction is the whole game.
When did your organization last take a genuine AI win from one team and deliberately walk it across to another?
The One Piece of Infrastructure I’d Put In First
So let me let you in on one concrete example of the human infrastructure that has to be built, because this is the piece I’d put in first. It’s what I call an AI Activation Hub.
An Activation Hub is a small, deliberate network whose actual job is horizontal coordination. It centralizes the growing pile of expertise, separates the real signal from the noise, and routes what’s useful to the specific people who need it. When that claims team builds something that works, the Hub is the reason the other team two floors away hears about it. It’s yokoten, given a home and an owner inside a modern company.
And the reason it matters so much right now is that individual AI talent has raced way out ahead of organizational capacity to use it. One recent survey found that AI super-users are delivering roughly 5x the productivity of their peers, and yet only 29% of organizations report significant ROI from generative AI.
The talent is already in the building. The results are real and enormous at the individual level. What’s missing is any mechanism to spread them. A separate reading from Microsoft’s 2026 Work Trend Index tells the same story from the other side: only about one in ten workers say AI has actually changed how work gets done, while three times as many executives are convinced they’ve already achieved enterprise-wide impact. The wins are staying local, and the people at the top can’t see it.
An Activation Hub is one answer to that. It doesn’t replace the rest of the support system, and it doesn’t work as a governance gate that every project has to pass through for approval. What it does is act as the connective tissue that lets a good idea compound instead of dying where it was born.
Where This Goes
The shift is this: stop treating AI as a tool you install and start treating it as a capability you have to help the organization absorb. That means building the human infrastructure on purpose, and the coordination layer is the piece almost everyone skips. Get it in place and your wins stop being lucky one-offs and start becoming how the whole company learns. That’s a real part of the road from where you are now to genuinely AI-native.
That coordination layer is exactly what we build together inside the Running Hyperadaptive Organizations class, where we bring Hyperadaptive concepts to life. You can find more information at hyperadaptive.solutions/class.
And if you want the full picture of how the pieces fit, that’s what the book is for. Hyperadaptive: Rewiring the Enterprise to Become AI-Native is available now at hyperadaptive.solutions/book.
Great article, Melissa. Seems to me it's the Layer 3 (Social layer) discussed in Gene Kim's "Wiring the Winning Organisation" that needs some attention.