Human-led AI transformation

Every technology shift comes with the same promise: this one will be different, this one will be fast, this one just needs the right rollout plan. I've spent fifteen years inside Microsoft watching PC give way to cloud, cloud give way to mobile, and now mobile giving way to whatever we're calling this AI moment. The rollout plan is never the hard part. The humans are.

Here's what nobody puts on the slide: organizations don't actually adopt technology. People do, one uncomfortable habit change at a time, and then the organization gets to take credit for it. McKinsey will tell you AI-mature companies see 1.7x the return of their peers. BCG will tell you the first sixty days determine whether a transformation sticks or slides back into old habits. Both are true, and both miss the point. The multiplier isn't the model. It's whether the person using it decided to actually change how they work.

I like to think of this as corporate epigenetics. The organizational DNA — the org chart, the strategy deck, the values poster in the break room — barely moves. What changes is expression: which behaviors get switched on, which habits get switched off, which of the same old people start operating in a genuinely different register. AI doesn't rewrite the genome. It flips switches. And switches only flip because a person flips them, not because IT pushed an update.

Which brings me to the group everyone forgets to design for: individual contributors. Leadership sets the direction, and I'll get to them. But leadership doesn't write the code, run the numbers, or draft the brief at 4pm on a Thursday. ICs do. They are the actual surface area where Humanᴬᴵ either happens or doesn't, and most transformation programs treat them as an audience for a rollout email instead of the fulcrum of the whole effort. Gartner puts the ratio of managers to ICs somewhere around 1:10 in most enterprises. If your change strategy only reaches the 1, you've engineered your own bottleneck.

So what does it actually take to get the 10?

Not a mandate. Mandates produce compliance, and compliance produces the minimum viable use of a tool — the equivalent of using a Ferrari to idle in traffic. What moves adoption is permission paired with proof: seeing a peer, not a slide, get real time back on real work. Deloitte found 70% of workers say they'd use AI more if they simply understood how it applied to their specific job. Not if they had more training modules. Not if leadership sent a more inspiring email. If they could see it work on the thing they actually do all day.

That's the job of leadership in this transition, and it's a narrower job than most executives want it to be. Not "champion AI" in the abstract — model it, specifically, visibly, on your own work, where your team can see you do it. Not "set the vision" — remove the friction that's actually stopping adoption, which is rarely a lack of vision and almost always a lack of time, trust, or a clear enough example. The people who need convincing aren't reading the AI strategy memo. They're watching whether their manager still does things the old way when nobody's checking.

Adaptation velocity — how fast a person or team actually changes how they work, not how fast the tool gets installed — is the real metric here, and it has almost nothing to do with the technology itself. It's a human variable. It responds to trust, to visible proof, to the sense that the person one desk over figured it out and didn't die. It does not respond to a rollout timeline.

Technology has always restructured the environments we operate in — that's not new, and it's not really about AI. What's new is the speed at which the environment is moving underneath us, which means the old approach of waiting for change management to trickle down through the org chart is too slow to matter. The environment is going to change. The only real question is whether you built the conditions for your people to change with it, or whether you're hoping the technology does the adapting for them.