Every AI rollout has a good first month. The keen ones dive in, the usage graph ticks up, someone screenshots it for the steering group, and everyone exhales. Then week six lands and the line goes flat. Not down. Just flat. The people who were always going to use AI already have, and everyone else hasn't budged an inch.

That flat line isn't a glitch. It's the AI adoption curve behaving exactly as it always does. And most organisations read it completely wrong, because they think the curve is something that happens to them rather than something they're supposed to drive.

Sixty years ago, a sociologist called Everett Rogers noticed that new ideas spread through a population in a predictable shape, and he sorted people into five groups by how quickly they take something up. Innovators, the tinkerers, about 2.5%. Early adopters, the keen and influential, about 13.5%. Then the early majority at 34%, the late majority at another 34%, and the laggards bringing up the rear at 16%. Plotted over time it makes the famous bell curve. It's the backbone of pretty much every technology adoption model since.

Here's the bit that matters for your AI rollout. The two eager groups at the front are only about one in six of your workforce. The two majority groups in the middle are two-thirds of everyone you employ. That's the bulk of the organisation, and they behave nothing like the keen crowd at the front. Treat them as if they do and you've lost before you've started.

Innovators and early adopters need no convincing. They were using AI before you bought it a licence. They're the people already smuggling their own tools past IT, the ones who'd have found a way with or without your permission. So when you launch and they pile in, of course the numbers jump. You didn't earn that adoption. They self-selected into it.

And this is the trap, because that early spike looks and feels exactly like success. The graph's up, the enthusiasts are vocal, and it's tempting to declare the thing landed and move on to the next initiative. You haven't landed anything. You've counted the people who never needed persuading and mistaken them for the whole. The real job hasn't even started.

In the nineties, Geoffrey Moore added the crucial wrinkle Rogers missed. Between the early adopters and the early majority there's a gap. A chasm. And most things that fail, fail right there, because the people on either side of it want completely different things.

Early adopters take a punt on something because it's new and interesting. The early majority couldn't care less that it's new. They're pragmatists. They adopt when it's proven, safe, and someone they actually trust has already done it and lived to tell the tale. Hand them the same launch email that thrilled the enthusiasts and they'll do nothing at all. They're not saying no. They're saying "show me it works for someone like me first", and then they wait.

That waiting is the chasm. It's the same gap the pillar calls the missing middle: the stretch between "I know this tool exists" and "I choose to use it", the part of the journey nobody budgets for. And the numbers show how few organisations ever get across it. Only 7% have actually scaled AI across the business, 95% of generative-AI pilots delivered no measurable impact, and 42% of companies have now abandoned most of their AI initiatives. That's not a technology failure. That's a workforce full of organisations that won the keen few, hit the chasm, and stopped.

The reflex when the middle won't budge is to call them resistant. They're not. They're rarely ever resistant. They just want different things than the enthusiasts, and you never gave them any of it.

The early majority want proof and a reason. Not "AI boosts productivity", but the specific hour it hands back to someone doing their exact job, demonstrated by a colleague they rate. The late majority are warier still, and a good chunk of them aren't sceptical so much as flat-out worn out from the last five things that got hyped and quietly abandoned. The laggards will come last whatever you do, and that's fine, because you don't need everyone on day one.

The point is that these are four different audiences with four different objections, and the standard rollout aims one all-staff email at the lot of them. That isn't a launch. It's a notification. And notifications don't move pragmatists across a chasm.

Here's where most leadership teams get the curve fundamentally wrong. They treat it like weather, something that unfolds on its own if they're patient enough. It doesn't. Adoption curves get crossed on purpose, by someone doing the work, or they don't get crossed at all.

The work is marketing, aimed at your own people. You segment the curve instead of blasting it: give the enthusiasts room to run, then use them as the proof the early majority are waiting for. Because the way you cross a chasm is with social proof, and the early majority copy trusted peers, not steering-group memos. Find the champions, capture the specific local wins, and make them impossible to miss. Answer "what's in it for me" in each team's own language. Then keep going, because the middle needs the message more than once. That's the whole adoption playbook, and it's the difference between a rollout that stalls at 16% and one that reaches the two-thirds who were always the actual prize.

Quick clarification, because these two get muddled constantly. The adoption curve tells you who takes AI up and when, from the innovators through to the laggards. The change curve is a different animal: it maps the emotional journey a person rides through any disruption, from shock and resistance to eventual acceptance. One is about the spread across a population. The other is about what's going on inside a single head. You need both lenses, because someone can be sat bang in your early majority and still be stuck at the bottom of their own emotional dip, waiting for a reason to climb out.

The keen few were free. They cost you nothing and they'd have adopted regardless, so counting them as a victory is like taking credit for the sunrise. The majority in the middle is the entire game, and they don't move on enthusiasm. They move on proof, relevance, and the sense that someone actually thought about them.

So the next time an AI usage graph ticks up in week one, hold your applause. The question that matters isn't "how many jumped in early". It's "what's your plan for the two-thirds still standing on the wrong side of the chasm". If you don't have one, that flat line in week six is already booked in.