The AI Change Curve Nobody Is Managing

I've led enough transformations to know that every large-scale change moves through the same arc. It opens with energy and executive attention, then enters a harder stretch where the novelty wears off, competing priorities crowd back in, and the people expected to work differently start making quiet decisions about whether this is real or just the latest thing leadership is excited about. Anyone who has led an organization through a system implementation, a restructuring, or an operating model shift has watched it happen. The pattern is well understood, and the leaders who catch it early lead through it. The ones who don't watch the momentum drain away and spend the following months trying to reconstruct what went wrong. 

AI adoption is moving through that same arc right now, and most organizations are managing it as if it were exempt. The launch looked like progress: sponsorship from the top, tools in people's hands, training calendars filling up, pilots that worked, real enthusiasm in the room. It felt different from past transformations because the technology was genuinely exciting and the early results were genuinely real. But what I'm seeing across most enterprises now is the same thing that happens in any change that is not deliberately led through its middle: the energy that carried the launch has run out, and nothing has been built to replace it. 

How the plateau shows up 

Usage flattens. The early adopters who drove the first wave move on to the next interesting thing. The broader organization, never fully convinced, drifts back to how it worked before, often without anyone deciding to. The tool that impressed the leadership team in a March demo is the same tool sitting unused in June. And leadership, still anchored to the energy of the launch, is slow to register that the dynamic has fundamentally changed. By the time the plateau is visible in the numbers, the organization has usually been sliding for a quarter or more, long before anyone names it as a problem. 

Why the AI plateau gets misread 

What makes this arc dangerous is that the signals are easy to misread. When adoption stalls after a restructuring or a system migration, leaders generally recognize what they are looking at, because they have watched a change effort lose altitude before. When AI adoption stalls, the conversation goes somewhere else entirely. It gets blamed on the technology not being ready, on people needing more training, or on the use cases not being the right ones. That's not a technology problem; it’s a leadership one. The effort has entered the most predictable and hardest phase of any transformation, and no one has changed how they're leading it. While leaders debate models and use cases, the return on the investment quietly erodes. 

The phase that decides everything 

The early phase of any change runs on novelty, executive attention, and the charge that comes with something new. None of that is built to last, and none of it was ever supposed to. The middle phase runs on something else, and it has to be built on purpose. In our work leading organizations through enterprise-wide change, the difference between the teams that cross this gap and the teams that stall comes down to four things the stalling teams never put in place. 

1. Reset the expectation from optional to standard. If the breakdown is that people still treat the new way as an experiment, the fix is to make it the baseline. A pilot invites people to opt in; a standard tells them what good work now looks like. In practice, that means it shows up in what leaders ask for, what they review, and what they treat as finished. Until that happens, employees will reasonably treat it as optional, and optional things lose out to whatever is urgent. This is also where the productivity and quality gains show up: not when people try the tool, but when the expected output changes. 

2. Make reverting visible and hold people accountable. If the breakdown is that slipping back to the old way carries no cost, the fix is to give it one. In most organizations, quietly reverting is the path of least resistance, and because no one tracks it, there is nothing to stop it. Visibility comes first: when a team's return to the old workflow shows up in how work is reviewed and discussed, the quiet decision to opt out stops being quiet. But visibility only works if it leads to accountability. Treat the new way as a standing expectation, name it when a team drifts back, and hold people to it the way you would hold them to any commitment that mattered. 

3. Remove friction instead of repeating the message. If the breakdown is that leaders keep advocating for change without clearing the path to it, the fix is to spend less time selling and more time taking obstacles out of the way. Most plateaus are not motivation problems; people have heard the message. What stops them is friction: an approval that takes longer in the new way, a system that does not connect, a policy written for the old process. What moves adoption at this stage isn’t a better pitch. It’s a shorter path. 

4. Wire adoption into how people are measured and rewarded. If the breakdown is that adoption still depends on enthusiasm, the fix is to build it into the systems that outlast it. How people are measured, developed, and recognized is what tells them what the organization actually values. When none of those systems reference the new way of working, people correctly read adoption as a campaign that will pass. When they do reference it, the new way stops being an initiative competing for attention and becomes part of the operating environment. 

The bottom line 

In our work leading organizations through large-scale change, the difference between transformations that produce results and those that stall is almost never the quality of the strategy or the sophistication of the technology. It is whether leadership recognized the shift from the launch phase to the middle phase and changed how it led in response: resetting expectations, making reversion visible, clearing friction, and building adoption into how people are measured and rewarded. Skip that work, and every transformation follows the same path: strong launch, quiet plateau, slow fade. AI is no exception. The organizations that earn a return on what they have already spent will be the ones that lead AI adoption with the same phase-awareness and rigor they would bring to any enterprise transformation, because that is precisely what it is. 

Andrea Schnepf 

P.S.: We are digging into exactly this on June 17: the operating conditions and leadership behaviors that carry AI adoption through the phase where most organizations lose momentum. Join us: Embedding AI Into How You Work: From Adoption to Impact.