The pattern is familiar: AI tools roll out, early adopters love them, usage grows for a few months—then plateaus. Maybe 30% of the target population uses the tools regularly. The rest tried once, didn’t click, and drifted back to old habits.

That plateau is a Reinforcement failure. ADKAR’s Reinforcement phase is about sustaining the change—making the new way of working stick. Without it, adoption decays.

Why Plateaus Happen

Early adopters ≠ mainstream.
The people who adopt first are often the ones who were already curious, already experimenting. They have high intrinsic motivation. The mainstream needs different conditions: clearer payoff, lower friction, social proof. When we optimize for early adopters, we design for the wrong population. Reinforcement for the mainstream requires different incentives and support.

Novelty wears off.
Initial enthusiasm—“this is cool!”—fades. What remains has to be useful: the tool needs to earn its place in the daily workflow. If it doesn’t—if it’s marginally helpful or situationally useful—people stop reaching for it. Reinforcement means the tool keeps delivering value, not just a one-time wow.

Reverting is easy.
The old way is still there. The old habits are ingrained. Without ongoing reinforcement—reminders, nudges, integration into mandatory workflows—people slip back. The path of least resistance wins. Reinforcement requires making the new way the default.

No consequences for non-use.
If using the tool is optional and not using it has no downside, many people will opt out. Reinforcement sometimes needs teeth: “we expect everyone to use X for Y” or “we’re deprecating the old process.” That’s politically charged, but without it, voluntary adoption often stalls.

What Reinforcement Looks Like

1. Bake it into the workflow.
The tool isn’t optional—it’s the way we do the thing. Code review happens in the agent-augmented flow. Specs are drafted with the agent. When the new way is the default path, reinforcement is structural.

2. Measure and respond.
Track usage, but track outcomes too. Are people who use the tool shipping faster? Producing better work? Reinforcement includes feeding that data back—“here’s what we’re seeing”—and iterating when the numbers don’t support the narrative.

3. Recognize and reward.
Celebrate people who use the tool well and share their practices. Reinforcement is social: when the org visibly values the new behavior, others follow.

4. Remove the old path.
At some point, the old way has to go. Dual-track—“you can use the new tool or the old one”—prevents reinforcement. Sunset the old process when the new one is ready. That’s the hard part of Reinforcement, and the one many orgs avoid.

The 30% plateau isn’t inevitable. It’s what happens when we stop at Ability and skip Reinforcement. Sustaining change takes as much intention as creating it.