Concentrated AI adoption hides well. Usage numbers look fine. A few engineers are visibly fast. Nobody is complaining. It takes a specific look to notice that "the team uses AI" actually means "three specific people use AI."
Here are the signs worth checking for.
You can name the same people every time
If someone asks "who's using AI well on this team," and the same two or three names come up every time, regardless of who is asking or when, that is concentration. Broad adoption produces a longer, more varied list. Concentrated adoption produces a short, stable one.
Usage dashboards look better than actual workflows
Seat activation and session counts can look reasonable while usage stays shallow. An engineer who opens the tool twice a week to ask a quick question shows up as "active" in most dashboards, even though nothing resembling a real workflow exists for them. Dashboards measure access, not depth.
AI-assisted PRs all come from the same few authors
Look at your last month of merged PRs. If AI involvement, visible through commit messages, PR descriptions, or just knowing the team, clusters around the same few names, that is concentration showing up in shipped work, which is the version that actually matters.
New hires learn old habits, not new ones
Watch how a new engineer gets onboarded. If nobody mentions the team's AI workflow during onboarding, or the explanation is vague ("some people use Copilot, do what works for you"), the workflow is not established enough to teach. It is still living in a few individuals' habits.
The fast engineers are the same ones who were always fast
This is a subtler signal. If your visibly AI-accelerated engineers are the same people who were already your fastest, most tool-curious engineers before the rollout, the tool amplified an existing gap rather than closing one. That is not necessarily bad, but it means the rollout did not do what most leaders hope it does, which is lift the team's floor, not just its ceiling.
Nobody has shown anyone else their workflow
Ask directly: has any engineer walked another engineer through how they actually use the AI tool day to day? In teams with real adoption, this happens naturally, in code review comments, in pairing sessions, in casual Slack messages. In concentrated adoption, usage stays private. People do not share what they are not sure is a "real" way to use the tool.
The gap has not moved in months
If you rolled this out three or six months ago and the same small group is still the only visible group, the gap is not closing on its own. Concentrated adoption tends to compound, not resolve, because the early adopters keep getting better while nobody else has a reason to change.
What to do once you spot it
None of these signs mean the tool failed. They mean the rollout stopped at access instead of continuing to a real workflow. The fix is not more licenses or another announcement. It is turning what your early adopters figured out into something the rest of the team can actually learn and follow, with someone accountable for making sure it spreads.
If several of these signs sound familiar, book a call and I'll help you figure out what would actually close the gap. If I'm not the right person, I'll say so.