Most leaders trying to fix AI adoption skip a step: finding out what is actually happening right now.
They have a general sense. A couple of engineers seem fast. Someone mentioned using Copilot in standup. The license dashboard shows everyone has access. That is not a map. That is a guess dressed up as data.
You cannot fix uneven adoption without knowing where it is uneven.
Why the obvious approach doesn't work
The obvious move is a team-wide survey: "How often do you use AI coding tools, on a scale of 1 to 5?" This produces numbers, but not accurate ones.
Engineers who feel behind round up because they do not want to look like they are falling behind. Engineers who lean on the tool heavily sometimes round down, especially if there is any stigma around "AI wrote my code." Occasional, low-effort use gets forgotten entirely. The result is a chart that looks precise and means very little.
You need a method that does not rely entirely on people accurately rating their own habits.
A practical way to map real usage
Start with whatever usage data your tools expose. Most AI coding tools have some form of admin dashboard, even if it is limited to session counts or active seats. It is not the full picture, but it tells you who is opening the tool at all, which is a useful first filter.
Sample recent PRs. Pull ten to fifteen recent pull requests across different engineers and look for signals of AI involvement: commit messages, PR descriptions, code patterns, or just asking the author directly. This tells you more than a dashboard, because it shows whether usage is turning into shipped work.
Talk to five or six engineers one-on-one. Not a survey. A short, specific conversation: "Walk me through the last time you used [tool] on something real. What did you use it for? Did it help?" People are far more honest in a low-stakes conversation than on a form that might get shared with their manager.
Look for patterns, not individuals. You are not trying to build a scorecard for performance review. You are trying to understand: is usage concentrated in two or three people? Is it broad but shallow, everyone dabbling without a real workflow? Is it trending up, flat, or fading after an initial spike?
What the map usually shows
In most teams I have looked at this with, one of two patterns shows up.
The first is concentration: two or three engineers use the tool constantly and effectively, everyone else uses it occasionally or not at all. This means the problem is spread, not awareness. The team needs a shared workflow, not another announcement.
The second is shallow breadth: most of the team has tried the tool, usage shows up in the dashboard, but nobody has a real workflow. People ask it quick questions or use autocomplete, but nothing resembling a repeatable practice exists. This means the problem is depth. The team needs a specific standard to aim for, not more access.
These require different fixes. Treating a concentration problem like a depth problem, or the reverse, wastes the effort of fixing it.
What good looks like
A useful map does not need to be exhaustive. It needs to answer three questions clearly: who is using AI tools well, what are they actually doing with them, and is that pattern spreading or staying put. If you can answer those with real evidence instead of impressions, you know what to fix.
If you want a structured way to get this picture instead of guessing, book a call. If I'm not the right person to help, I'll say so.