2 min lesson
Instrumenting the 'current AI usage' dimension
Answer this as if it were happening now: "Reading the admin analytics, a team shows HIGH request volume but a LOW acceptance rate of committed AI code. What does that pattern mean and what do you do?" Say what supports your choice.
Step 1 of 2
Instrumenting the 'current AI usage' dimensionread adoption, don't guess it
'Current AI usage' is the dimension reps fudge the most. They log 'they use some AI' and move on. There's a concrete instrument for reading it. Once a customer is in (or in a pilot), the admin analytics give you two signals you read together: a team's or person's request volume and their acceptance rate of committed AI code. Plot them as a quadrant and each cell prescribes a different action.
- Signal pattern
- High volume + high acceptance
- What it means
- Power users. They've found the workflows that work.
- Your move
- Learn how they prompt (shared knowledge / shared transcripts), then distribute those learnings across teams.
- Signal pattern
- High volume + low acceptance
- What it means
- Trying hard, getting poor output.
- Your move
- An enablement and training opportunity, not a churn risk. Coach the prompting.
- Signal pattern
- Low volume
- What it means
- Not really adopting yet.
- Your move
- Drive awareness and adoption; find the one workflow that hooks them.
| Signal pattern | What it means | Your move |
|---|---|---|
| High volume + high acceptance | Power users. They've found the workflows that work. | Learn how they prompt (shared knowledge / shared transcripts), then distribute those learnings across teams. |
| High volume + low acceptance | Trying hard, getting poor output. | An enablement and training opportunity, not a churn risk. Coach the prompting. |
| Low volume | Not really adopting yet. | Drive awareness and adoption; find the one workflow that hooks them. |
Request volume × acceptance rate. Admins can enable shared knowledge / shared transcripts at team level to study exactly how a power user works.
People are blind to what's automatable in their own workflow. The canonical story: a Cursor engineer hand-posted a daily usage update to Slack for weeks before a teammate said 'why don't you just automate this?' (now an automation queries the data via MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. and posts it). Probe in discovery for the repetitive thing a champion does by hand every day. That's the most fertile ground for a quick, sticky early win.