2 min lesson
The recommended first automation
Make the call in this situation. "An exec asks you to report 'lines of AI-generated code' as the pilot's headline metric. What's wrong with it and what do you offer instead?" Explain what supports it.
Step 1 of 2
The recommended first automationlow risk, high stickiness
When a pilot team asks 'what's the first automation we should set up?', there's a data-backed answer: 'Summarize Changes Daily.' It posts a Slack engineering digest of notable repo changes each morning. Cursor's data science found it the stickiest automation at 89% retention. It's the ideal opener. Its blast radiusHow much breaks if a change goes wrong; the scope of potential damage. Press Enter for the full definition. is low, because it reads and reports and never touches code. It's visible to the whole team, and it builds the daily habit that makes everything above it on the ladder feel normal.
To use cloud agents, start guided setup from the Cloud AgentsAgents that run in a Cursor-managed virtual machine, check out the repo, do the work and open a pull request, then shut down, with no load on your laptop. Press Enter for the full definition. dashboard or the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition. in the desktop app: connect your Git provider, select the repos and add the env vars and secrets the agent needs to install dependencies. Cursor's docs say it can set up your dev environment in the cloud in less than 10 minutes, so a first run that takes a few minutes shouldn't spook the champion.
Once the agent has the code working, save a snapshot of its virtual machine so future agents start from it instead of rebuilding, and commit the configuration to .cursor/environment.json so the whole team reuses it. After that, engineers move freely between local and cloud from the client.
You measure against a four-lens scorecard and you set the baseline before you start. A pilot with no day-zero baseline proves nothing, because every result answers 'compared to what?'
- Lens
- Adoption & capability
- What it captures
- Are people using it and at what rung?
- Example signals
- WAU/DAU, ladder rung reached, % on governed workflows
- Lens
- Flow
- What it captures
- Is delivery actually moving faster?
- Example signals
- Cycle time, PR-to-merge, deploy frequency
- Lens
- Quality & safety
- What it captures
- Are we shipping safer, not just faster?
- Example signals
- Escaped-defect rate, BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. catches, policy coverage
- Lens
- Experience & trust
- What it captures
- Do engineers and security trust it?
- Example signals
- Developer sentiment, security sign-off, retention of usage
| Lens | What it captures | Example signals |
|---|---|---|
| Adoption & capability | Are people using it and at what rung? | WAU/DAU, ladder rung reached, % on governed workflows |
| Flow | Is delivery actually moving faster? | Cycle time, PR-to-merge, deploy frequency |
| Quality & safety | Are we shipping safer, not just faster? | Escaped-defect rate, BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. catches, policy coverage |
| Experience & trust | Do engineers and security trust it? | Developer sentiment, security sign-off, retention of usage |
Four lenses, baselined day zero. 'Lines of AI code' is NOT a lens; it's the wrong headline.
Learn more
Optional practice
Test yourself on The recommended first automation
QA pilot team in the Expand phase asks 'what's the first automation we should turn on, and how long will cloud onboarding take?' Give them a concrete, low-risk answer.