BestAngle

Insights

What holds up in practice.

Notes for engineering leaders on working with coding agents, improving delivery, and deciding where AI fits.

Engineering leadership

Just Enough Process in the Age of AI

AI can help an engineer open a pull request sooner. If review and integration still take three days, customers may see little difference. The team may even spend more time sorting through changes.

Pick one recent change and trace it from the first decision to production. How long did it wait for clarification, review, or a release? Give the largest delay an owner and test one change. Track lead time alongside rework and defects so that a faster step doesn’t hide a slower system.

Lessons from BookMaestro

Give agents a definition of done.

Describe the user behavior you expect and give the agent examples it can test. Keep each issue on its own branch, review its changes independently, and check that the tests cover the behavior you asked for.

A branch per issue gives reviewers a clear boundary. It also makes it easier to abandon a bad approach without pulling unrelated work out with it.

AI adoption checklist

Five questions before you start a pilot

  1. Which workflow are we improving, and who owns it?
  2. How does the workflow perform today?
  3. What data may the tool use, and where may it go?
  4. Which outputs need human review before they reach users?
  5. What result would justify expanding the pilot?
Explore the AI Readiness Assessment →