AI Workflow
Overview
AI workflow is the practical process for using AI tools to accelerate engineering work without losing correctness or judgment.
Why It Matters
AI can speed up search, drafting, and scaffolding, but only if the output is reviewed like code.
Core Concepts
- Use AI for leverage, not authority.
- Verify anything that affects behavior or risk.
- Keep prompts grounded in the actual codebase.
Mental Models
Treat AI as a fast assistant that needs good constraints, context, and verification.
Best Practices
- Ask for targeted help.
- Provide relevant file and system context.
- Review outputs before merging.
Common Mistakes
- Copying AI output blindly.
- Giving vague prompts with no constraints.
- Using AI to skip understanding the code.
Trade-offs
AI can increase speed dramatically, but unreviewed output can also increase risk and noise.
Decision Framework
| Task | AI fit |
|---|---|
| Drafting | Strong |
| Refactoring | Good with review |
| Security-critical logic | Use carefully |
Examples
- Ask AI to summarize a PR before review.
- Use AI to draft a checklist from an incident.
Checklists
- Did I verify the output?
- Is the prompt specific?
- Did AI help me understand rather than replace understanding?
Senior Engineer Notes
Senior engineers use AI to compress routine work, not to outsource judgment.