In 30 seconds
- AI makes it cheap to add code, including bad code.
- Unreviewed AI code is a fast route to debt.
- AI is also a strong tool for refactoring and tests.
How AI creates debt
- Duplicate code generated instead of reusing existing code
- Inconsistent patterns across files
- Code nobody on the team fully understands
- Large changes merged with shallow review
How AI reduces debt
- Writing missing tests around legacy code
- Mechanical refactoring at scale
- Explaining old code before it's changed
- Updating dependencies and deprecated APIs
Habits that keep code healthy
- Small changes with real review
- Clear conventions the AI is given and checked against
- Track duplication and complexity trends
- Book some AI time for debt reduction every sprint
See technical debt.
Myth vs reality
Myth · AI code is debt-free because it's new.
Reality · New code can be debt on day one.
Myth · Debt matters less if AI can rewrite anything.
Reality · Rewrites still break behaviour and need review.
Myth · More code means more progress.
Reality · More code often means more maintenance.
FAQ
How do we spot AI-driven debt?
Rising duplication, inconsistency and slower changes over time.
Should we limit AI use?
Limit unreviewed output rather than AI itself.
Can AI help with legacy systems?
Yes, especially for explaining code and adding tests.