For teams actively using AI coding assistants (Copilot, Cursor, Windsurf, etc.), I'm noticing a frustrating pattern: the more complex your codebase, the more time developers spend preventing AI from breaking things.

Some common scenarios I keep running into:

* AI suggests code that completely ignores existing patterns

* It recreates components we already have

* It modifies core architecture without understanding implications

* It forgets critical context from previous conversations

* It needs constant reminders about our tech stack decisions

For engineering teams using AI tools:

1. How often do you catch AI trying to make breaking changes?

2. How much time do you spend reviewing/correcting AI suggestions?

3. What workflows have you developed to prevent AI from going off track?

Particularly interested in experiences from teams with mature codebases.