Review a pile of PRs that your coworkers, both human and AI, have produced, checking for gotchas, product direction, alignment, and architectural concerns. You don't trust AI code outright. You internalize it, think about it, and understand its implications and tradeoffs.
