If you're trying to build a culture where AI experiments actually happen, steal this practice.

At the end of every sprint or monthly review, require teams to share one thing that didn't work — with the same level of detail as the wins. What was the hypothesis? What happened? What did you learn?

You're not celebrating failure. You're institutionalizing learning.

The signal you send about misses shapes what people are willing to try.