An organization wants to learn from a phased AI rollout before expanding. Which two practices create a useful learning loop? Select 2 answers.

1. Feed documented lessons into the roadmap, controls, and user enablement before the next phase.
2. Review outcome, quality, adoption, and incident evidence with the relevant owners.
3. Discard failures so later teams can repeat the same discovery process.
4. Count every rollout phase as successful regardless of evidence.
5. Expand exposure before resolving a known material safety failure.