Plan for failure

One-minute script

An AI tool can be unavailable, wrong, or unexpectedly different after a change. A safe workflow needs a practical way to stop, continue, and recover without improvising under pressure.

Try this

Name a stop trigger, fallback process, and person responsible for recovery.

Ask together

What happens if this tool disappears right now?

Facilitator notes

Choose a fallback the team can actually perform. Record dependencies that would prevent it from working.

Source notes

The NIST AI RMF Core calls for contingency processes, safe failure, monitoring, incident response, recovery, and change management.