Scenario

You are the Engineering Manager for a robotics software team deciding how to ship a new autonomy capability. The feature could unlock a major customer expansion, but it depends on a not-yet-proven perception change, a refactor in the control stack, and a tighter release window than the team has used before. The product lead wants speed, the safety lead wants more validation, and the field team is already seeing customer pressure to commit to a date. You need to decide which risks are acceptable, which must be reduced, and which should block launch.

Question

How do you decide which technical risks are worth taking?