Design a runtime policy engine for action approval, refusal, and escalation.

Instruction: Describe how you would decide at runtime whether the assistant may act, refuse, or hand off.

Context: Assesses whether the candidate can design a practical architecture and explain the main tradeoffs. Describe how you would decide at runtime whether the assistant may act, refuse, or hand off.

Official answer available

Preview the opening of the answer, then unlock the full walkthrough.

I would give the engine access to identity, workflow state, action type, data sensitivity, confidence signals, and approval history so it can decide whether the next step should proceed, be refused, or be escalated. The same action may deserve different treatment in different...

Upgrade to view official answer

Related Questions