Salesforce Decode
Salesforcedecode
Back to questions
AgentforceIntermediatetrust-layerdegradationux

Implement graceful degradation when Einstein Trust Layer blocks response

Real World Scenario

Trust Layer toxicity filter occasionally blocks legitimate medical terminology; nurses see "cannot help" with no escalation path.

Expected Answer

• Custom blocked-response handler with domain-specific fallback message • Automatic human queue offer when block occurs twice in session • Tune sensitive topic policies with domain exception list reviewed clinically • Log block reason codes for policy refinement not user shame • Alternative phrasing suggestion to user when safe • Monitor false positive rate by industry vocabulary • Separate clinical agent policy profile from consumer retail

Follow-Up Questions & Answers

Click to expand — each follow-up includes a direct, interview-ready answer

Direct answer: Custom blocked-response handler with domain-specific fallback message Also consider: Automatic human queue offer when block occurs twice in session In practice: Tune sensitive topic policies with domain exception list reviewed clinically Balance speed of delivery with maintainability.

Architect Perspective

Trust Layer false positives are operational incidents — tune policies with domain experts continuously.