Salesforce Decode
Salesforcedecode
Back to questions
AgentforceArchitect (Hardest)hipaahealthcarecomplianceguardrails

Architect HIPAA-compliant agent guardrails for patient support

Real World Scenario

A hospital system deploys agents for appointment scheduling and billing questions. PHI must never appear in logs sent to external model providers, and agents must not diagnose conditions.

Expected Answer

• Scope agents to administrative intents only—scheduling, billing, directions—not clinical advice • Apply minimum necessary PHI in prompts; use internal patient tokens instead of full identifiers in LLM context • Configure BAA-covered Einstein features and verify data residency requirements • Implement output filters blocking diagnostic language, medication recommendations, and symptom triage • Restrict agent actions to read-only PHI lookups unless authenticated patient identity is verified • Maintain access logs meeting HIPAA audit trail requirements with role-based transcript access • Conduct annual risk assessment including agent-specific threat scenarios and breach drills

Follow-Up Questions & Answers

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

Test at three levels: unit (isolated logic), integration (end-to-end flow), and UAT (real user scenarios). Scope agents to administrative intents only—scheduling, billing, directions—not clinical advice. Use sandbox data that mirrors production volume. Healthcare agents require narrow intent scope and defense-in-depth. Document the decision in an ADR and align with enterprise standards.

Architect Perspective

Healthcare agents require narrow intent scope and defense-in-depth. If clinical questions are in scope, you need clinician-in-the-loop—not stronger prompts alone.