Access boundaries
Integrations and user permissions should be limited to the referral tasks a deployment requires, such as reading referral details, updating status, or attaching approved documents.
Security and trust
Referral automation should act inside clear clinical and operational boundaries.
Clayful is designed around scoped access, configurable guardrails, auditability, and human review for sensitive or ambiguous referral work.
Workflow guardrails
Clinic rules define when automation can proceed and when a case must pause for staff review, including urgent symptoms, denied authorizations, missing clinical information, or patient complaints.
Human review
AI agents do not replace clinical judgment. Sensitive, unclear, or high-risk cases are routed to authorized staff with the referral timeline and supporting context.
Communication controls
Patient outreach can be limited by channel, message type, consent status, language, and retry cadence so communication remains predictable and respectful.
Retention practices
Referral records, logs, and message history should be retained according to customer policy, contract terms, and applicable healthcare record obligations.
AI limitations
Clayful should use approved knowledge sources and workflow rules. When the system lacks confidence or lacks a permitted action path, it should escalate instead of guessing.