Front office · AI agents
AI Front-Desk Agent
A tool-using agent that handles routine patient requests end to end, around the clock and in the patient's language. Its tools read and write real systems (the patient lookups below call this app's live FHIR API), and its guardrails escalate anything clinical to a human right away.
LLM tool useFHIR R4 read/writeidentity verificationguardrailshuman handoff
62%
requests resolved end-to-end
24/7
EN / ES coverage
Reschedule + outside MRI (Spanish)
Channel: SMS
Verified session
Press “Play conversation” or step through it.
Agent trace
Every tool call, its arguments and the live result
// tool calls will appear here
Tools available to the agent
lookup_patientFHIR Patient search + identity verification (name + DOB)get_appointments / find_open_slotsFHIR Appointment + scheduling rulesreschedule_appointmentWrites back to the EMR (HL7 SIU^S14)get_coverage / get_prior_auth_statusCoverage + Prior Auth Copilotsend_secure_linkForms, imaging upload, payments (expiring)safety_screen / handoff_to_humanRed-flag detection → on-call escalation
Guardrails
- • No clinical advice. Symptoms are screened first; red flags skip scheduling and go straight to the on-call clinician.
- • Identity before PHI: name + DOB match, and write actions need a verified session.
- • Minimum necessary: tools return only the fields the task needs.
- • Scoped writes: the agent can book, reschedule and send links, never edit clinical data.
- • Full audit trail of every message and tool call, plus a human takeover button in the staff console.
- • LLM under a BAA with zero data retention, prompt-injection filtering on inbound text, and evals run before each release.