Patient request 58y · description: chest pain on exertion. Pre-triage pattern activated. Notifying the on-call doctor, automatic booking suspended.
AGENT · PATIENT ACQUISITION TRIAGE
Patient leads arrive qualified to the right doctor or operator.
Patient Acquisition Triage receives patient booking requests from forms, chat, messaging channels, and webhooks from third-party booking platforms. It qualifies by requested service (specialty, procedure), recognises preliminary clinical signals for pre-booking triage, and assigns the lead to the right doctor or operator. The clinical decision stays with the professional; the agent sets the stage. Integration with the healthcare management system is delivered during the project.
Patient Acquisition Triage at work.
I'll call the patient back for a preliminary assessment. AI Act disclosure already done at opening.
Call taken over by the doctor. GDPR Art. 9 audit traced.
Why it exists.
Private healthcare facilities — clinics, polyclinics, medical centres, telemedicine services — receive patient requests from multiple channels. Web forms, chat, WhatsApp Business, webhooks. The operational challenge is not the number of channels — it is the internal flow. Front-office staff must identify the right specialty, decide whether the stated urgency is compatible with a wait of days or needs immediate attention, assign to the right doctor or operator.
How it qualifies the request.
The agent activates on every new request. In a few seconds: it extracts structured data (specialty, description of need, preferences), openly declares itself an automated system in compliance with AI Act Art. 50, recognises preliminary clinical signals relevant to pre-booking triage, assigns to the right specialty or operator according to the rules configured by the clinical team. The sequence remains in the audit registry for GDPR Art. 9 audit.
The clinical decision stays with the doctor.
The agent does not diagnose, does not prescribe, does not replace clinical assessment. The perimeter is consistent with MDR-SaMD avoidance. For cases with a recognised preliminary clinical urgency, the agent does not proceed with booking: it immediately notifies triage staff or the GP with the clinical signal prominently shown.
Who it serves and where it applies.
Medical director
Sees the first phase of the patient flow become structured, with declarative rules written by the clinical team rather than case-by-case interpretation by front-office staff. Pre-booking triage quality becomes measurable against the pre-agent baseline.
Marketing chat «I'd like to book a cardiology visit, I had a strange episode…»
Service Cardiology
Urgency To assess · 48h
Contact callback requested
Routed to cardiology desk · CRM updated
Front-office staff
Recover the time spent manually qualifying inbound requests. They handle only the cases that require human judgment — an ambiguous request, a patient with a complex history, cases the agent has escalated as urgent. Standard requests arrive already routed.
outpatient visit · 18:24
«the patient reports exertional dyspnoea for about two months…»
I50.9 Heart failure, unspecified
R06.0 Dyspnoea
EHR populated · physician signature required
Assigned doctor
Receives the patient request with the context already prepared: a specialty consistent with the stated need, any preliminary clinical signal highlighted, the history of prior interactions if the patient is already in the system. Visit preparation does not start from scratch.
Weekly audit 4 systems
HIS · accesses OK
LIS · reports OK
EHR · traceability VERIFY
RIS · privacy OK
1 anomaly · alert to healthcare authority lead
A concrete example.
Saturday evening, a form request.
At a private specialist polyclinic, a patient request arrives via the website form at nine on a Saturday evening. The agent receives it in real time. Data extraction: patient aged 58, requesting a cardiology appointment, description "I've had a few episodes of chest pain on exertion over the last few weeks; I'd like to book a visit." The first automated response openly declares that the interlocutor is a system.
Signal detected: chest pain on exertion.
The agent recognises preliminary clinical signals: chest pain on exertion, marked as requiring clinical triage attention before standard booking under the clinical team's declarative rules. The agent does not proceed with automatic booking. It immediately notifies the on-call triage doctor on the configured work channel.
The doctor calls back within half an hour.
The doctor sees the notification with the clinical signal prominently shown, the full description, the phone number for immediate contact. They decide to call the patient back within half an hour for a preliminary clinical assessment and correct referral. The full event — request, AI Act disclosure, clinical signals, doctor notification, decision — remains in the audit registry for GDPR Art. 9 audit and healthcare regulatory inspection.
Configuration and technical resources.
The pre-booking triage rules are declarative, written by the facility's clinical team. They define: symptom-to-specialty mapping for standard routing, clinical signals that require triage attention before booking, AI Act disclosure consistent with the facility's practice, patient data retention according to the applicable GDPR Art. 9 constraint. The rules live in the customer's repository, versioned, validated at agent startup.
- Language
- TypeScript (Node.js)
- LLM model
- customer's choice: Anthropic, OpenAI, Mistral, open source models hosted internally, AWS Bedrock for a private model
- Built-in controls used
- pii-detector, prompt-injection, topic-guardrail, message-length-limit
- Native delivery channels
- WhatsApp, Telegram, Slack, HTTP OpenAI-compatible (form webhook)
- Corporate email and healthcare portal integration
- dedicated adapter built during delivery by the Exelab team
- Healthcare management system and booking system integration
- dedicated adapter built during delivery (systems vary widely by facility)
- Triage rules
- declarative, written by the clinical team, versioned, validated at startup
- MDR-SaMD perimeter
- the agent does not diagnose, does not prescribe, does not replace the clinical professional
- AI Act Art. 50 disclosure
- mandatory, configured per instance
- Memory
- persistent per instance, pgvector + PostgreSQL FTS on patient context
- Patient data retention
- configurable per instance under GDPR Art. 9
- Registry
- append-only, queryable with a standard SQL client (GDPR Art. 9 and healthcare regulatory audit inspectable)
Frequently asked questions about the agent.
No. Patient Acquisition Triage does not diagnose, does not prescribe, does not replace the clinical professional. The pre-booking triage rules are written by the facility's clinical team and indicate which signals require triage attention before standard booking. The agent follows the rules; the clinical decision stays with the doctor. The perimeter is consistent with MDR-SaMD avoidance.
Yes — and it is mandatory. The first automated response openly declares that the interlocutor is an automated system, provides the front-desk contact for anyone who prefers to speak to a person, then proceeds with qualification. Compliance with AI Act Art. 50 is built into every deployment for systems in contact with end users in the healthcare sector.
Patient data is treated as special category data under GDPR Art. 9. The applicable legal basis is typically Art. 9.2.h (healthcare management by professionals bound by professional secrecy); the specific processing structure must be defined with the facility's DPO in a pre-deployment DPIA. Retention is configurable per instance. The audit registry traces all agent decisions for GDPR and healthcare regulatory inspection.
Healthcare management systems vary widely. Integration with the facility's specific system — clinic management software, regional booking system, national health record — is delivered during the project by the Exelab team. Commercial discovery qualifies the integration patterns for the real customer case.
From a 30-minute conversation to the squad in production.
A 30-45 minute conversation to understand how Patient Acquisition Triage would configure to the facility's case. Which intake channels, which clinical pre-triage rules, which healthcare management system.