General-purpose AI wasn't built for the front door of primary care. CareRequest Labs is building a vertical LLM: a conversational agent made specifically for patient intake in EU primary care.
Hi, I need my blood pressure tablets renewed and I'd like a GP review this week.
Thanks. I've noted a repeat prescription and a GP review. Is anything else worrying you today?
Our mission
General-purpose chatbots weren't built for the front door of primary care. Patients describe needs in their own words, in many languages, often while unwell. Our Lab builds agents that understand that conversation, ask the right follow-up questions and turn it into structured, consented data clinicians can act on.
One focus: the first conversation between a patient and their care provider.
Designed around GPs, pharmacies and virtual GP services.
Multilingual, GDPR-aligned and designed for European healthcare systems.
Research areas
A language model tuned to primary care intake: the vocabulary, request types and follow-up logic patients and clinicians actually use.
Agents that guide patients through intake, triage questions and consent in natural conversation, on any channel.
Clear limits built in: no diagnosis, safe escalation of urgent symptoms and a human in the loop for every clinical decision.
Every conversation becomes structured data mapped to FHIR R4, ready to enter the clinical system with a full audit trail.
Research partners
We don't build alone. Our Lab works with research partners who bring deep expertise in applied AI and data modelling.
CeADAR is Ireland's national AI research and development lab. Together we carry out research and joint model work on the CareRequest healthcare agent.
With Data2Sustain we model our data across the CareRequest API and application layer, so every request is structured consistently from the first conversation.
Lab principles
We keep the original conversation alongside the structured record.
Our agents support intake. They never diagnose or replace a GP or pharmacist.
Every step is logged, versioned and explainable.
FHIR R4 and EHDS-ready architecture, built for the systems providers already use.