TL;DR

Clarence Medical Centre, a GP surgery in Rhyl, North Wales, has reported that its fully integrated AI-powered digital assistant handled more than 6,000 messages and supported 1,100 interactions in the three weeks since going live on 19 March. Practice manager Gwyn Hughes says conversations average 48 minutes, with some extending to nearly 100 conversational turns — suggesting patients are using it for substantive care navigation rather than quick queries.

How it differs from a chatbot

The assistant, branded “Clarence”, is built on the same clinical pathways as NHS 111 online, meaning it triages patients against the same decision logic a 111 call handler would use. That matters: most GP chatbot pilots over the past five years have been thin wrappers on pre-written FAQs, which patients quickly learned to bypass. Clarence is structured to assess need and route to the right service — registration, appointment booking, knowledge-base answers or escalation to a clinician.

Of 371 individual users, 346 engaged in what Hughes called “fully engaged” conversations. The number of phoned-in queries has dropped, freeing surgery lines for urgent and complex cases. Informal patient feedback has been “consistently positive”, though the practice flags that formal feedback submissions are thin — a familiar pattern for quietly useful digital tools.

Why this lands

Clarence Medical Centre believes it is among the first UK surgeries to deploy a fully integrated AI assistant of this type. The deployment is small, but the pattern — a single practice, Welsh primary care, a named practice manager willing to talk numbers — is exactly the evidence base NHS Wales and England have been missing for AI in primary care. Most of the sector’s AI conversation has been about radiology, cardiology and back-office automation; front-door patient interaction has lagged.

Hughes’s framing is notable too: the assistant is “redistributing” human interaction, not replacing it. That is the same language the Rhyl team uses internally, and the same principle EY invoked this week for audit work — a consistent narrative thread across UK AI deployments.

Looking forward

The critical next data point is clinical outcomes: does Clarence-style triage lead to better-matched appointments and fewer missed red flags? Three weeks of usage data is not enough to answer that, but it is enough to justify funding a longer evaluation.