Clinic Guidance
CLINICAL RISK CLASSIFICATION
CONFIDENCE AND ABSTENTION
STRUCTURED KNOWLEDGE CAPTURE
BRIEF
A private clinic runs on more than treatment. Its owner-director handles the bookings policy, the complaints approach, the supplier contacts, the rota rules and the judgement calls that sit between them. Very little of it is written down. When the director is in clinic, travelling or simply off duty, the team still needs answers and the phone rings at all hours.
Arquin was asked to explore a single idea. How could the director’s working knowledge be captured, structured and made available to the team as a first port of call. The aim was not to replace the director, but to stop them being the only place the answers live.

CHALLENGE
Language models are fluent whether or not they are right. Point one at a clinic’s notes and it will answer anything, including questions it has no business answering, in the same confident tone. In most settings that is an inconvenience. In a clinic it is a governance problem: one plausible but wrong answer, relayed by reception, is enough to undo the trust the tool depends on.
The difficulty is that the line between operational and clinical is thin, and it often runs through the middle of a single question. “Can she fly tomorrow?” is aftercare. “Her lip has gone pale, can she come in today?” is a possible complication and a booking question at once. The system had to classify every question before answering it, identify the parts that belong with a clinician, and recognise when its sources simply did not cover what was asked. Knowing when not to answer is harder than answering.
The knowledge itself was the second problem. Very little of it was written down, and it would not be written down just because a system asked for it. Capture had to happen naturally, at the moment the director answered a real question, and be turned into structured guidance without adding to his workload.

SOLUTION
Arquin designed the system around three tiers of answer, each with its own rules. Operational questions are answered from the director’s written guidance. Aftercare questions are answered only from a protocol signed off by the clinic’s medical lead, repeated as written and never supplemented. Clinical questions are routed to the prescriber on duty, and where a question mixes the two, the clinical part is separated and handled first.
When no guidance exists, the system says so and passes the question to the director, with the staff member’s context and a note of urgency. He sees open questions in one place, answers from a draft prepared from his existing guidance, and his answer is saved as standing guidance. The next person to ask receives it immediately, with its source.
The result is a tool that answers the questions it can, declines the ones it should, and becomes more useful each time the director answers something new.
GOVERNANCE
In a clinical setting, governance has to be part of the design from the start rather than a disclaimer added at the end. Three principles shape how this system handles what it knows.
Clinical boundaries by design. The system does not give clinical advice. Questions about symptoms, suitability, medicines or complications are routed to a named clinician. This is a structural rule, not a setting.
Named sources, not general knowledge. Every answer is drawn from the clinic’s own guidance and shows where it came from. Aftercare information carries the name of the person who approved it and its review date. Nothing is filled in from the model’s general knowledge.
Gaps are escalated, never guessed. Where guidance does not exist, the question goes to the director. Client names and details are kept out of the knowledge base, with a check at the point of entry.
WORKFLOW

SCOPE & ROADMAP
This prototype was deliberately limited. It demonstrates the core loop: ask, answer from approved sources, escalate what is not covered, and capture the director’s answer as standing guidance. The aftercare protocol shown is example content, to be replaced by the clinic’s own signed-off material.
The designed roadmap adds capability in stages. The first step is a production build on UK-hosted infrastructure, with individual sign-in, role-based access and a full audit trail. Then review cycles that prompt the director and medical lead when guidance is due for review. Finally, reporting on the questions asked most often, showing where written policy or staff training would remove the need to ask at all.
Each step extends what the system can do. None relaxes the principles above.
APPROACH
Arquin builds in stages, proving each layer before adding the next, so that value and risk are both visible at every point. This prototype was built quickly to test the experience with realistic questions before any investment in infrastructure. The technology is chosen to fit the problem, not the reverse. And the hardest questions, such as what the system may answer, who stays accountable and where its boundaries sit, are treated as design decisions from day one, not compliance added at the end.