What AI Should and Shouldn't Do at a GP Office
AI can relieve reception staff from routine tasks — but the boundary between administration and clinical judgement is absolute. Here is what can responsibly be automated.
The line that cannot be moved
AI tools are gradually becoming more common in administrative settings — including in healthcare. But in a GP practice, one boundary must always hold: the separation between administration and clinical judgement. No technology can substitute where professional expertise is required.
This means AI-based automation has a clear and valuable role — but also non-negotiable limits.
What AI is well suited for
Routine tasks that require no clinical judgement are natural candidates for automation. Appointment booking and cancellations follow clear rules: if slots are available, they can be booked without a member of staff having to step in. The same applies to answering questions about opening hours, the practice’s address, and which services it offers.
An AI assistant handling phone calls can identify what the patient needs, resolve routine enquiries directly in the conversation, and provide the office with a tidy summary. This frees health secretaries’ time for tasks that genuinely require human judgement.
What AI should not do
Anything involving medical assessment belongs with clinical staff. That obviously includes questions about symptoms, diagnoses, and treatment — but it equally includes judging whether an enquiry has clinical relevance. Deciding whether a patient who says “I have stomach pain” needs a prompt appointment is a clinical decision, not an administrative one.
Urgent situations are a special case. If a patient indicates that something is urgent, that they are in significant pain, or that they are worried about something serious, they must be routed immediately to a member of staff or to the emergency GP service. An AI assistant that cannot distinguish urgent calls from routine ones is not safe for use in healthcare.
Trust and transparency
For patients to trust an AI-based system, it must behave predictably. That means the assistant should always make clear what it is, that patients can ask to speak to a person, and that the system never gives the impression of offering medical advice or assessment.
Transparency is not just an ethical requirement — it is the foundation for automation to work in a high-trust environment like a GP practice. Patients who know what they are talking to and what it can help them with are better placed to use the service appropriately.
AI at the GP practice is not an either/or question. Used within clear boundaries, it can provide genuine relief and better accessibility. Used without them, it risks eroding trust — and at worst, causing harm.