Agentic AI in GP Surgeries: What to Check Before You Buy

Agentic AI in GP surgeries is often described in terms of autonomy. For a practice buying a service, the more useful question is what the system is authorised to do, what it actually completes and where a person remains responsible.

A product label does not establish that an appointment is booked, a request reaches the right team or the patient understands what happens next. Those outcomes need to be demonstrated.

Define the administrative task

Consider a fictional patient asking to change an appointment. Recording the request, sending it for review and completing a change are separate outcomes. Ask the supplier to identify which outcome is available in the proposed service and how staff can see whether it succeeded.

Repeat the exercise with a correction or a second request. Review what the receiving team sees, including any unfinished work. Keep fictional details in demonstrations.

Ask about limits as carefully as capabilities

• Which actions are available now in this practice’s configuration?
• Which require setup, an additional agreement or further validation?
• What happens when a connected service is unavailable?
• How can the patient reach a member of staff?
• Who owns an incomplete request?

Ask for a written distinction between current capability and planned work. This helps a practice compare the service it can buy today, rather than a broad description of what AI might eventually do.

Evaluate the journey without needing private technical details

A buyer can observe the patient experience, the resulting request and the staff workflow without publishing a supplier’s internal implementation. Concentrate on the outcome, its limits and the evidence needed for the decision.

Use the NHS DTAC framework alongside the organisation’s review. Keep a record of unresolved questions and the people responsible for answering them before deployment.

Apply the same questions to EMMA

EMMA by QuantumLoopAI supports administrative patient access for NHS GP surgeries during surgery hours. It does not provide clinical advice or make clinical triage decisions. The actions available must be confirmed for the practice’s proposed configuration.

Bring the tasks that matter to your surgery, then ask QuantumLoopAI to show the agreed journey from call to next action. The value of the demonstration is that staff can inspect a concrete workflow and identify what still needs checking.

For a PCN, repeat important checks across the different practice configurations. A successful example at one site does not establish that every surgery has the same setup.

Explore EMMA for GP surgeries or book an EMMA demonstration to review the workflow and evidence relevant to your practice.

For a related practical guide, read Intelligent Call Routing for NHS GP Surgeries.