AI receptionists for GP surgeries support patient contact by answering calls, capturing requests and helping the practice manage administrative work. Choosing a service starts with your patients, your reception team and the workflow a request follows after the call.
The most useful comparison is a demonstration of the same tasks, using the same questions and evidence requirements for every supplier. A convincing voice alone does not show whether a system will work well in your practice.
This guide is written by QuantumLoopAI, the supplier of EMMA. It gives practice managers, GP partners and PCN teams a practical framework for evaluating AI reception. It is a supplier-authored guide, rather than an independent ranking of vendors.
What should an AI receptionist do for a GP surgery?
Start by mapping the requests your reception team handles: appointment requests, changes to bookings, prescription enquiries, questions about services and requests that need a member of staff. For each request, define what successful handling means.
Taking a request, passing it to another system and completing a booking are different outcomes. Ask suppliers to show where each task ends, who is responsible for the next step and how the practice can see whether that step succeeded.
At QuantumLoopAI, EMMA is an AI patient access system for NHS general practice. EMMA operates during surgery hours. Clinical decisions remain with appropriately qualified staff. The service should be evaluated against the workflows agreed for your practice.
Seven questions to ask every AI reception supplier
1. What happens when a patient needs a person?
Ask for a demonstration of a patient requesting human help, a conversation the system cannot complete and a caller who needs additional communication support. Follow the handover through to its actual destination. Check the alternative when the receiving team is busy or unavailable.
2. What can it do in our existing systems?
Ask the supplier to demonstrate the exact workflow in your EMIS Web or SystmOne configuration and any consultation tools you use. Request a written distinction between functionality that is live for your practice, functionality that needs configuration and functionality that is planned.
For an appointment workflow, follow a request from patient identification to the practice’s record. Establish whether the system captures a request for review or completes an action, and how failures are surfaced. Avoid relying on a general integration logo as evidence of a specific task.
3. How does it support patients with different access needs?
Use realistic test conversations, including interruptions, background noise, regional accents and a patient asking for information to be repeated. Include people who use the telephone because they find online services difficult. Establish the supported languages and the alternative when the conversation cannot continue.
4. How are patient information and system activity governed?
Ask for the current assurance documents relevant to the product version and your intended use. Identify the people responsible for reviewing security, data protection, accessibility and clinical safety. Record the questions still outstanding before deployment.
The NHS Digital Technology Assessment Criteria provide an assessment framework across these areas. Use the NHS DTAC guidance alongside your organisation’s own review. A marketing badge should be followed by the underlying documents and their scope.
5. What happens during an outage?
Request a demonstration of the fallback process when the AI service, telephone connection or connected system is unavailable. Establish how the practice learns about a failure, what the patient hears and how incomplete requests are recovered.
6. What does the full service cost?
Compare subscription, setup, usage, integration, support and exit costs over the same period. Ask how the quote changes if patient numbers, call volumes or the number of practices changes. Separate staff capacity released from cash savings that can actually be realised.
7. What evidence will show whether it works?
Agree a baseline before introducing the service. Useful measures include abandoned calls, completed requests, successful human transfers, repeat contacts about the same issue, staff corrections and patient feedback. Define each measure consistently and report the denominator as well as the percentage.
A practical scorecard for practice managers
| Area | Evidence to request | Question to resolve |
|---|
| Patient access | Observed call and handover tests | Can this patient complete their request or reach a person? |
| Workflow fit | Demonstration in the relevant configuration | Where does the request arrive and who acts on it? |
| Accessibility | Representative patient testing | What alternative is available if this route is difficult? |
| Assurance | Current documents and named reviewers | Does the evidence cover this product version and use? |
| Reliability | Fallback demonstration and support arrangements | What happens to unfinished requests? |
| Cost and outcomes | Complete quote and agreed baseline | What improvement would justify the total cost? |
Record each criterion as demonstrated, partly demonstrated or awaiting evidence. Treat an unresolved essential requirement as an outstanding decision, even if other parts of the demonstration score well.
How to organise a useful demonstration
- Choose several common patient requests and one difficult handover scenario.
- Run the same scenarios with every shortlisted supplier using fictional patient details.
- Include reception staff and the people who receive the resulting requests.
- Write down the actions completed, the work left for staff and any unanswered questions.
- Agree the measures, responsibilities and review points for any subsequent pilot.
For a PCN or ICB, repeat this exercise across different practice configurations. A workflow demonstrated at one surgery does not establish that every site has the same setup or requirements.
Questions about EMMA by QuantumLoopAI
Is EMMA available outside surgery hours?
EMMA operates during surgery hours. Discuss your practice’s existing arrangements for calls outside those hours when reviewing the proposed service.
Does EMMA make clinical decisions?
EMMA supports patient access and administrative workflows. It does not provide clinical advice or make clinical triage decisions.
Can we check how EMMA would fit our current workflow?
Yes. Bring your clinical system, consultation tools and the tasks you want to review to a demonstration. Ask the team to identify what is available for your proposed configuration and what needs further validation.
Where can we find more information?
Read about EMMA for GP surgeries and the assurance information available from QuantumLoopAI. For a review of your practice’s needs, book an EMMA demonstration.
