Loading...
Receptionist · General practice
An AI GP receptionist for admin, with clinicians on triage
An AI GP receptionist can take admin calls off your phone lines. Clinical triage stays with your clinicians, and every call that needs a person gets one.
In focus →
Live routingHover to pause
Incoming
Checking…
AI handles0
A person decides0
Where the line sits
What an AI receptionist should do in a GP surgery
Practice phone lines are busiest when the team is stretched thinnest.
Much of that demand is admin: appointment changes, repeat prescription queries, opening hours and test-result call-back requests.
An AI receptionist for GP practices can handle that admin and pass everything else to a person.
It must not assess symptoms, give clinical advice or decide who is seen first.
This guide sets out where that line sits for an AI GP receptionist, and what NHS England, the MHRA and the CQC expect.
Admin the AI can handle
- Opening hours, directions and practice information
- Moving and cancelling appointment types the practice approves
- Routing repeat prescription queries to the right team
- Logging test-result call-back requests
- Taking messages for the admin team
What stays with the practice team
- Any symptom, clinical question or triage decision
- Urgent or worried callers
- Complaints and concerns
- Callers who may be vulnerable or need adjustments
- Anyone who asks to speak to a person
The line
Admin to the AI GP receptionist, triage to clinicians
The safest design keeps the AI firmly on the admin side of the desk.
In NHS England’s modern general practice model, the practice assesses each patient’s needs and allocates them to the right professional.
An AI GP receptionist should support that flow, not replace the assessment step.
It collects admin details, answers approved questions and hands anything clinical to your care navigators or clinicians.
If a caller mentions symptoms, the AI passes the call on and makes no attempt to judge urgency.
The emergency message, such as when to call 999, is written and approved by your clinicians.
Patients who cannot or will not use the AI must always be able to reach a person.
85%
AI acts alone
130
65% of 200 calls
Goes to a person
70
35% of 200 calls
Illustrative batch with a typical confidence spread. Your real split comes from a pilot on your own work. Some decisions go to a person whatever the score.
Regulation
Where software becomes a medical device
The MHRA decides whether software is a medical device based on its intended purpose.
Its guidance on software applications gives examples of symptom checkers that may and may not be devices.
Software that only signposts people to care is unlikely to be a device, while software that matches symptoms to conditions may be.
An admin-only AI GP receptionist stays well clear of that line by design.
The CQC’s GP mythbuster on AI expects clinical risk work under DCB0160, with a named clinical safety officer.
It also expects human oversight, with AI used as a support tool rather than a replacement for it.
Any NHS AI receptionist project needs that clinical safety and data protection work, and we do not make clinical safety decisions for you.
Our AI assurance review can then test the human oversight design, alongside your own clinical safety work.
AI drafts
Check
Person decides
Learns
01 / 04
Telephony and transparency
Fitting your phone system and telling patients
NHS England’s guidance on telephone journeys covers call-backs, queue messages and routing on advanced cloud-based telephony.
GP phone AI should sit inside that journey, not bolt on beside it.
The GP contract requires new telephony to be bought through the Advanced Telephony Better Purchasing Framework.
So check with your supplier and ICB how any AI layer fits your contract before you commit.
The CQC says that because AI is new, practices need to tell people they are using it.
It also expects a non-digital route to care, so no patient is shut out.
Our wider UK buyer’s guide to AI receptionists covers recording, data location and handover checks.
NHS England’s guidance on telephone journeys covers call-backs, queue messages and routing on advanced cloud-based telephony.
How it works
Four steps, one owner
Each system ships with a named person accountable for what it does.
01
01
Map the calls
Sort a sample of calls into admin, clinical and sensitive, with your practice manager and a clinician.
02
02
Agree the boundary
List the admin tasks the AI may do, and the triggers that pass a call straight to a person.
03
03
Safety and data protection
Complete your DCB0160 clinical risk work and DPIA, led by your clinical safety officer and data protection lead.
04
04
Launch small and audit
Start with one call type, review transcripts and escalations weekly, and log any incident.
Questions
Asked and answered
Further reading
From the source
NHS EnglandModern general practice model ↗How practices assess need and allocate patients to the right professional.NHS EnglandHow to improve telephone journeys in general practice ↗Choosing and optimising advanced cloud-based telephony.Care Quality CommissionGP mythbuster 109: Use of artificial intelligence (AI) in GP services ↗What CQC looks for when a practice uses AI.Medicines and Healthcare products Regulatory AgencyMedical devices: software applications (apps) ↗When software, including symptom checkers, counts as a medical device.NHS EnglandGuidance on the use of AI-enabled ambient scribing products in health and care settings ↗NHS England’s guidance on AI scribes and ambient voice technology.
Keep reading
Related pages
AI receptionistAI receptionist UK →AI receptionist UK: an honest buyer’s guideSectorsAI for dentists →AI in dentistry, with the dentist still making the callGovernanceAI assurance →AI assurance, with evidence a person can stand behindTrainingAI governance course →An AI governance course for the people who design the controls
Next step
Tell us the work. We’ll show you where the line goes.
A 30-minute call. We map one workflow, what the AI could take, and where a person must stay in the loop.