A GP I know described her first week with an AI scribe like this: "It was brilliant in morning surgery, useless in home visits, and I kept forgetting to turn it on for telephone consultations." This is what happens when you adopt a tool without thinking about where it fits in your actual working day.
AI documentation tools are designed for a specific scenario: a clinician and a patient in a room, having a conversation that needs to be recorded. That covers a lot of general practice. But it does not cover all of it.
Successful integration means understanding where AI fits naturally into your workflows, where it needs adaptation, and where it simply does not belong. This lesson maps AI onto the actual working day of a GP practice.
Face-to-face consultations
This is where AI documentation tools work best. The ambient microphone captures the conversation, generates a structured note, and the clinician reviews and files it.
But even in face-to-face consultations, there are workflow decisions to make.
When do you start the tool? Most clinicians activate it as they greet the patient. Some wait until after the initial pleasantries. Find what feels natural and be consistent.
When do you review the note? There are two approaches. Some clinicians review the note immediately after each consultation, while the details are fresh. Others batch their reviews, doing all their notes at the end of a surgery session. The immediate approach catches more errors. The batch approach is faster overall. Most clinicians settle on a hybrid — quick scan immediately, thorough review at the end of the session.
How do you handle the physical examination? AI scribes cannot observe what you see and feel during an examination. You need to either narrate your findings aloud ("Abdomen is soft, non-tender, no organomegaly") or add them manually afterwards. Develop a habit for this — the notes are incomplete without examination findings.
Some clinicians dictate a brief examination summary immediately after examining the patient: "For the record, examination of the chest showed bilateral clear air entry, no wheezes." This feels odd at first but becomes natural quickly, and it ensures the AI captures your findings.
Telephone and video consultations
Telephone consultations are the second-largest proportion of GP workload, and AI tools handle them differently depending on the setup.
If you are using a headset with a desktop microphone, most AI scribes can capture both sides of the conversation. If you are using a handset, the tool typically only captures your voice, which makes the transcript incomplete.
The practical approach: Use a headset or speakerphone for telephone consultations where you want AI documentation. Ensure the tool is configured to capture both audio channels. Test this before relying on it — some setups capture audio unevenly, with the patient’s voice too quiet for accurate transcription.
Video consultations generally work well, as the audio capture is similar to a face-to-face setting. Check that your video platform and AI tool are compatible — some combinations cause audio conflicts.
Always inform the patient that AI documentation is being used, even on the telephone. "I am using an AI documentation tool to help me make notes during our conversation. The recording is processed to create a clinical note and then deleted. Are you happy to proceed?" This takes ten seconds and protects you professionally.
Home visits and out-of-surgery work
Home visits present practical challenges for AI documentation. Mobile signal may be unreliable. Background noise — televisions, pets, family members — can interfere with transcription. The clinical environment is less controlled.
Some clinicians use their mobile phone with the AI app during home visits. Others prefer to make notes manually during the visit and dictate a summary into the AI tool afterwards, using it as a transcription and structuring tool rather than an ambient scribe.
A practical workflow for home visits: Make brief handwritten notes during the visit. Back in the car, dictate a structured summary into the AI tool. Review the generated note before leaving for the next call. File it when you return to the surgery.
This hybrid approach gives you the benefit of AI structuring without the technical challenges of ambient capture in unpredictable environments.
Administrative tasks
AI documentation is not limited to consultations. Several administrative workflows in general practice can benefit from AI assistance.
Referral letters. AI can draft referral letters based on consultation notes and clinical records. Review carefully — check clinical details, dates, and the urgency classification. A two-week wait referral that AI has categorised as routine is a serious safety issue.
Patient information leaflets. AI can help draft plain-English patient information based on clinical guidelines. Useful for condition-specific leaflets that supplement standard resources. Always have a clinician review the final version.
Meeting summaries. Practice meetings, significant event analyses, and clinical governance discussions can be documented with AI assistance. Particularly useful for lengthy meetings where comprehensive minutes are needed.
Protocol drafting. Practice policies and protocols can be drafted by AI and then reviewed and adapted by the appropriate clinical lead. This can save significant time on documents that are important but time-consuming to write from scratch.
For administrative tasks, the risk profile is different from clinical documentation. An error in a practice protocol draft is less immediately dangerous than an error in a consultation note. But always review AI output before using it — the principle of human oversight applies everywhere.
Where AI does not fit
Not every workflow benefits from AI. Be honest about where it adds value and where it adds complexity.
Quick follow-ups. A thirty-second call to tell a patient their blood results are normal does not need AI documentation. Type the note yourself. It is faster.
Group consultations. AI scribes are designed for two-person conversations. A family meeting with four people, or a joint consultation with another professional, produces a transcript that is hard for the AI to structure accurately.
Procedures. Minor surgery, joint injections, coil fittings — these have specific documentation requirements that templates handle better than AI narrative generation.
Highly sensitive content. As covered in Module 3, there are consultations where the presence of any recording technology changes the dynamic. Trust your clinical judgement about when to turn it off.
The goal is not to use AI for everything. The goal is to use AI where it genuinely helps and to have good alternative workflows for everything else.
Building your workflow map
I recommend that every practice creates a simple workflow map: a one-page document that shows which tasks use AI documentation and which do not.
AI-assisted: Face-to-face consultations, telephone consultations (with headset), video consultations, referral letters, meeting summaries.
Manual/template: Quick follow-ups, procedures, group consultations, prescriptions, coding and QOF data entry.
Clinician’s discretion: Home visits, sensitive consultations, complex multi-problem appointments where the clinician prefers to document manually.
Make this visible. Laminate it. Put it in consulting rooms. It helps new team members understand the practice’s approach quickly, and it prevents the inconsistency that comes from everyone making individual decisions about when to use the tool.
Key Takeaway
AI documentation works best in face-to-face and telephone consultations with proper audio setup. Home visits need a hybrid approach. Administrative tasks like referrals and meeting summaries benefit from AI drafting. Not everything needs AI — quick follow-ups, procedures, and group consultations are often better without it. Create a simple workflow map so the whole team knows what goes where.