If you want to know who would benefit most from AI in your practice, do not start in the consulting rooms. Start at the practice manager’s desk, where the rota, the CQC folder, three draft policies, a recruitment advert, and forty unread emails are all due by Friday.
Practice management is drowning in exactly the kind of work general-purpose AI handles well: structured documents, routine correspondence, and text that follows patterns. Almost none of it involves patient data. This lesson is a tour of the manager’s workload with an AI lens.
Everything here uses the same pattern as Lesson 3: human supplies facts and constraints, AI drafts, human verifies and signs. The examples change; the discipline does not.
Policies and the practice handbook
Every practice maintains dozens of policies — infection control, chaperones, complaints, lone working, business continuity. Most were written years ago, updated under pressure, and formatted differently from each other.
AI is a strong policy partner in three ways. It drafts new policies from a clear brief. It updates old ones — paste your existing policy (policies contain no patient data) and ask what a current version should add, remove, or restructure. And it harmonises: give it your template and ask it to reformat an old policy to match.
AI knows what a policy usually says, not what the current national requirement is. For anything with a regulatory anchor — safeguarding, infection control, employment matters — verify requirements against the current source: CQC guidance, national safeguarding procedures, ACAS, your LMC. AI structures the document; the requirements come from the horse’s mouth.
The same applies to CQC preparation more broadly. AI can turn your evidence notes into a well-organised statement, draft your practice’s responses to the key questions, and build checklists from published inspection frameworks. It cannot tell you what inspectors currently emphasise — that comes from the CQC’s own material and recent local experience.
Meetings and minutes
Minute-taking is a quietly expensive job — someone senior spends the meeting typing instead of thinking, then an hour tidying afterwards.
The safe pattern: rough notes go in, structured minutes come out. Type quick bullet points during the meeting, then ask AI to turn them into minutes with attendees, decisions, and actions with owners. Five minutes of tidying replaces an hour of writing. Practice meetings about operational matters contain no patient data, so this is green zone work — with one exception worth flagging: if a meeting discusses individual patients (a safeguarding meeting, a significant event about a named case), those notes stay out of general AI tools.
What about recording meetings and letting AI transcribe them? Now you are processing your colleagues’ voices and words — personal data of staff, if not patients. It can be done properly, but it needs consent from those present, a clear policy, and a tool your practice has actually assessed. For most practices, bullet-points-in, minutes-out delivers most of the benefit with none of the complexity.
Recruitment and rotas
Job adverts and descriptions. AI drafts a strong advert from your bullet points: the role, the hours, the practice’s character, what makes the job worth having. It also produces interview question sets, scoring templates, and induction checklists. Verify pay scales and contractual details against current national terms — AI will confidently state last year’s figures.
Shortlisting is different. Using AI to score or filter actual applications means feeding it applicants’ personal data and letting it influence decisions about real people — which raises fairness, bias, and legal questions well beyond a drafting task. Keep AI on the employer’s side of recruitment: describing the job, not judging the candidates.
Rotas. The rota is a constraint puzzle — sessions, leave, skill mix, contractual limits. AI chat tools can help you think through rota design and draft the rules; dedicated rota software then does the reliable computation. What AI does well immediately is the correspondence around the rota: the leave policy, the swap-request process, the note explaining the Christmas cover arrangements before anyone has asked.
The daily correspondence
Then there is the inbox itself: the ICB request, the supplier query, the patient’s non-clinical email about parking, the letter to the community trust about a shared process. Drafting these with AI — particularly the delicate ones where tone matters — is a small, constant saving that compounds.
For NHS staff there is a practical route: Copilot Chat is available within NHSmail at no cost, inside the NHS’s own Microsoft environment. For non-clinical drafting it is a sensible default — already where your work email lives, and covered by NHS agreements rather than a consumer sign-up. The Module 2 rule still holds: no patient-identifiable data in your prompts — an approved environment is not a licence to paste in clinical records.
Teach the pattern to the whole team, not just the manager. A receptionist who can ask Copilot to “make this reply firmer but still polite” or “explain our appointment system in simpler words” saves minutes on every awkward email — and the team members doing the most repetitive writing get the most benefit.
Key Takeaway
Practice management work — policies, minutes, recruitment materials, rota correspondence, the daily inbox — is almost entirely patient-data-free and ideally shaped for AI drafting. Verify regulatory and contractual facts against current sources, keep AI away from judging job applicants and out of patient-specific meeting notes, and use Copilot Chat in NHSmail as the practical default for the whole team.