You cannot buy the future in advance, and you should not try. But there is a difference between a practice that will meet the next five years standing up and one that will meet them flat-footed — and the difference is built now, out of decisions that cost little and pay off in every version of the future.
This lesson is deliberately practical. Lesson 3 described what is coming from the centre; Lesson 4 described what it means for the role. The question left is what a sensible partnership does about it — without gambling on any particular prediction being right.
The answer is what strategists call no-regrets moves: preparations that pay off whether the future arrives fast or slow, whichever tools win, whatever the MHRA Commission recommends. There are five.
Move 1: Get your data house in order
Every committed national programme — the single patient record above all — makes your clinical data more visible and more consequential. Miscoded diagnoses, unfiled results, ambiguous problem lists: today these are local irritations. From 2027, clinicians across the system will read your records; from around 2028, your patients will read them in the NHS App.
The preparation is unglamorous: coding hygiene as a standing agenda item, a rolling tidy-up of problem lists, consistent use of the searches and recall discipline from Module 5. Practices with clean data will also get more from every AI tool they adopt, because every tool reads the record. Bad data in, bad AI out — at system scale.
NHS England is running engagement sessions on the single patient record with GPs and practice managers through 2026. Send someone. It is the cheapest early intelligence you will get on the change that most directly touches your data responsibilities — which, remember, stay with the practice.
Move 2: Build the governance muscle before you need it
Module 4 gave you the framework — Decide, Prepare, Pilot, Embed — and Module 5 scaled it to a one-page policy. The preparation move is to make this muscle routine while the stakes are low.
A practice that has completed one full, well-documented adoption — DPIA, training records, measurement, review dates — can repeat the process for anything the next five years brings. A practice that has never done it once will improvise under pressure when an ICB programme or an irresistible tool arrives. Governance is a capability, not a document; capabilities are built by repetition.
Keep the artefacts current too: DPIAs reviewed annually, your DSPT submission taken seriously rather than ritually, a named AI lead who reads supplier updates. When a procurement asks what your practice’s AI governance looks like, the answer should already exist.
Move 3: Develop the team broadly, not just the enthusiast
The capability that compounds fastest is a team that understands this technology at working level: what it can do, what it cannot, what must never go into it. That is Modules 1, 2, and 5 — and they were written for the whole practice team, not just clinicians.
The preparation move is to make that knowledge organisational rather than personal. If your practice’s AI competence lives in one enthusiastic partner, it leaves when they do. Build it into induction, into protected learning time, into the shared prompt library from Module 5. The RCGP has been clear that GPs are adopting AI faster than guidance is arriving — the practices that are safe in that gap are the ones whose whole team knows the rules.
Move 4: Keep your commitments reversible
In a fast-moving market, the expensive mistake is not choosing the wrong tool — it is choosing any tool in a way you cannot undo. The supplier landscape will consolidate; the registry is self-certified and evolving; the regulatory framework is being redesigned as you read. Some of today’s vendors will not exist in five years.
Contract short. Prefer annual terms to multi-year lock-ins, even at slightly higher cost. The option to change course is worth a premium right now.
Demand your data back. Any contract should specify that your data — and anything derived from it — is exportable in a usable format at exit. A tool you cannot leave is a tool that owns you.
Prefer standards. Tools that integrate through recognised NHS routes and open standards can be replaced; bespoke integrations calcify. Ask the boring question in every demo: “and how would we leave?”
Beware of the equipment-cupboard version of AI strategy: buying things because the future feels urgent. The future does not need your purchase orders yet. Buy tools that solve a problem you have this quarter, on terms you can exit — and let the committed national infrastructure come to you.
Move 5: Watch the few things worth watching
Finally, assign someone — realistically, your AI lead — to watch the short list of developments that genuinely change decisions: the MHRA National Commission’s recommendations (expected from summer 2026), the AVT procurement framework, Copilot’s arrival in primary care, and the single patient record milestones. Lesson 6 turns this into a sustainable routine rather than a doom-scroll.
That is the whole preparation: clean data, practised governance, a broadly capable team, reversible commitments, and a watchful eye. Notice what is absent — no crystal ball, no speculative spending, no bet on any particular future. A practice that does these five things will be ready for whichever future actually turns up.
Key Takeaway
Prepare with no-regrets moves, not predictions: clean clinical data (the single patient record makes it consequential), governance practised until it is muscle memory, AI literacy across the whole team rather than one enthusiast, contracts kept short and exitable in a consolidating market, and a named watcher on the few developments that matter. Cheap now, valuable in every version of the future.