The weakest CPD is an attendance certificate; the strongest begins with a genuine problem encountered in practice and ends with an identifiable change in knowledge, behaviour or patient management, which happens to be exactly the shape a well-handled clinical question already has. That alignment is this article's whole argument: an ANP who asks real questions, verifies real sources and reflects briefly is generating revalidation-grade evidence continuously, and the only missing piece is usually capture. The timing sharpens the point: the NMC plans to consult between September and December 2026 on a revised Code and revalidation, with stronger expectations around the safe and effective use of AI and digital technologies signalled, which makes documented, reflective, verification-centred AI use not just good CPD but the direction of regulatory travel.
The workflow, run on a real question
Take the fictional but recognisable trigger: "What monitoring is required after increasing this medicine in a frail patient with declining renal function?" The workflow converts it into evidence in seven moves, each a line or two. The learning need, stated: what the case exposed that you did not securely know. Sources consulted: the Ask-iatroX orientation, then the documents themselves, the guideline's surveillance recommendations, the exact SmPC's monitoring and warnings sections, any shared-care protocol, each opened, per the routing this cluster teaches. How the sources differed: the register differences and any genuine conflict, named rather than smoothed, because noticing purpose-differences between sources is itself the competency being evidenced. What was learned: the delta, in your own words, closed-book, the generation step that makes this learning rather than reading. What action was taken: the plan changed, the review booked, the patient counselled, the escalation made, the practice half without which CPD is a diary. How patient safety was protected: the verification behaviours, run and recorded, the 90-second screen and its full-review promotion where the red flags fired. And what will be reviewed later: the follow-up that closes the loop, plus the topic's entry into spaced review so the knowledge survives past the reflection.
The reusable reflection template
Six questions, answerable in five minutes, mapped to what revalidation actually rewards. What prompted the question? The practice anchor. What did I initially believe? The honest baseline that makes the delta visible. What evidence changed or confirmed my view? Sources, dated. How did I verify it? The behaviours, not the vibes. What will I do differently? The behaviour change, specific enough to audit. How will I know the change helped? The review criterion, which converts reflection into a small closed loop. Add the competency mapping, one line noting which RPS competencies the episode evidences, the full mapping at /blog/rps-prescribing-competency-framework-ai-era, and one further line the AI era specifically earns: reflection on the tool's limitations in this episode, what the synthesis missed, where the source correction happened, which is precisely the critical-use evidence the coming expectations point toward.
Making capture effortless
The workflow fails only at capture, so engineer capture down to nothing: the question, sources and reflection recorded at the moment of resolution, not reconstructed at revalidation season, and held where the learning happened, which is the design argument for a platform where the question, the sources opened, the related practice questions and the CPD record live in one loop, iatroX's configuration, stated plainly as the product this workflow was built to run on and equally runnable, with more friction, in any document you will actually maintain. The compounding return is the point: one captured question per working week is fifty pieces of practice-anchored, verification-evidenced, competency-mapped reflection per year, a revalidation portfolio that assembles itself from work you were doing anyway, and a prescriber whose learning record is, for once, a true picture of their learning.
Frequently asked questions
Does AI-assisted learning count as valid CPD?
The learning counts, evidenced as above: what revalidation assesses is the reflection, verification and practice change, and AI's role, honestly recorded including its limits, strengthens rather than taints the record under every signalled direction of travel.
How many of these entries does a portfolio need?
Quality and spread beat volume: a modest number of complete loops across your scope's territory, each with the action and review halves present, outweighs any quantity of certificates, per the participatory-learning logic revalidation already encodes.
Can the same workflow serve appraisal conversations?
Directly: the captured loops are the raw material appraisal is supposed to discuss, and arriving with a year of them changes the conversation from reconstruction to development, which is what both processes were always for.
What if a captured question later proves to have been answered wrongly?
Capture the correction as its own entry: the follow-up that found the error, the source that settled it and the practice adjustment made is among the strongest evidence of reflective practice a portfolio can hold, and the audit trail only exists because the first entry did.