Every postgraduate exam guide quietly assumes a student's calendar. Working doctors do not have one. The binding constraint on MRCP, MRCGP, MRCEM or SCE preparation is not intelligence or resources; it is that revision must happen in 30 to 45 minute fragments around a rota that does not care about your exam date. The right way to choose AI tools, then, is to stop asking which has the best content and start asking which is engineered for fragments.
The five capabilities that matter for time-poor candidates
Mobile-first delivery, because the fragment happens wherever you are: native apps, instant question loading, sessions that survive interruption. Short adaptive sessions, because with 25 questions a day at stake, every question must be chosen well; adaptive sequencing that chases your weaknesses is worth more to a busy registrar than to anyone else. Automatic planning, because you do not have spare executive function to be your own project manager; the plan must read your performance and redistribute itself when nights delete a week. An always-available tutor, because your revision happens at 11pm when no senior is around to ask; a Socratic layer on missed questions replaces the teaching the rota stole. And continuity across devices and exams, because this is rarely your last exam, and analytics that persist compound.
How the tools score
iatroX is built around exactly this profile: adaptive sets sized to the window you have, spaced reviews batched for commutes, a study planner wired to your bank performance, the Socratic Tutor on every miss, and timed mocks for protected days, in native iOS and Android apps. Two structural fits for working doctors specifically: one £99 annual subscription spans the Royal College and specialty banks, including all 13 SCEs, rather than a fresh purchase per sitting; and the platform stays useful between exams, with askiatroX answering the day job's clinical questions from cited UK guidance and My CPD capturing the reflection your appraisal needs anyway.
Geeky Medics serves the clinical-assessment seasons well, with virtual patients and offline guides suited to ward life. UWorld and AMBOSS are excellent for the US-facing, with mature apps and offline modes. General-purpose AI remains the free late-night explainer, fenced away from being your question source.
A caution specific to the time-poor: generation-heavy workflows, converting your own notes into cards and questions, quietly create curation work, reviewing, correcting and organising the output. Students have slack for that; you do not. Curated content that arrives ready is a feature, not a limitation, at this stage of a career.
Three sample weeks
Foundation doctor, MSRA ahead: 20 adaptive questions on four weekdays, reviews on the fifth, one 45-minute timed block at the weekend; Tutor on every miss, planner absorbing the on-call set. GP registrar, AKT term: lunchtime question sets between surgeries, askiatroX resolving the guideline questions the day generates, Sunday mock fortnightly, SCA rehearsal held to one consultation a week until the AKT passes. Medical registrar, SCE season: specialty-filtered adaptive sets daily, guideline reading only where errors cluster, mocks on the two protected days the month actually contains.
The pattern across all three: small, scheduled, adaptive, relentless. The exam does not know you were busy. The right tools make sure your thirty minutes a day never forgets it either.
Frequently asked questions
How many questions a day is actually enough?
Twenty to thirty, chosen adaptively and worked properly, sustains a serious campaign for most postgraduate exams; the number matters less than the constancy and the follow-through on misses. The trap is calibrating to good weeks: set the floor at what survives your worst rota block, and let the planner raise volume in the generous weeks rather than mourning the thin ones.
What do I do with a week of nights?
Downgrade, do not abandon. Nights weeks convert to reviews-only mode, ten minutes of due questions at some survivable hour, which preserves the spaced-repetition thread that a full stop severs; a connected planner redistributes the displaced new material automatically, which is precisely the feature this life stage buys. The candidates who resume smoothly are the ones whose system never fully stopped.
How should I spend actual study leave?
On the things fragments cannot do: full timed mocks under exam conditions, the deep reading your error clusters have earned, and honest recalibration of the plan against your analytics. Spending precious protected days on ordinary question volume, the work your commute already does, is the classic misallocation; the fragments fund the daily engine so that study leave can fund the rehearsals.
Is it worth studying at all in a five-minute gap?
Yes, if and only if the tool makes it frictionless: five minutes clears a review batch, and review batches are where retention actually lives. The deeper value is habit integrity, since a system you touch daily, however briefly, is a system that survives the term. This is the entire argument for fragment-first platform design, and the fairest test of whether your chosen app has it.