The schedule / Entry 05 / Exam preparation
SCS Cockpit.
Retired, job doneExam passed / ACMA CGMA, 2026
The study engine I built for my own CIMA Strategic Case Study. The exam rarely fails people for not knowing things. It fails them for generic theory, no recommendation and missing the facts in front of them. I built against those three, used it until exam day, and passed.
The Strategic Case Study is the last CIMA exam: a fictional company you've studied for weeks, a clock, and a run of emails asking what you'd actually do. Knowing the theory isn't the hard part. The guidance names three ways people fail: generic theory that could apply to any company, no real recommendation, and missing the facts the scenario hands you.
Most study material tests what you know. I wanted something that tested how I answered, under time, against those three.
| Failure mode | What the tool does about it |
|---|---|
| Generic theory | Every drill is anchored to the case company, and Application is marked on whether the answer used its numbers |
| No recommendation | Judgement is marked on a clear call, not a balanced menu of options |
| Missing scenario facts | The command verb is flagged before you write, with the shape of answer it needs |
A drill is composed, not pulled from a bank: a case-specific scenario seed, a competency, a template and an anchor from the study material. It arrives as an email with tasks attached, and a timer changes colour as time runs down. Afterwards the answer is scored on the five dimensions the exam marks: role fit, application, judgement, structure and ethics.
Behind it, 54 source PDFs were normalised into 166 study items and 23 frameworks, loaded from three packs and de-duplicated at startup. Spaced repetition decides what comes back tomorrow. One button turns any drill into an examiner prompt I could hand to an outside AI for a second opinion.
From: Finance Director, Northmoor Composites plcOur largest customer wants 90-day payment terms instead of 30, in return for a three-year volume commitment. The board meets tomorrow. Recommend whether we accept, and what we'd need in place to protect ourselves.
Verb: recommendMake a call, give the reason, say what has to be true for it to work.
Debrief: clear recommendation, good structure. Ethics bolted on at the end. Thread it through.
Retired, job done. The real exam: passed.
A made-up company and a sample marking, built for this page. It doesn't read what you type. The real drills used the exam's own case material, which stays with CIMA.
The exam is three hours under one clock, and the feature that would have rehearsed exactly that never got built. Mock exam mode, three linked sections, a three-hour clock and a structured debrief, went to the top of the priority list in April. It was still at the top after the exam. I built short timed drills instead. They worked, but they weren't the whole exam.
The highest-value feature on the list stayed on the list.
| Ref | Project | Summary | Status |
|---|---|---|---|
| 01 | OpsPilot | An AI operations layer in Microsoft Teams. Reads 6,124+ documents, drafts answers I approve. | Production |
| 02 | Trading Bot | Four AI agents that veto my trades. 115+ sessions, zero impulsive trades let through. | Paper trading |
| 03 | Uni-Nation Digital | Website, inventory PWA and Hindi voice stock entry for the family fabric business. | Being repaired |
| 04 | WhatsApp Smart | A Chrome extension that gives WhatsApp scheduling, AI replies and summaries. | Prototype |