Nobody remembers their first business case fondly. You remember, instead, the precise moment the Strategic Case you drafted on Tuesday stopped agreeing with the Financial Case you finished on Friday. You remember hunting for the origin of a number that three colleagues had quoted and not one could source. You remember the spreadsheet that worked beautifully right up until somebody opened it. And you remember the dawning realisation, somewhere around the fourth draft, that you were spending ninety per cent of your energy on the mechanics of the document and ten per cent on the decision it was supposed to inform.
If any of that rings true, take comfort. It was never you. It was the process.
Between us, we have spent the better part of a working life in and around the business case—designing the methodology, writing the Treasury guidance, training the practitioners, reviewing the results, and occasionally despairing at how much human talent gets poured into the wrong end of the task. Which is why we want to make an argument that may sound surprising coming from the two people arguably most invested in the rigour of the method: a great deal of what you currently do by hand should be handed to a machine. And, at long last, the machine is good enough to take it.
In defence of the method
First, let us be clear about what should not change.
The Five Case Model is one of Britain’s quietest success stories. Born in HM Treasury, refined over decades and across dozens of departments, adopted by governments around the world and accredited internationally through APMG, it has earned its place not through bureaucratic inertia but through intellectual force. It compels a proposal to answer five awkward questions, in order. Is there a compelling case for change—the Strategic Case? Does the preferred way forward deliver value for money—the Economic Case? Can a viable deal be struck—the Commercial Case? Can it be afforded—the Financial Case? And can it actually be delivered—the Management Case?
Answer those honestly, in that sequence, and you have done something rather profound: you have made it genuinely difficult to spend public money badly. That is the entire point. The model is not paperwork. It is a thinking discipline that happens to wear the clothes of a document.
So nothing that follows is a criticism of the method—quite the opposite. The tragedy of the modern business case is that the discipline, the part that creates value, is routinely crowded out by the drudgery, the part that creates only stress and overtime. The model asks for judgement. The process, maddeningly, demands typing.
A £1.5 billion typing problem
And the typing is not cheap. By reasonable estimates, the UK public sector spends in the region of £1.5 billion a year on external consultancy to help produce business cases, and generates well over 27,000 of them annually. Behind those figures sit all the things that never make it onto a slide: the backlogs, the slipped timelines, the capable civil servants quietly seconded to formatting duty, the worthwhile projects that stall not because the idea was weak but because the document took half a year to assemble.
Said out loud, it is faintly absurd. We possess a world-class method for deciding how to invest public money, and we then routinely delay the investment—sometimes past the point of usefulness—because writing it up takes longer than the decision deserves. The option appraisal that could have been thrashed out in a single workshop becomes a quarter of version control. The evidence base that ought to be a living argument decays into a graveyard of broken hyperlinks and half-remembered assumptions.
None of this is the fault of the people doing the work. They are, overwhelmingly, skilled professionals labouring under impossible deadlines with tools designed for a previous century. The genuine miracle is that the cases get finished at all.
The opportunity, and the trap
So here is the opportunity. And here, just as importantly, is the trap.
The opportunity is almost embarrassingly obvious once you name it. Most of what makes business case drafting slow is structured, repetitive and rules-based—which is precisely the work that modern artificial intelligence does well. Assembling a coherent first draft across all five cases. Keeping the narrative and the numbers in agreement with one another. Linking every claim back to its source. Checking that the short-list actually descends from the long-list, and that the affordability statement squares with the economic appraisal. A capable large language model, properly harnessed, can do all of this in minutes—consistently, tirelessly, at three in the afternoon or three in the morning, without once losing the will to live.
The trap is subtler, and it is the reason we are writing this together rather than leaving the technologists to it. A business case is not, fundamentally, a document. It is an argument for spending other people’s money, and it must be owned—truly owned—by a named human being who can stand in front of a board, a committee, or a public inquiry and defend every line of it. You cannot automate accountability. You should never want to.
This is the line that separates a serious instrument from a party trick. There are already tools that promise to turn a one-sentence prompt into a finished business case at the press of a button, and there will be more. Run a mile. A business case that no human has thought hard about is not an asset; it is a liability with a contents page. The aim is not to remove the person from the process. It is to remove them from the photocopying, so they can get back to the thinking.
A real example—declared interest and all
Which brings us, with appropriate transparency, to Business Case Guru. In the interests of full disclosure: one of us architected the Five Case Model and now sits on BCG’s advisory board, precisely because the tool gets this balance right. We would rather tell you that than have you find it out—and you are welcome to weigh the argument accordingly.
BCG is, in the plainest terms, software that walks a user through the Five Case Model and shoulders the heavy lifting of drafting, while leaving every act of judgement exactly where it belongs: with the human. You begin by giving it your project—the context, the objectives, the supporting documents. It then works through the model with you, section by section, proposing structured, evidence-based content for each of the five cases. Every claim it makes is linked back to its source, so you are never asked to take a number on faith. An “Ask BCG” function lets you interrogate any part of the draft—why this option, where that figure came from, which assumption is doing the quiet work underneath—and you approve each section before moving on. When you are finished, it exports cleanly to Word and Excel, ready for the review process you already trust.
The effect, on BCG’s own evidence, is the sort that makes a finance director sit up straight: a 70 to 80 per cent reduction in the time and cost of producing a business case. Documents that once took months now arrive in days. As the team behind it like to put it, they do not merely write business cases—they accelerate decisions. And in an age of shrinking budgets, accelerating good decisions is not a luxury. It may be the single best investment a public body can make.
The guardrails are the point
If that sounds too good to be true, this is the part that should reassure the sceptics—and we would gently encourage you to remain one.
Because automation in the public sector lives or dies on trust, BCG is built so that the human is not merely “in the loop” but holds the pen from the first word to the last. AI outputs are advisory only. Nothing is published, submitted or actioned without explicit human review; the user accepts, edits or rejects every section, of every case, every single time. The system is deliberately engineered to abstain rather than to bluff—where the evidence is thin or the inputs ambiguous, it stops and asks, instead of manufacturing a confident-sounding answer. It does not take decisions, it does not act autonomously, and it does not replace professional judgement. It never pretends otherwise.
The plumbing matters too—as anyone who has endured an information-governance review will appreciate. BCG runs on Anthropic’s Claude models hosted on AWS in the United Kingdom, with UK data residency, UK and EU GDPR compliance, end-to-end encryption and ISO 27001 certification. Your data is never used to train anyone’s model. What you put in remains yours; what comes out can be checked, line by line, against the public sources it cites.
That pairing—serious capability bolted to serious guardrails—is the whole game. It is the difference between a tool that flatters you with a finished-looking document and one that quietly makes you a sharper, faster, more defensible practitioner. AI suggests; the human decides. On a poster, that reads like a slogan. Built into the architecture, it becomes a promise.
Good news for the certified practitioner
We suspect some readers—the APMG-certified among you in particular—are wondering whether all this is good news or a redundancy notice politely phrased. Allow us to put the worry to bed.
The certified practitioner does not become less valuable in this new world. They become considerably more so. When the mechanical 80 per cent of the work collapses into an afternoon, what is left is the 20 per cent that was always the actual job: the quality of the strategic argument, the honesty of the option appraisal, the robustness of the assumptions, the awkward conversations with stakeholders, the judgement calls that no model and no machine will ever make for you. Those are exactly the capabilities the Five Case Model was designed to cultivate, and that APMG certification exists to validate.
Automation does not dilute expertise. It concentrates it. It takes the practitioner who used to spend three weeks assembling a document and hands back two and a half of those weeks for thinking. The result is not fewer business case professionals. It is better ones—producing better cases, faster, and spending their hard-won skill on judgement rather than on justified margins and broken cross-references.
A deliberately modest claim
We have both been in this field long enough to be allergic to hype, and “AI will revolutionise everything” is currently the most over-promised sentence in the English language. So we will close with a deliberately modest claim, which we nonetheless believe without reservation.
The Five Case Model was always about reaching good decisions quickly and defensibly. For thirty years the method has been ready and the tooling has not. That has now changed. The future of the business case is not a machine that thinks on your behalf. It is a disciplined human, trained in a proven method and equipped with an assistant that does the heavy lifting and leaves the judgement well alone.
That future already has a working example, and it goes by the name of Business Case Guru. Smarter, faster, cheaper—and unmistakably, accountably human.
Radical, isn’t it?