AI for Business Cases
A business case asks someone to spend money and take risk on your say-so, which is why the document is really two things wearing one template: an argument (structure, options, evidence — where AI genuinely helps) and a commitment (the numbers, the benefits you’ll be measured against, the recommendation — which are yours in a way no tool changes). Handled that way, AI takes weeks out of the assembly without ever touching the accountability.
Why this task matters
Good ideas die in bad business cases constantly: the template intimidates, the options section gets written to make the preferred answer look inevitable, the benefits are optimistic prose nobody intends to measure. And they take forever — weeks of assembling context, formatting options and wrestling structure, which is time the actual analysis never gets. The frustration isn’t the thinking; it’s that the document eats the thinking’s schedule.
The traditional workflow
- Gather context: strategy links, prior attempts, current-state pain
- Develop options (often after the answer’s been chosen, honestly)
- Estimate costs and benefits; negotiate them with Finance
- Draft the document; fight the template
- Review, revise, resubmit; the decision meeting asks the one unanswered question
How AI can help
Draft
- The document skeleton in your template, populated from your notes
- Current-state and background sections from source material
- Parallel-structure options so the comparison is honest and readable
Summarise
- Supporting studies and prior papers into evidence paragraphs
- The finished case into the one-page version the committee actually reads
Analyse
- Options stress-testing: ‘argue the case for the option I’m dismissing’ — the cheapest red-team you’ll ever run
- Consistency: do the numbers in the summary match the tables? Do benefits map to problems stated?
- Gap-checks against your template and against the questions decision-makers always ask
What must stay human
Every number is yours: costs, savings, timelines — AI has no idea what your organisation’s day rates or realistic adoption curves are, and it will invent both fluently. The benefits are promises you’ll be measured against; commit to them personally or not at all. Option development is strategy, and the do-nothing option deserves honest human effort. The recommendation is accountability itself. And reading the decision-makers — what this committee funds, fears and has been burnt by — is organisational judgement no document tool has.
Traffic light assessment
Structure, background drafting, consistency and gap checks from your own material. Assembly work over sources you hold; verification is a read-through with your own numbers.
Options drafting, benefits articulation, evidence synthesis. Useful acceleration, but every figure gets replaced with your verified numbers and every benefit becomes a commitment you’ve personally sanity-checked.
The financial model, the recommendation, anything commercially sensitive in an unapproved tool. Invented numbers in funding documents end careers. Models are built and owned by humans; sensitive commercial content follows the confidentiality red lines.
Example prompt
For the honest-options problem — AI as your red team:
Below are my three options for replacing our document management approach, with my notes on each. Tasks: 1) Rewrite the three options in strictly parallel structure — description, what it solves, effort, risks — so a reader can compare them fairly; 2) Then argue the strongest honest case FOR option one (do nothing), which I’m inclined to dismiss; 3) List the questions a sceptical CFO would ask about my preferred option that my notes don’t answer. Constraints: no invented figures anywhere — where a number is needed and missing, write [FIGURE NEEDED]. [paste options notes]
The risks
Invented numbers are the fatal risk: AI will supply plausible costs, savings percentages and industry benchmarks that survive review right up until someone checks — every figure comes from your model or gets flagged as needed. Optimism bias compounds: generated benefits prose sounds committed while committing to nothing measurable; make every benefit specific, owned and dated. Business cases often carry commercially sensitive strategy — approved tools only. And the polished-document trap is real: fluent prose can make a weak case look strong, including to its own author.
A better workflow
The current way
- Weeks lost to assembly and formatting
- Options written to flatter the chosen answer
- Benefits as prose; the measurement conversation deferred forever
The AI-assisted way
- AI builds the skeleton and background from your notes in the first week
- Options rendered in parallel structure; the dismissed option red-teamed honestly
- You spend the reclaimed time on the model, the benefit commitments and the pre-meetings that actually decide outcomes
What improves
- The document stops eating the analysis’s schedule
- Comparisons get fairer, which decision-makers reward with trust
- The sceptic’s-questions list surfaces weaknesses before the committee does
- Your case arrives with your numbers, your benefits, your name — defensibly
Key takeaways
- AI accelerates the argument; the commitment — numbers, benefits, recommendation — stays yours
- Never accept a generated figure: your model or [FIGURE NEEDED], nothing between
- Use AI as a red team on the options you’re tempted to dismiss
- Benefits become specific, owned and dated, or they’re just prose
- The freed weeks belong to the model and the pre-meetings, not more formatting