KNOWLEDGE BASE · GUIDE
How to Build an AI Business Case Your CFO Will Sign
A CFO signs an AI business case when it arrives in their format: hours and costs as inputs, a 12-month model with three scenarios, a payback period, and the worst case stated plainly. Build the case before you ask, and defend it line by line.
Gather three honest numbers first
- Hours the work costs today — the repeated, checkable work: reporting packs, quoting, replies, document assembly. Name the people whose weeks change.
- Loaded cost per hour — salary, not billing rate. The method example on /scan uses a loaded cost figure against twelve hours a week of manual reporting (skillpilotadvisory.ai/scan, fetched 10 Oct 2026).
- What can actually be removed — freelance spend, overtime, external support. If nothing can be cancelled, the cash benefit is zero and the honest model says so. The healthcare case on /examples shows exactly this state: 67.2 hours released monthly, assumed cash benefit zero, "do not claim cost savings" (/examples, fetched 11 Oct 2026).
The 12-month model
| Line | Input | Where it comes from |
|---|---|---|
| Hours released/month | baseline hours x adoption % x time-reduction % | Your time logs, not vendor claims |
| Cash benefit/month | hours released x loaded cost, ONLY if removable | Payroll/overtime/contractor records |
| Initial cost | tool + setup + training | Quote or product price |
| Running cost/month | subscription + maintenance | Same |
| Net benefit (12mo) | cash benefit x 12 - costs | Calculated |
| Payback month | when cumulative benefit crosses cost | Calculated |
Run it three ways — conservative, expected, upside. The Scan example lands at payback in month 5 under the conservative scenario; the /examples cases range from ~1-month payback to a first-year loss of RM30,300. Both live on the record, because a case that only shows upside does not get signed twice (/examples and /scan, fetched 10-11 Oct 2026).
The CFO summary (one page)
- The decision: fund / fix / stop, stated in the first line.
- The worst case: what you lose if adoption lands at the low end — the telco case on /examples shows first-year net of only RM1,485 and payback at 11.4 months where a one-month delay turns the year negative.
- The owner: one named sponsor answers for the number monthly.
- The check: a monthly sentence, on the record — the line that moved, whether it is outside your normal range, who is accountable.
Sources and context
- Malaysia's National Guidelines on AI Governance and Ethics (MOSTI, Nov 2024) expect accountability and human oversight for AI use — a signed business case with a named sponsor is that expectation, operationalised.
- Most SME AI pilots die between a promising test and a funded rollout; the gap is the approval path (/pillar/ai-pays-off, live 11 Oct 2026).
- The board-test failure pattern: spend rises, pilots multiply, cost/attrition/revenue/cycle time barely move (skillpilotadvisory.ai/about, live 11 Oct 2026).
Scan builds this with you
Scan's business-case step turns your hours and costs into exactly this model — three scenarios, payback, CFO summary, spreadsheet pack you can defend line by line — and the readiness answers pre-fill it. 365 days of access (skillpilotadvisory.ai/scan, fetched 10 Oct 2026). Human first? The advisory founding sprint ends in a fund/fix/stop decision (skillpilotadvisory.ai/advisory, fetched 10 Oct 2026).
FAQ
What does a CFO actually check? The inputs' honesty, the scenario spread, the payback month, and who answers for the result.
Can I use this template free? Yes — the page and /examples give the full method; Scan automates the model if you want it in one login.
What if my numbers show a loss? Publish the loss to yourself and stop or rescope — the property case on /examples did exactly that.
How long to build? "The case takes the time it takes to find three honest numbers" (live /scan FAQ).