AI automation business case: a one-page template
A good business case is short, honest about assumptions and clear about how success will be judged. Here is a one-page structure you can copy, with notes on what decision-makers look for in each section.
The template
| Section | What to write | Tip |
|---|---|---|
| 1. Problem | The task, who does it, and why it matters (cost, delays, errors, staff frustration) | One or two sentences. Name the specific task. |
| 2. Proposal | What will be automated, how, and what stays with people | Say what the AI does and what a person still checks. |
| 3. Baseline | Current volume, time per unit, hours per week, error rate | Measured, not guessed — see how to measure. |
| 4. Assumptions | Time saved %, adoption %, ramp-up, loaded hourly cost | Show the source of each number. Pilot data beats benchmarks. |
| 5. Costs | Setup (design, integration, data, training, pilot) and running (licences, usage, maintenance) | Use the hidden costs checklist. |
| 6. Return | Hours saved, value of time, payback, year-1 ROI, 3-year net value | Include conservative and optimistic cases. |
| 7. Use of freed time | What the saved hours will be used for | The answer turns time into money. |
| 8. Risks and controls | Errors, data protection, adoption, vendor dependence — and the control for each | Short table: risk, likelihood, control. |
| 9. Pilot plan | Who, how long, what is measured | Two to four weeks is usually enough. |
| 10. Decision criteria | What result means scale, adjust or stop — and the decision date | Agree this before the pilot starts. |
Show a range, not a point
Decision-makers trust a business case more when it shows sensitivity. Run the calculator three times — conservative, expected and optimistic time savings — and present all three:
| Case | Time saved | Payback | 3-year net value |
|---|---|---|---|
| Conservative | 20% | — | — |
| Expected | 35% | — | — |
| Optimistic | 50% | — | — |
Fill in from the calculator. If the conservative case still pays back within your hurdle, the decision is easy.
What reviewers look for
- Specificity. "Invoice data entry for 1,500 invoices a month" beats "finance efficiency".
- Measured baselines. Two weeks of real data is more persuasive than any vendor benchmark.
- All the costs. A case listing only licences signals inexperience.
- A clear stop rule. Knowing when you would abandon the project makes approval easier, not harder.
- Honesty about evidence. Reference independent research where it helps — see what the research says.
Frequently asked questions
How long should a business case be?
For a small or mid-sized project, one or two pages is enough. Put the detailed baseline data and calculations in an appendix or a linked spreadsheet. Decision-makers want the problem, the cost, the return, the risks and the stop rule — quickly.
Should we include soft benefits?
Yes, but separately. Faster customer response, fewer errors or better staff morale are real, but they are hard to price. List them as additional benefits and keep the core return based on measured time saved, so the case does not depend on them.
What hurdle should the project clear?
Many small businesses use a simple payback rule — for example, payback within 12 months in the conservative case. Larger organisations may require a minimum ROI or net present value. Agree the hurdle with whoever approves the budget before you build the case.
Who should sign it off?
The budget holder, plus the manager of the team doing the work. Involving the team's manager early improves adoption, which is one of the biggest drivers of the actual return.
Example: one-paragraph summary
"Our three-person finance team spends 24 hours a week keying supplier invoices. We propose an AI extraction workflow that creates draft entries for approval, with flagged exceptions reviewed by a person. Based on a two-week baseline and a three-week pilot showing a 55% net time saving, we expect to free about 600 hours a year, worth roughly $27,000 at loaded cost, for a setup cost of $6,500 and running costs of $3,600 a year — payback in about five months. We will redeploy the time to supplier reconciliation, currently two weeks behind. We will review at 90 days and stop if net savings fall below 30%."