How to Build a Business Case for AI Your Leadership Team Will Approve
By Kesley Kernes · 26 October 2026 · Updated 29 October 2026 · 8 min read
A business case for AI is a short document that shows your leadership team what a specific AI change will cost, what it should return, how you will measure it and what could go wrong. The ones that get approved are narrow, start with a measured baseline and treat every saving as an estimate to be tested, not a promise. Here is how to build one for an Australian small or medium business, step by step.
If you are still working out which problem to tackle first, our complete guide to AI readiness assessments covers how to find and rank opportunities across your teams. This article picks up once you have a candidate and need to win the budget.
Why most AI business cases stall
Most AI business cases stall because they ask leaders to fund a technology rather than fix a problem. "We should get Copilot for everyone" is a purchase request. "Our accounts team spends about 40 hours a week rekeying supplier invoices, and we want to test whether AI can halve that" is a business case.
The other common reasons a partner, board or leadership team says "not yet":
- No baseline. Nobody measured how long the work takes today, so any saving is a guess.
- Gross savings only. The case counts the hours saved but forgets licences, set-up time, training and ongoing support.
- Time saved treated as cash saved. Freeing up ten hours a week only becomes money if those hours go somewhere useful.
- No risk section. Privacy, accuracy and who owns the tool are left for later, which is exactly what makes a careful leader nervous.
What to include in a business case for AI
A good AI business case for an SME fits on two or three pages. Leaders want enough to make a confident decision, plus a clear point at which to stop if it isn't working.
| Section | What it answers | Typical length |
|---|---|---|
| The problem | What is slow, costly or risky today? | One paragraph |
| Baseline | How much time and money it costs now | A few lines of numbers |
| Proposed change | What AI will do, and what people will still do | One paragraph |
| Costs | Set-up and running costs over three years | A short table |
| Benefits | Time, cost, quality or revenue gains, with assumptions | A short table |
| Risks and controls | Privacy, accuracy, security, ownership | Five to eight bullets |
| Measures and decision point | How you will know, and when you will decide | A few bullets |
How to build the business case, step by step
1. Pick one process, not a programme
Start with a single, repetitive process that one team owns: quoting, invoice processing, rostering, client onboarding, monthly reporting. Narrow cases are easier to measure, cheaper to test and much easier to approve.
2. Measure the baseline before you promise anything
Baselines win arguments. For two weeks, have the people doing the work log roughly how long the task takes, how often it happens and how often it needs rework. A simple tally in a shared spreadsheet is enough.
Then put a cost on that time. For a rough guide, the ABS reported that average weekly ordinary time earnings for full-time adults were $2,083.70 in May 2026. Over a 38-hour week that is about $55 an hour, or roughly $61 once you add the 12% superannuation guarantee. Your own payroll figures are better, and remember that payroll tax (where it applies), leave and overheads push the true cost higher.
3. Count every cost, not just the licence
Leaders lose trust quickly when costs appear after approval. Include:
- Software licences or usage fees, for the number of people who will actually use the tool
- Set-up and configuration time, whether internal or external
- Training and the time staff spend learning, which is real cost even when nobody sends an invoice
- Ongoing support, monitoring and updates
- A contingency, because first projects rarely go exactly to plan
Show costs over three years, not just year one.
4. Estimate benefits conservatively, and show your working
Use the low end of what you think is realistic. A conservative case that beats its estimate builds far more trust than an exciting one that falls short. Write down every assumption next to the number it produces, so anyone reviewing it can challenge the assumption rather than the whole case.
Benefits usually fall into four groups: time saved, direct costs avoided (such as overtime or outsourcing), quality gains (fewer errors and less rework) and revenue gains (faster quotes or quicker follow-up). Count only the ones you can measure.
5. Name the risks and the controls
Every AI business case needs a short risk section, and it is often what tips a cautious board towards yes. Cover what data the tool will touch, how accuracy will be checked, who owns the process and what happens if the tool gets something wrong.
If personal information is involved, check your obligations. The Privacy Act generally applies to businesses with annual turnover over $3 million, which can include many established SMEs, along with some smaller businesses such as health service providers. The OAIC's guidance on using commercially available AI products is a useful starting point. From 10 December 2026, covered businesses must also explain in their privacy policy when they use computer programs and personal information to make decisions, or do things substantially and directly related to making them, that could reasonably be expected to significantly affect people's rights or interests. This is general information only, so get professional advice for your situation. Our complete guide to AI policy for Australian businesses covers the wider governance side.
6. Set a decision point
Finish with a date and a measure. For example: "After eight weeks, we will compare processing time against the baseline. If time per invoice has not dropped by at least 25%, we stop or change the approach." A clear exit makes approval feel low-risk, because it is.
A worked example: illustrative only
Here is how the numbers might look for a business with 40 staff. Every figure below is an assumption made up for the example, not a quote, benchmark or expected result. Your numbers will differ.
Assumptions:
- Eight people in finance and operations each spend about 5 hours a week on manual data entry and report preparation: 40 hours a week in total
- Staff cost of about $61 an hour (the ABS average above plus super, before other on-costs)
- AI handles 40% of that work once it is bedded in, saving 16 hours a week
- 46 working weeks a year after leave and public holidays
- Set-up cost of $20,000 and running costs of $8,000 a year
The maths:
- Hours saved: 16 × 46 = 736 hours a year
- Value of that time: 736 × $61 = about $44,900 a year
- Net annual benefit: $44,900 minus $8,000 = about $36,900
- Payback: $20,000 ÷ $36,900 = about 0.54 years, or roughly six and a half months
- Three-year net benefit: ($36,900 × 3) minus $20,000 = about $90,700
Notice that payback uses the net benefit after running costs, not the gross saving. It is an easy step to miss, and one of the first things a sharp finance manager will check.
Also be honest about what that $44,900 means. Unless you reduce overtime, avoid a new hire or redirect those hours into revenue-earning work, it is capacity, not cash. Say which of those you expect, and how you will track it.
Test it before you scale it
The strongest business cases replace assumptions with real results as quickly as possible. A small, time-boxed proof of concept on your own systems turns "we think it saves 40%" into a figure measured on your own work, which is a far easier number to take to the board. Our complete guide to AI proof of concept projects explains how to scope one and what "good enough to keep" looks like.
Frequently asked questions
How long should an AI business case be?
For most small and medium businesses, two or three pages is plenty. Lead with a one-paragraph summary of the problem, the cost, the expected benefit and the decision point. Put detailed calculations in an appendix.
What payback period is reasonable for an AI project?
There is no standard figure, and it depends on your business and its appetite for risk. Many leadership teams are more comfortable with small first projects that can show results within a few months. Agree the payback threshold with your finance lead before you build the case, so the target isn't moved afterwards.
Should we count time saved as cost savings?
Only if the time turns into something measurable: less overtime, fewer contractors, a hire you no longer need, or more billable or sales work. Otherwise, report it as capacity released and track where that capacity goes.
Get a first estimate in 60 seconds
Building a credible business case for AI is one of the first things we work through in an Opportunity Audit: measuring the baseline, ranking opportunities and putting honest numbers in front of your leadership team. If you'd like a starting point before you talk to anyone, the ROI Calculator gives you a conservative estimate, in about 60 seconds, of what your repetitive work costs and what automating it could be worth over three years.
Written by Kesley Kernes, Founder and Principal Consultant of NovoPro. Kesley holds an MBA and a Lean Six Sigma Black Belt, previously established an RPA Centre of Excellence, and helps small and medium businesses across Newcastle, the Hunter and Australia use AI and automation to make and save money.