Most AI business cases die in the room where they're presented. Not because the idea is bad. Because the case is built around the wrong thing.
If you're trying to get sign-off on an AI project, here's what actually works: a clear picture of what the current situation is costing you, a realistic view of what changes, and numbers that hold up to scrutiny. That's it.
This guide walks through how to write an AI business case that finance teams and decision-makers actually approve, with a worked example and a template you can use straight away.
Why most AI business cases get rejected (and what finance teams actually want to see)
Finance teams aren't hostile to AI investment. They're hostile to vague investment. When a business case comes through with phrases like "drive efficiency" and "unlock scale" but no concrete numbers, it gets parked. Often permanently.
The people reviewing these proposals aren't asking "is AI a good idea?" They're asking "what does this cost, what do we get back, and when?" If your business case can't answer those three questions clearly, it won't get approved regardless of how good the underlying idea is.
What finance teams want to see is straightforward: current state cost, proposed solution, projected ROI, and risk mitigation. Four components. Every approved AI business case I've seen, and we've built a fair few for clients, covers all four.
The common mistake: leading with technology instead of outcomes
The most common reason AI proposals fail is that they lead with the tool instead of the problem.
"We want to implement an AI workflow automation system" tells your CFO nothing useful. It sounds expensive and risky.
"Our team spends 22 hours a week manually processing supplier invoices. At our fully-loaded staff cost, that's £34,000 a year doing a task that can be automated" — that gets attention.
Start with the pain. The technology is just the mechanism for fixing it.
The anatomy of a compelling AI business case
A business case that gets approved doesn't need to be long. It needs to be clear. Here's what it should cover.
Quantifying the current cost of doing nothing
This is the section most people skip, and it's the most important one.
Before you can justify the investment, you have to make the status quo look expensive. Because it usually is. People have just stopped noticing.
The calculation is simple:
How many hours per week does the process take across all staff involved?
What's the fully-loaded hourly cost? Salary plus NI plus benefits, typically 1.3x to 1.5x base salary.
Multiply hours by cost by 52 weeks.
Add error rate cost: rework time, customer complaints, lost revenue from mistakes.
Add delay cost: slow turnaround times affecting customer satisfaction or cash flow.
When we worked through this with a removals company client, the number came out at just over £60,000 a year tied up in manual job scheduling, confirmation emails, and driver communication. That's before factoring in the two or three jobs a month that went wrong because something fell through the cracks.
Once you've got that number, the conversation about investment changes completely.
Estimating ROI without overpromising
The temptation here is to present the best-case scenario. Don't. Finance teams are trained to spot inflated projections, and one number that looks unrealistic undermines the whole document.
Run three scenarios: conservative, base, and optimistic. Be explicit about the assumptions behind each.
Conservative: automation handles 50% of the task volume, staff still involved in edge cases, 6-month implementation period before full benefit realises.
Base: 70-75% automation rate, full implementation in 3-4 months, staff time redirected to higher-value work rather than headcount reduction.
Optimistic: 85%+ automation rate, faster implementation, secondary benefits from better data and reporting.
Showing three scenarios signals that you've thought critically about this. It also protects you. If the conservative case still has a positive ROI, you've made an argument that's hard to dismiss.
On payback period: for most AI builds we do, clients are looking at 6-18 months to break even depending on build complexity and process volume. That's a reasonable benchmark to work from.
A worked example: service business automation case study
Here's what a real business case looks like in practice. The numbers are based on a client in the equipment hire sector, a service business with a small team, high manual workload, and a clear bottleneck in their order and enquiry process.
Before state: time, cost, error rate
The team was handling around 120 inbound enquiries a week. Each one required a manual response: checking stock availability, calculating hire costs, sending a quote, following up if no reply came back.
Time per enquiry: roughly 18 minutes across the full cycle. Total weekly time: 36 hours. Fully-loaded staff cost: £22 per hour. Annual cost of the process: just under £41,000.
Error rate: around 8% of quotes had a pricing or availability mistake that required correction. Each correction took 25 minutes on average and occasionally cost a job. Conservative cost of errors: £6,000-8,000 per year in rework time, not counting lost revenue.
Total annual cost of current state: approximately £47,000-49,000.
Proposed build: what changes and how
The proposed solution was a custom AI system that handles inbound enquiries end to end. It reads the incoming request, checks live stock data, calculates the correct hire rate, drafts and sends the quote, and triggers a follow-up sequence if no response comes within 48 hours.
Staff only touch the exceptions: unusual requests, large orders, or customer escalations. Everything standard moves automatically.
Build cost: £18,000. Monthly support and maintenance: £800 per month.
Projected return: conservative, base, and optimistic scenarios
Conservative scenario: 60% of enquiries handled fully automatically, error rate drops to under 2%, 4-month implementation.
Annual staff time saving: £24,600
Annual error reduction saving: £5,000
Total annual benefit: £29,600
Build plus year 1 support: £27,600
Year 1 net: +£2,000. Year 2 net: +£19,900
Base scenario: 75% automation rate, full benefits from month 4.
Annual staff time saving: £30,750
Annual error saving: £6,500
Total annual benefit: £37,250
Year 1 net: +£9,650. Year 2 net: +£27,650
Optimistic scenario: 85% automation, secondary benefit from staff redeployment to sales activity.
Annual direct saving: £42,000+
Year 1 net: +£14,000+
Even the conservative case shows positive ROI in year 1. That's the argument that gets approved.
Common objections and how to address them
A few you'll hear in almost every sign-off meeting, and what actually works in response.
"What if it gets things wrong?" The current process already has an 8% error rate. The question isn't whether the automated system is perfect. It's whether it's better than what you're doing now. It will be, and staff handle the exceptions.
"We tried automation before and it didn't work." Worth asking what was used. Template-based tools and generic integrations hit their ceiling fast with complex operations. A custom build is a different category of solution. It's built around your specific process, not adapted from a generic workflow.
"This feels like a lot of money." Go back to the cost of doing nothing. If the current process costs £47,000 a year and the build costs £18,000, the risk of not acting is actually higher than the investment cost. The question is whether you can afford to keep running it manually.
"What about our team — will this replace people?" In most of the builds we do, the answer is no. Staff are doing less repetitive work, not fewer jobs. The point is to stop your best people spending a third of their week on tasks a system can handle.
Free AI business case template
If you want a structured template to work through for your own situation, the core sections are:
Executive summary — one paragraph covering the problem, proposed solution, and headline ROI figure
Current state analysis — process description, time audit, fully-loaded cost calculation, error and delay costs
Proposed solution — what changes, what doesn't, what staff involvement looks like after
Financial projections — conservative, base, and optimistic scenarios with clear assumptions
Risk assessment — what could go wrong, how it's mitigated, what the fallback looks like
Implementation timeline — phased rollout, key milestones, when benefits start to realise
Recommendation — clear ask, next step, decision needed
If you want us to do the heavy lifting, run a proper audit of your operation, quantify the actual costs, and build the business case numbers, that's exactly what our initial audit is for. It's refundable against the first build, and you walk away with the numbers regardless of what you decide next.
If this sounds like your business, we should talk. Book a free conversation at amplconsulting.ai.
FAQ — People Also Ask
What should an AI business case include?
An AI business case needs four core components: a quantified cost of the current state, a clear description of the proposed solution, projected ROI across conservative and base scenarios, and a risk section addressing what happens if results come in lower than expected. Without all four, finance teams are left filling in the gaps themselves. And they usually fill them in with a no.
How do you calculate ROI for an AI project?
Start with the annual cost of the process you're automating: hours per week multiplied by fully-loaded staff cost multiplied by 52. Add error and rework costs. That's your baseline. Estimate what percentage the automation removes, model it across three scenarios, then subtract build cost and ongoing support. Dividing build cost by annual saving gives you payback period.
How long does it take to see ROI from AI automation?
For most custom builds, payback is 6-18 months depending on process volume and build complexity. High-volume, repetitive processes with clear inputs and outputs tend to pay back faster, sometimes within 6 months. More complex builds involving multiple integrated systems take longer to fully realise, but the conservative ROI should still be positive within year one.
How do you justify AI investment to a sceptical finance team?
Lead with the cost of doing nothing. Scepticism usually softens when the status quo has a price tag attached to it. If your current process costs £40,000 a year in staff time and the automation investment is £15,000, the frame shifts. The question becomes "can we afford not to?" Show three scenarios, not one, and be conservative in your assumptions. Credible numbers beat optimistic ones every time.
What is an AI business case template?
An AI business case template is a structured document covering seven sections: executive summary, current state analysis, proposed solution, financial projections, risk assessment, implementation timeline, and recommendation. The most important section is current state analysis. Without a clear cost of the problem, the rest of the document has nothing to argue against.
Do I need a consultant to build an AI business case?
Not necessarily. If you have time for a proper process audit and you're comfortable with financial modelling, you can work through it yourself using the template structure above. Where an outside view helps is in the numbers. Most teams underestimate the true cost of manual processes because they've normalised the inefficiency. An external audit usually finds the cost is higher than people thought, which makes the business case stronger.
