Before you spend a single pound on AI automation, you need to answer one question: will this actually pay off? It sounds obvious, but most small and mid-sized businesses either skip the maths entirely — jumping in on gut feel — or they get paralysed by uncertainty and do nothing at all. Neither approach serves you well. The good news is that calculating the ROI of an AI automation project doesn't require a finance degree or a spreadsheet the size of a beach towel. You need a simple framework, a few honest numbers, and about an hour of your time. Here's exactly how to do it.
Step 1: Identify What You're Actually Paying For Right Now
ROI calculations start with the cost of doing nothing — what economists call your baseline. Before you can measure a saving, you need to know what you're currently spending.
Start by identifying the specific task or process you want to automate. Then answer three questions:
How many hours does this task take per week? Be honest. Include everyone who touches it — not just the person doing the bulk of the work, but whoever checks it, chases it, or fixes mistakes from it.
What does that time cost you? Take the fully-loaded hourly cost of each person involved. A rough rule of thumb: multiply their annual salary by 1.3 (to account for employer taxes, benefits, and overhead), then divide by 1,800 (typical working hours in a year). A £35,000-a-year employee costs you roughly £25 per productive hour.
What are the downstream costs of errors or delays? This is the number most people forget. If a manual process causes one missed client follow-up a week, what's the average value of a lost deal? If invoicing errors mean you chase payments an extra two weeks late, what's the cash flow impact?
Let's make this concrete. Say you run a 12-person consultancy, and your operations manager spends eight hours every week manually pulling data from your CRM, formatting it into reports, and emailing updates to project leads. At £28 per hour, that's £224 a week — roughly £11,600 a year — in labour alone. And that's before you count the two hours your senior consultant spends each week correcting figures and chasing missing updates.
That's your baseline. Write it down. It's the number you're trying to beat.
Step 2: Estimate the True Cost of the Automation
This is where people often get tripped up — they look at a monthly software subscription and think that's the whole cost. It isn't. A realistic cost estimate for an AI automation project includes:
- Build or setup cost: Whether you hire an agency like BrightBots or use a no-code platform yourself, there's an upfront cost to designing and deploying the workflow. For a moderately complex automation — say, connecting your CRM to your project management tool with AI-assisted logic in the middle — expect anywhere from £1,500 to £6,000 for professional setup.
- Ongoing software costs: Tools like Zapier, Make (formerly Integromat), or custom AI pipelines typically run £50–£400 per month depending on complexity and volume.
- Maintenance and iteration: Automations need occasional tuning. Budget roughly 5–10% of the build cost annually for updates.
- Staff time to transition: There's always a short learning curve. Factor in a few hours of your team's time in the first month.
For a typical mid-market workflow automation, you might be looking at a year-one total cost of £5,000–£9,000, dropping to £1,500–£5,000 in subsequent years once the setup cost is absorbed.
Step 3: Calculate Your Payback Period and Annual Return
Now you have two numbers: your annual cost of the current process, and your annual cost of the automation. The maths from here is straightforward.
Annual saving = Cost of current process − Cost of automation
Payback period (months) = Upfront investment ÷ Monthly saving
ROI (%) = (Annual net saving ÷ Total annual cost of automation) × 100
Back to our consultancy example. The current process costs £11,600 per year in direct labour (we'll leave the downstream costs aside for now to be conservative). The automation — connecting the CRM to a reporting dashboard with an AI layer that formats and distributes updates automatically — costs £3,500 to build and £120 per month (£1,440 per year) to run.
- Year-one total cost of automation: £4,940
- Year-one saving: £11,600 − £4,940 = £6,660
- Year-one ROI: (£6,660 ÷ £4,940) × 100 = 135%
- Payback period: £3,500 ÷ (£11,600 ÷ 12 − £120) = roughly 4.5 months
A 135% return in year one, with payback in under five months. That's a project worth committing to.
In practice, Hogarth Legal Services — a growing 20-person law firm — went through almost exactly this exercise before automating their client onboarding process. They were spending around 14 hours per week across two fee-earners and a paralegal manually collecting client documents, chasing signatures, and updating their case management system. After deploying an AI-assisted onboarding workflow that integrated their email, e-signature platform, and case management tool, that 14 hours dropped to under two. At their billing rates, recovering even half of that time to fee-earning work represented over £40,000 in additional annual capacity — far outstripping the £5,500 first-year cost of the project.
Step 4: Account for the Risks and Soft Benefits
A rigorous ROI calculation doesn't just count the wins — it acknowledges the uncertainties too. Two adjustments worth making:
Apply a confidence discount. If you're not certain the automation will deliver the full saving, reduce your projected saving by 20–30% before running your ROI calculation. If the numbers still stack up, you're on solid ground. If a 30% shortfall kills the business case, you may want to start with a smaller pilot.
Don't forget soft benefits — but don't lead with them either. Reduced staff frustration, faster client response times, fewer dropped handoffs, and scalability without headcount are all real and valuable. They just don't belong in your primary financial calculation. Once you've got a hard-number case that stands on its own, mention these as supporting evidence for the decision, not the main argument.
One final check: ask yourself what the cost of delay is. If your current manual process is costing you £960 per month in wasted labour, every month you wait to automate is £960 you don't get back. For most projects we've seen, a delayed decision costs more than a wrong one.
Conclusion
Calculating the ROI of an AI automation project is less about financial wizardry and more about intellectual honesty. Measure what the current process genuinely costs — including hidden costs like errors, delays, and staff frustration. Get a realistic quote for the automation. Run the numbers with a conservative discount applied. If the payback period is under 12 months and the year-one ROI is above 80%, you almost certainly have a project worth pursuing. If the numbers are marginal, start smaller — automate one slice of the process, prove the saving, then expand. The businesses that benefit most from AI automation aren't the ones with the biggest budgets. They're the ones that did the maths first.