Before you spend a penny on AI automation, you should be able to answer one question with a number: what will this actually return? Too many businesses — from busy dental practices to fast-growing consultancies — dive into automation projects on gut feel, then struggle to justify the cost six months later. The good news is that calculating ROI on an AI automation project isn't complicated. You don't need a finance degree or a data science team. You need a clear method, honest inputs, and about an hour. Here's how to do it properly.
Step One: Quantify the Problem You're Actually Solving
ROI calculations fail at the starting line when businesses measure the wrong thing. Don't start with the technology — start with the pain. Specifically, put a number on the time, errors, or missed revenue the current process is costing you right now.
The three most common cost buckets in automation projects are:
Labour time. How many hours per week does your team spend on this task? Be honest and granular. If your office manager spends 45 minutes every morning copying enquiries from a web form into your CRM and then sending a follow-up email manually, that's roughly 3.75 hours per week. At a fully-loaded staff cost of £35 per hour (salary plus employer NI and benefits), that's £131.25 per week — or £6,825 per year — for one task alone.
Errors and rework. Manual processes break. Data gets entered twice. Emails get missed. If your team spends two hours a week chasing mistakes caused by manual hand-offs, add that to your baseline cost. At the same £35/hour rate, that's another £3,640 annually.
Missed revenue. This one is often overlooked. If slow follow-up on inbound leads means you're converting 15% instead of 25%, and your average client is worth £4,000, you're leaving real money on the table. Estimate conservatively — even if just one extra conversion per month is realistic, that's £48,000 in additional annual revenue within reach.
Add these three buckets together and you have your baseline cost of inaction. In this example: £6,825 + £3,640 + £48,000 = £58,465 per year sitting in a broken process.
Step Two: Estimate the True Cost of Automation
Now look at the other side of the ledger. AI automation projects have two types of costs: build costs and running costs.
Build costs cover the initial setup — connecting your tools, configuring the AI agent, testing workflows, and staff training. For a mid-complexity automation (say, connecting a form, a CRM, and an email tool with AI-driven routing and personalisation), a boutique agency typically charges between £2,000 and £8,000 depending on complexity. Simpler single-step automations can come in under £1,500. Enterprise-grade, multi-system builds with custom logic can reach £15,000–£25,000.
Running costs include the monthly subscription fees for automation platforms (tools like Make or Zapier run £30–£150/month at SMB scale) and any AI API usage costs (OpenAI or similar, which for most SMB use cases costs £20–£100/month). Budget a realistic all-in running cost of £100–£300 per month for most business automation projects.
So for a mid-range project: £5,000 build cost + £200/month running cost = £7,400 in year one, dropping to £2,400 in year two.
Step Three: Apply the ROI Formula (It's Simple)
Once you have your baseline cost and your automation cost, the calculation is straightforward:
ROI = (Annual Benefit − Annual Cost) ÷ Annual Cost × 100
Using the example above:
- Annual benefit: £58,465 (recovered labour + rework + missed revenue)
- Year one cost: £7,400
- Year one ROI: (£58,465 − £7,400) ÷ £7,400 × 100 = 690%
Even if you strip out the revenue upside entirely and only count the operational savings (£10,465), the ROI is still (£10,465 − £7,400) ÷ £7,400 × 100 = 41% in year one, improving sharply in year two as the build cost disappears.
Your payback period — the point at which the automation has paid for itself — is equally useful. Divide the build cost by the monthly saving. A £5,000 build saving £875/month in labour and rework pays back in under 6 months.
Real-world example: A London-based HR consultancy with 12 staff was manually processing contractor timesheets — collating spreadsheets, chasing approvals by email, and re-entering data into their invoicing system. The process consumed roughly 8 hours of admin time per week. After an AI automation project costing £4,200 to build (connecting their timesheet tool, Slack, and Xero), the process now runs unattended. Labour saving: £14,560/year. Running cost: £1,800/year. Net year-one return: £8,560. Payback period: 3.5 months.
Step Four: Account for Risk and the Soft Benefits
A robust ROI calculation also stress-tests the downside. Ask yourself: what if the automation only delivers 50% of the projected benefit? Halve your numbers and rerun the formula. If the project still shows a positive return at half-efficiency, that's a strong signal to proceed.
You should also factor in soft benefits — those that don't show up directly in a spreadsheet but absolutely affect your bottom line:
- Staff satisfaction. Removing repetitive tasks reduces attrition risk. Replacing a team member typically costs 50–200% of their annual salary in recruitment and training.
- Scalability. An automated process handles 10x the volume with no extra cost. Manual processes don't scale — they require more headcount.
- Compliance and audit trails. Automated workflows log every action. This matters enormously for law firms, healthcare providers, and financial services businesses where manual processes create compliance gaps.
- Speed to customer. Automated follow-up within 5 minutes versus a human doing it the next morning isn't just an efficiency gain — research consistently shows that response speed is one of the top drivers of conversion.
These don't need precise numbers to be included in your thinking. But acknowledging them prevents you from undervaluing a project that looks borderline on hard figures alone.
Conclusion
The honest reason most businesses skip this calculation is that it feels like extra work before the actual work. But a 45-minute exercise mapping your current costs, your automation costs, and a simple ROI formula can be the difference between a project that gets approved and funded properly — and one that gets stalled, underfunded, or abandoned halfway through. Start with the problem, put a number on it, be conservative on the benefits, and be realistic on the costs. If the numbers work at 50% efficiency, they'll almost certainly work in practice. And if they don't add up even optimistically, that's equally valuable information — it means you should solve a different problem first.