In short
Return on investment (ROI) is a standard way to compare the money a project returns with the money it costs, usually expressed as a percentage over a set period such as a year. A project that costs £10,000 in its first year and returns £15,000 of benefit in that year has an ROI of 50%. The arithmetic is simple. The judgement lies in what you count as benefit and cost, and how confident you are in each figure.
Step 1: measure the baseline
Before anything changes, record the figure the project is meant to improve. That might be hours per week spent on a task, the number of calls that go unanswered, the time taken to reply to a web enquiry, or the share of enquiries that turn into bookings. Two to four weeks of real data is far more useful than an estimate made in a meeting.
Without a baseline, any result after launch is a guess, and it becomes very hard to decide whether to keep paying for the tool.
Step 2: value the benefits
| Time saved | Hours saved per week, multiplied by the loaded hourly cost of the people doing the work, and by working weeks in the year. |
|---|---|
| Revenue recovered | Extra enquiries answered or converted, multiplied by your conversion rate and average order value. Use your own historic figures. |
| Errors avoided | Re-work, refunds or missed appointments prevented, valued at what each one costs you today. |
| Capacity | Hours freed for higher-value work. Only count this as cash if the time will be used for something that earns or saves money. |
Step 3: count every cost
Include the obvious costs: software subscriptions, usage charges such as call minutes or AI credits, and any build or integration work. Then add the less obvious ones: staff time for set-up and testing, training, someone’s time each week to review outputs and handle exceptions, and changes to other systems the project relies on.
Spread one-off costs over a sensible period, often one to three years, and be clear which period you are using when you quote the ROI.
Step 4: use scenarios and a payback period
Forecasts for new tools tend to be optimistic. A useful habit is to calculate three cases, such as a conservative, a working and an ambitious scenario, and make the decision on the conservative one. If the project only pays off in the ambitious case, it is a risky bet.
Payback period is often easier for a small business to act on than a percentage. Divide the upfront cost by the monthly net benefit to see how many months it takes to get your money back. Then measure the real figures after launch and compare them with the forecast at your first quarterly review.
A worked example with round numbers
Suppose a team spends 10 hours a week re-typing enquiry details from email into a CRM, at a loaded cost of £20 an hour. That is £200 a week, or about £9,200 a year across 46 working weeks. If an automation removes 35% of that work in the working scenario, the benefit is roughly £3,200 a year. If the automation costs £1,500 to set up and £50 a month to run, the first-year cost is £2,100, giving a first-year ROI of around 52% and a payback of roughly seven months. These figures are illustrative only. Use your own.
When this matters
Relevant before you sign off spend on an AI tool or a build, and again at each review. The free calculator at /ai-roi-calculator estimates the hours and annual capacity AI could recover from your admin load under 20%, 35% and 50% scenarios, with no email required.
Questions
What is the formula for AI ROI?
The standard formula is (benefit minus cost) divided by cost, over a stated period. For AI projects, the benefit is usually time saved, revenue recovered and errors avoided.
Should we trust ROI figures from AI vendors?
Treat them as marketing until you can see how they were calculated. Your own baseline and your own conversion figures are the numbers to rely on.
What if the benefit is hard to put a number on?
Measure what you can, such as response times or hours, and record the rest as a qualitative benefit. Do not invent a cash value for it to make the case look stronger.