Monthly AI return = (hours saved × the hourly cost of the people doing the work) + mistakes avoided + faster replies, minus (tool costs + setup spread over the year + the time people spend reviewing). Write your baseline numbers down before you launch, or you’ll never know what changed.
- Measure before you launch. No baseline, no ROI.
- Count review time as a cost. It’s real.
- Track four numbers monthly: hours saved, straight-through rate, problems caught, response time.
- Likes and “feels faster” aren’t numbers.
The simple formula
A worked example
Here’s an illustrative example for a small office handing invoice entry and order-status emails to AI. The figures are made up to show the method. Use your own.
The biggest returns often don’t show up as hours. A lead answered in two minutes instead of two days, or a renewal that didn’t lapse, can be worth more than a month of saved typing. Note those wins separately, even if you don’t put a dollar figure on them.
The four numbers to track every month
Write your baseline before you launch
Before switching anything on, write down how long the task takes today, how many mistakes happen, and how fast replies go out. That’s your baseline. Then check the same numbers after a month. The one-week time audit gives you most of this for free.
A 30-minute monthly check-up
- Look at your four numbers against the baseline.
- Read five items that were flagged for review. Were the rules right?
- Fix one rule or one instruction based on what you saw.
- Decide: keep it as is, give it more freedom, or add the next task.
Where AI returns usually come from
When you present the result to a partner or your team, lead with the hours and the caught mistakes, because they’re concrete. Then mention speed and consistency as the extra value you’ll keep an eye on.
Common mistakes to avoid
Subtract the minutes people spend checking AI’s work, or the number will look better than reality.
Give a new workflow a few weeks to settle before judging it.
Posts generated or emails drafted aren’t returns. Hours, errors and response time are.
Volumes change. Re-check the four numbers every month.
We don’t publish prices, because the honest answer depends on your volume and how much checking the work needs. What we do is help you write down a baseline before anything is built, then measure against it. If the numbers don’t work on paper, we’ll tell you, and you’ll keep the time audit either way.
Where to go from here
The fastest way to learn this is to try it on one real task this week. If you want a guided start, take the free course. If you’d rather talk it through, book a call.
Questions, answered
How do you calculate ROI on AI for a small business?
Add up hours saved times the hourly cost of the people doing the work, plus mistakes avoided, then subtract tools, setup spread over the year, and the time spent reviewing AI’s work.
How long does it take to see a return from AI?
For a simple, high-volume task, you can usually see whether it’s working within the first month, as long as you measured a baseline first.
Should I include review time in the cost?
Yes. Reviewing AI’s work takes time, and counting it keeps your estimate honest.
What if the ROI is negative?
Pick a higher-volume task, reduce checks on proven low-risk work, or stop. A clear “no” is a useful result.