Quick answer
To measure the ROI of an AI tool in a small business, pick one repeatable task, record how long it takes and what it costs today, use the AI tool on that task for 30 days, then compare the value of the time saved or extra sales against the full cost of the tool. The formula is simple: (value gained − total cost) ÷ total cost × 100. Count the time spent checking and fixing AI output as a cost, not a saving.
Key takeaways
- Measure one task, not “AI” in general
- Record a baseline before you start, or you will be guessing
- Include checking and fixing time in your costs
- Run the trial for about 30 days with one named owner
- Decide with numbers: keep, change or cancel
📘 Main article: This guide follows our article Per Seat vs Per Token AI Pricing: Why It’s the Wrong Question (and What to Ask Instead), which explains why the real question is not how you pay for AI, but whether you can prove it creates value.
In this guide
- What does ROI mean for an AI tool?
- Which task should you measure first?
- How do you record a baseline?
- How do you run a fair 30-day trial?
- How do you calculate the ROI?
- Worked example: AI for customer emails
- What if the benefit is not time saved?
- Keep, change or cancel?
- Frequently asked questions
- The bottom line
- Related posts
Most small businesses now pay for at least one AI tool. Far fewer can say whether it is earning its keep. As we explain in our main article on per seat vs per token AI pricing, the pricing model matters less than knowing what the tool actually delivers.
This guide gives you a simple, honest way to measure that, with no spreadsheet skills needed.
What does ROI mean for an AI tool?
Return on investment (ROI) compares what you gain from the tool with what it costs you. For most small businesses, the gain shows up as one of three things:
- Time saved on a task you already do
- Extra revenue from work you could not get to before
- Fewer mistakes or costs, such as refunds or missed invoices
The cost is not just the subscription. It includes setup time, learning time and the time spent checking the AI’s work.
Which task should you measure first?
Choose a task that happens often, takes real time and has a clear finish. Good first candidates:
| Task | How often | What “done” looks like |
|---|---|---|
| Replying to common customer emails | Daily | Reply sent, no follow-up needed |
| Writing product or listing descriptions | Weekly | Description published |
| Chasing unpaid invoices | Weekly | Invoice paid |
| Turning a video into a blog post | Weekly | Post published |
| Summarising sales calls into the CRM | Daily | Notes saved and next step set |
💡 Tip: Avoid starting with a vague goal like “marketing” or “productivity”. If you cannot say when one unit of the task is finished, you cannot measure it.
How do you record a baseline?
Before using the AI tool, time yourself (or your team) doing the task five to ten times. Write down:
- Minutes per task, on average
- How many times the task happens each month
- Hourly cost of the person doing it (wage plus overheads, or your own target hourly rate)
- Quality: errors, complaints or rework, if any
Multiply minutes per task by tasks per month and divide by 60 to get hours per month. Multiply by the hourly cost to get your monthly baseline cost.
How do you run a fair 30-day trial?
- Name one owner who decides how the AI is used and checks its work
- Use the same task and the same quality bar as your baseline
- Log every use: time to prompt, time to check, time to fix
- Track the bill, including any usage or overage charges
- Note quality issues, such as wrong facts or off-brand wording
⚠️ Warning: Do not count time “saved” if the output then needed rewriting. Checking and fixing time is part of the AI’s cost. If you skip it, your ROI will look far better than it is.
If the tool charges per use, our guide How to Work Out the Real Cost of an AI Task Before You Switch Models shows how to work out the real cost per finished task.
How do you calculate the ROI?
Use this formula at the end of the month:
- Value gained = (baseline hours − hours with AI) × hourly cost, plus any extra revenue you can clearly link to the tool
- Total cost = subscription or usage fees + setup and learning time (valued at your hourly cost)
- ROI % = (value gained − total cost) ÷ total cost × 100
A positive ROI means the tool paid for itself in that month. Setup time is usually a one-off, so month two is often a fairer test than month one.
Worked example: AI for customer emails
The figures below are an illustrative example, not real results.
| Item | Before AI | With AI |
|---|---|---|
| Emails per month | 200 | 200 |
| Minutes per email | 6 | 2.5 (including checking) |
| Hours per month | 20 | 8.3 |
| Cost at £20 an hour | £400 | £166 |
| Tool cost | — | £40 |
| Setup and learning (one-off) | — | 3 hours = £60 |
Value gained = £400 − £166 = £234. Total cost in month one = £40 + £60 = £100. ROI = (£234 − £100) ÷ £100 × 100 = 134%. In month two, without setup time, it would be higher.
What if the benefit is not time saved?
Measure the result the time was used for. Freed-up time only creates value if it goes somewhere useful. Decide in advance what you will do with it, for example one extra sales call a day or one new product page a week, and track that result.
This is the approach Nate B Jones recommends to leaders in the video behind our main article: put the freed-up time on a revenue project you could not resource before, and count what it earns.
💡 Tip: Write one line in your log each week: “Time saved this week was used for…”. If that line keeps saying “nothing in particular”, the saving is not real value yet.
Keep, change or cancel?
| Result after 30–60 days | What to do |
|---|---|
| Clear positive ROI, quality steady | Keep it, and try the same method on a second task |
| Small or break-even ROI | Change the task, the prompts or the plan (for example, fewer seats) |
| Negative ROI or quality problems | Cancel, or move back to a cheaper plan |
| Unclear, because nobody logged it | Restart the trial with a named owner |
📘 Main article: Wondering whether to pay per user, per use or per result? Read Per Seat vs Per Token AI Pricing: Why It’s the Wrong Question (and What to Ask Instead), and see our companion guide on outcome-based AI agent pricing.
Your next step: pick one task today, time it five times, and write the numbers down. That baseline is the hardest and most valuable part. Then compare pricing models with our real-cost method.
Frequently asked questions
How long should I test an AI tool before judging ROI?
About 30 days is enough for a daily or weekly task. Judge month two as well, because setup and learning time make month one look worse.
What is a good ROI for an AI tool?
Any clearly positive ROI with steady quality is worth keeping. Be wary of very high figures if you have not counted checking and fixing time.
Do I need special software to measure AI ROI?
No. A simple notebook or spreadsheet with task counts, minutes and costs is enough.
How do I value my own time as a business owner?
Use a realistic target hourly rate, such as what you would pay someone else to do the task, or what your best paid work earns per hour.
What if the AI tool improves quality rather than speed?
Track the cost of the problems it reduces, such as refunds, complaints or rework, and count that as value gained.
The bottom line
You do not need a data team to know whether AI is paying off. Measure one task before and after, count every cost honestly, and decide with the numbers. Your next step: time one repeat task five times this week and record the baseline.
Related posts
- Per Seat vs Per Token AI Pricing: Why It’s the Wrong Question (and What to Ask Instead)
- How to Work Out the Real Cost of an AI Task Before You Switch Models
- How to Draft With a Cheap AI Model and Refine With a Premium One
