The useful outcome
Estimate the return on one AI project using your own numbers, set a baseline before you start, and know what result would make it worth keeping.
Good first-project signals
This approach is a good fit when…
- You are being pitched AI tools and want to know if they pay off
- You want a number before you spend money or staff time
- You need to explain the decision to a partner or manager
Measure one task, not “AI”
Asking what AI is worth to your business is too broad to answer. Asking what it is worth to speed up your weekly quotes, or your replies to new inquiries, is something you can measure. Pick one repeated task and work out the numbers for that alone.
If you are not sure which task to pick, the AI workflow opportunity guide helps you find a good candidate.
Step 1: set a baseline
Before trying anything, write down what the task costs now. Use a week or two of real numbers rather than guesses.
- How many times the task happens per week
- Minutes per occurrence, including fixing mistakes
- Who does it and roughly what their time costs per hour
- Any knock-on effects: slow replies, missed follow-ups, rework
Do this: Time the task for real on at least five occasions. Most owners underestimate how long routine tasks take.
Step 2: estimate the gain honestly
Estimate how much faster the task would be with AI, and include the time to review and correct the output. A draft that takes two minutes to generate and eight to fix only saves time if the old way took longer than ten.
Then decide what the saved time is worth. Time is only valuable if it goes somewhere useful, such as more sales calls, faster service, or fewer late evenings. Be clear about where it goes. The value calculator lets you put in your own hours and rates.
Step 3: count all the costs
The subscription is usually the smallest cost. List everything before comparing.
- Tool subscriptions, per seat or per use
- Setup time: configuring, writing instructions, connecting systems
- Training time for staff
- Ongoing review and maintenance time
- Outside help, if any
Step 4: put it together
A simple version: (hours saved per month × value of an hour) − monthly tool and maintenance costs = monthly return. Divide your one-time setup cost by the monthly return to get the payback period in months.
Illustrative example, not a real client: a task done 20 times a week at 15 minutes each takes about 21 hours a month. If AI cuts it to 6 minutes including review, that frees roughly 13 hours a month. Multiply by what an hour of that person’s time is worth, subtract tool costs, and you have a number you can judge.
Step 5: test before you commit
Your estimate is a hypothesis. Run a small, supervised test for two to four weeks, measure the same numbers as your baseline, and compare. Keep the project only if it beats a result you wrote down before you started.
The guide to running a small AI pilot shows how to structure that test.
Use it with your team
AI ROI worksheet
Fill in one sheet per task you are considering.
- Task and who does it
- Times per week
- Minutes per occurrence today
- Estimated minutes with AI, including review
- Value of an hour of that person’s time
- Monthly tool and maintenance cost
- One-time setup cost
- Result that would make it worth keeping
Measure what matters
What to track
- Actual minutes per task during the test compared with the baseline
- Errors or rework compared with the old way
- Whether the saved time went somewhere useful
Common traps
What to avoid
- Estimating ROI for “AI” instead of one task
- Leaving out review time and setup costs
- Trusting vendor case studies over your own baseline
Direct answers
Frequently asked questions
How do you calculate AI ROI for a small business?
Pick one task, measure what it costs today, estimate the time saved including review, value that time, and subtract all costs. Then test it for a few weeks and compare real results with your baseline.
What is a good ROI for an AI project?
Decide that before you start. For a small business, a reasonable bar is a short payback period on setup costs and a clear use for the time saved. If you cannot say where the saved hours will go, the return is hard to count.
Why do AI projects fail to deliver ROI?
Common causes are choosing a task that is too rare or complex, ignoring review time, skipping staff training, and never measuring a baseline, so nobody can tell whether it worked.
Can you measure AI ROI before buying a tool?
You can estimate it with a baseline and honest assumptions, and many tools offer trials that let you test on real work. Our AI readiness checklist helps you check the basics first.
Want a second set of eyes?
Turn the guide into a practical next move.
If you want to know whether AI will pay off before you spend anything, start with a free 15-minute assessment. We’ll look at one task and give you an honest read on the return.