Most AI business cases fail on arithmetic, not on technology. They count the hours saved and ignore what it costs to get them. Here is a method you can run on a napkin before you talk to a vendor.
The basic formula
Annual benefit minus annual cost gives you the net gain. Setup cost divided by the monthly net gain gives you payback in months.
Counting the benefit
- Hours saved: items per week, times minutes saved per item, divided by 60, times 52.
- Value of an hour: use the fully loaded cost of the person doing the work (wages plus taxes and benefits), not their wage alone.
- Realization factor: the share of saved time that turns into real value. Be conservative. If nobody changes how they spend the freed time, the factor is close to zero.
- Recovered revenue: jobs booked, invoices collected sooner, renewals kept. Count only what you can trace.
- Avoided cost: overtime, outside agency or temp labor you no longer buy.
Counting the cost
- Software: the tool itself, usually per user or per usage.
- Setup: configuration, integrations, testing. Include your own time.
- Review time: a person checking outputs. Many owners forget this and it can eat half the savings early on.
- Error handling: the time spent fixing what the AI got wrong.
- Ongoing ownership: someone has to maintain it as your business and the vendor change.
- Training: staff hours to learn the new process.
A worked illustration
The numbers below are made up to show the mechanics. They are not benchmarks, and your figures will differ.
| Line | Illustration |
|---|---|
| Items per week | 120 |
| Minutes saved per item | 6 |
| Hours saved per year | 120 x 6 / 60 x 52 = 624 |
| Loaded cost per hour | $40 |
| Gross value of time | $24,960 |
| Realization factor | 50% |
| Counted benefit | $12,480 |
| Annual tool and review cost | $6,000 |
| Net annual gain | $6,480 |
| One-time setup | $3,000 |
| Payback | about 5.6 months |
Change the realization factor to 100% and the case looks wonderful. Change it to 20% and the project barely breaks even. That sensitivity is the real finding: the decision depends on what happens to the time you save, so decide in advance.
Saved time is only money if it goes somewhere
In Homebase's 2026 survey of small business decision-makers, 87% of AI adopters said it saved them time, and 57% said they used the saved time to grow the business, while 53% used it on tasks they had been neglecting.[2] Those are good uses, but they are not cash savings unless you track them. Write down where the hours will go: more estimates sent, faster collections, less overtime, a hire you do not need to make.
Run three cases, not one
- Conservative: half the time savings you expect, higher review time.
- Expected: your best honest estimate.
- Stretch: what happens if everything goes right.
If the conservative case still pays back inside a period you are comfortable with, proceed. If only the stretch case works, the project is a bet, not an investment. Set your payback threshold before you look at the numbers so you do not talk yourself into it afterward.
Measure before you start
Time the current process for two weeks. Count items and minutes, and log errors. Without a baseline you will be arguing about whether things improved. Bluevine found that 52% of AI-using small businesses reported a return and 24% had not seen one yet.[1] The difference is often whether anyone measured.
Do not approve a project because the stretch case looks great. Approve it because the conservative case is acceptable and you know who will check the number after 90 days.
Sources
Figures are quoted from the sources above and were reviewed in October 2026. Where a source is a secondary summary of other research, we say so in the text. Survey results reflect each survey's own sample and definitions. Examples labeled as illustrations use made-up numbers to show the method. Nothing here is a forecast or a guarantee of results.
Want this worked out for your business?
Tell us what is wasting your time.