Knowing how to measure AI ROI is what separates a useful tool from an expensive experiment. Our founder, Rachel Faciana, came to AI through finance: in 2020 she founded Western Reserve Consulting, a fractional CFO and accounting firm for small and midsize businesses. She looks at AI the same way she looks at any other investment: what does it cost, what does it return, and how will we know? This guide walks through the method we use, step by step.
Step 1: Measure your baseline first
You can't measure improvement without a starting point. Before anything is built, write down what the task costs today. The simplest formula is:
Loaded labor cost is more than a wage. It includes payroll taxes, benefits, and other costs of employing someone. Your accountant or payroll provider can help you estimate it.
Get the hours from the people who actually do the work. Ask them to track the task for a couple of weeks rather than guess. Write down error rates and turnaround times too, since you'll want those later.
Step 2: Count every cost, not just the build
The price of building a tool is only part of the picture. Split your costs into two groups.
One-time costs
- Design and build of the tool or automation
- Data cleanup needed before the tool can work
- Training your team to use it
- Staff time spent testing and giving feedback
Ongoing costs
- Software subscriptions and AI platform licenses
- API usage, which often rises as you use the tool more
- Maintenance: monitoring, fixes when a connected app changes, and model updates
- Ongoing improvements as your process changes
Leaving out ongoing costs is the most common way ROI estimates go wrong. A tool that looks great in year one can disappoint if its upkeep was never counted. Our post on what AI consulting costs explains how these costs are usually priced.
Step 3: Calculate payback and ROI
Two simple formulas answer most of the questions an owner will ask.
- Payback period = one-time costs ÷ net monthly benefit. Net monthly benefit is the monthly value of time saved, minus ongoing monthly costs. This tells you how many months it takes to earn back the upfront investment.
- ROI = (total benefit − total cost) ÷ total cost. Run it over a set period, such as one year and two years, and include both one-time and ongoing costs.
Look at more than one time period. One-time costs land up front, while benefits build month after month, so a two-year view usually tells a fuller story than year one alone.
A worked example of AI ROI for a small business
Say three people on an accounting team each spend 4 hours a week entering vendor invoices. That's 12 hours a week, or 600 hours over a 50-week year. At a loaded labor cost of $40 an hour, the baseline is $24,000 a year.
Now say an automation pulls invoices from email into the accounting system for review, and cuts that time by two-thirds. That frees up 400 hours a year, worth $16,000.
On the cost side, say the build is $10,000 one time, and software, API usage, and upkeep run $200 a month, or $2,400 a year.
- Net annual benefit: $16,000 − $2,400 = $13,600, or about $1,133 a month
- Payback period: $10,000 ÷ $1,133 ≈ 9 months
- Year-one ROI: ($16,000 − $12,400) ÷ $12,400 ≈ 29%
- Two-year ROI: ($32,000 − $14,800) ÷ $14,800 ≈ 116%
In this example, no one loses their job. The team gets 400 hours back to follow up on overdue invoices, clean up vendor records, and help with month-end close. That's the point of AI done well: it takes repetitive work off people's plates so they can focus on higher-value work.
It also means the $16,000 is capacity, not cash in the bank. Be honest about that in your numbers. The return comes from what your team does with the time.
How to treat soft benefits honestly
Some of the best benefits of AI are hard to put a dollar value on:
- Fewer errors, such as fewer mistyped invoice amounts or missed follow-ups
- Faster response times for customers, vendors, and staff
- A better employee experience, with less tedious work and more time for work people find meaningful
Track these, but keep them separate from your core ROI math. If a decision only works once you add generous guesses about soft benefits, treat that as a warning sign. Report them as supporting evidence: "errors fell from X to Y" is more credible than an invented dollar figure.
Step 4: Keep measuring AI ROI after launch
ROI isn't a one-time calculation. Check actual results against your estimate after the first month, then each quarter.
- Re-measure hours spent on the task, using the same method as your baseline
- Compare actual software and API costs to your estimate
- Note how often the tool runs and how often someone has to step in
- Ask the team what they're doing with the time they got back
This is built into how we work. Our AI Readiness Assessment ranks your top opportunities by estimated hours saved, cost, and difficulty. Plan onboarding sets baseline metrics for ROI reporting. The Grow plan includes a monthly ROI report, and Canopy includes a live ROI dashboard. Both are based on labor rates you provide.
Want to know where AI could pay off in your business? Start with our free AI readiness scorecard, or book a free 30-minute discovery call and we'll talk through your numbers together.