How to Calculate AI ROI for a Small Business

AI ROI for a small business is the annual gain (hours saved plus errors avoided plus any added revenue) compared with the total project cost over the same period. If net gain divided by cost is positive and payback comes in under 12 months, the project is usually worth doing.

The numbers below are a hypothetical example in realistic ranges so you can copy the logic into a spreadsheet. Replace them with your own.

What goes into total cost

Many companies compare only the software subscription and get surprised later. Add everything:

• Setup and integration with your ERP, CRM or spreadsheets (one-time, typically a few thousand to a few tens of thousands of dollars depending on complexity).
• Tool subscriptions and AI API usage, which grows with volume.
• Internal hours spent preparing data, checking answers and training staff.
• Maintenance: tweaks, new cases and monthly monitoring.

Rule of thumb: use the full first-year cost, setup included, as the denominator. In later years cost drops and ROI improves.

Step 1: hours saved

Measure how long the task takes today instead of guessing. Time it for a week or sample it. Then estimate the remaining time with AI, including human review.

Example: sales answers 400 quote requests a month at 12 minutes each. With AI drafting the reply, each takes 4 minutes. Saving: 8 minutes x 400 = 3,200 minutes, about 53 hours a month. At a fully loaded labor cost of 12 dollars an hour, that is roughly 640 dollars a month, or 7,700 a year.

Step 2: errors avoided

Count how many errors happen and what each one costs: rework, reshipped freight, a wrongly produced part, a discount granted by mistake. Resist the urge to inflate this number.

Example: 10 order-entry mistakes a month at an average of 90 dollars each. If automation halves them, the gain is 5 x 90 = 450 dollars a month, or 5,400 a year.

Step 3: added revenue, only if measurable

Answering quotes in hours instead of days may win more deals, but that gain is uncertain. Treat it as a bonus. If you include it, use a conservative guess, such as winning back one or two deals per quarter, and show ROI with and without it.

Step 4: build the spreadsheet

Create four rows: gain from hours, gain from errors, first-year cost and result. With the example numbers:

• Gain from hours: 7,700 dollars a year.
• Gain from errors: 5,400 dollars a year.
• Total gain: 13,100 dollars.
• First-year cost (setup 4,800, tools and API 190 a month or 2,280, and 40 internal hours at 12 or 480): 7,560 dollars.
• Net gain: 5,540 dollars. ROI = 5,540 / 7,560, about 73% in year one.
• Payback: 7,560 / (13,100 / 12), roughly 6.9 months.

Traps that fool your own math

1. Hours saved only turn into money if they are redeployed to productive work or avoid a hire. Otherwise it is capacity, not cash.
2. Account for the AI's own error rate and review time. The example above already includes review.
3. Run a pessimistic case with 30% lower gain and 20% higher cost. If it still comes out positive, the project is robust.
4. Recheck after 90 days with real data and correct the spreadsheet.

For automation and AI projects, start with a process that has clear volume and simple measurement. That gives you a first proven ROI and a base for the next project.

Frequently asked questions

What is a good ROI for an AI project?

For a small company, payback within 12 months is usually acceptable. The more measurable the task (defined volume and time), the more trustworthy the calculation.

How do I calculate an employee's hourly cost?

Add monthly salary, payroll taxes and benefits, then divide by hours actually worked. The result is higher than the plain hourly wage and is the right number to use.

What if the ROI comes out negative?

Review the scope: volume may be too low to justify automation, or the process may need simplifying first. Sometimes the best decision is to wait.