How to measure AI ROI.

1. The basic AI ROI formula

The core formula is the same as any other investment:

Net ROI (%) = (Gain − Cost) / Cost × 100

What makes AI ROI different from, say, a CRM ROI model is not the math — it is what goes on each side of the equation, and how honest you are about both.

2. What belongs on each side of the equation

Gain side

• Hours saved × fully loaded hourly rate. "Fully loaded" means salary plus benefits plus overhead — typically 1.3× to 1.6× base rate. Do not use base rate alone; it undercounts.
• Errors avoided × cost per error. A misinvoiced order, a missed compliance deadline, a duplicate payment. Each has a dollar cost that often dwarfs the hours-saved calculation.
• Revenue retained or captured. Leads followed up on time that would have gone cold. Customers re-engaged. Renewals closed.
• Hire-avoidance savings. If the AI lets you not hire a FTE you would otherwise need, include base salary, benefits, recruiting cost, and ramp time. This is often the largest line.
• Capacity unlock. If the AI means your existing team can handle 2× the volume at the same headcount, that is a revenue multiplier.

Cost side

• Build cost. The consulting fee, the internal stakeholder time, the tools and platform subscriptions for the year.
• Ongoing operating cost. API calls, monitoring, maintenance, iteration. Often 10-20% of build cost annually.
• Change management. Training, documentation, the inevitable "why is this different now?" meetings. Usually 10-15% of build cost.
• Governance overhead. If the project requires logged decisions or compliance review, include that effort. See the governance pillar.

3. A worked SMB example

Real case: a beauty manufacturer with a 3-person accounts payable team. AI invoice automation build:

• Cost: $8,000 build + $2,400/year tools + $1,200 training + $800 change management = $12,400 year one
• Gain: Avoided 1 FTE hire ($65,000 fully loaded) + 3% error reduction on $2.1M AP volume ($63,000) + 5 hours/week saved across existing team at $65/hr ($16,900) = $144,900 year one
• Net ROI: ($144,900 − $12,400) / $12,400 = 1,068%
• Payback period: $12,400 / ($144,900 / 365 days) = 31 days

This is a real outcome, not a spreadsheet exercise. We publish it because the numbers are defensible — every line has a source in the client's own operations.

For a generic version you can run against your own business, use the free ROI calculator.

4. Three traps that break AI ROI models in month six

Trap 1: Counting hours that never got reinvested

If AI saves a salesperson 8 hours a week but they just take longer lunches, the ROI does not exist. Saved time only becomes ROI when it is reinvested into revenue-generating or margin-protecting work. Build the reinvestment plan into the project scope from week one.

Trap 2: Using base salary instead of fully loaded rate

A $60K salary is $90-$96K fully loaded. Using base rate undercounts gains by 50% and makes the project look worse than it is. Always use fully loaded.

Trap 3: Ignoring ongoing costs

Build cost is one-time. API costs, monitoring, and iteration are recurring. A year-one ROI that ignores year-two operating costs is fiction. Model three years, not one.

5. How AI ROI compounds over 12-24 months

The interesting number is not the first-year ROI. It is the second-year and third-year ROI — where the build cost is already paid off and only the operating cost remains on the cost side. A 340% first-year ROI often becomes 800% in year two and 1,200% in year three, assuming the workflow remains stable.

That is why single-project ROI measurement misses the bigger story. The businesses getting the most out of AI are measuring portfolio ROI — total gain across all AI initiatives divided by total investment. That number is what we track with clients on the fractional CAIO retainer.

6. FAQ

What is the typical ROI on AI for a small business?

Across Heed AI Solutions SMB engagements, first-year ROI averages 340 percent. A well-scoped $8,000 pilot typically pays back in about 74 days. Individual results range from 150 percent (conservative efficiency gains) to 1000 percent+ (hire-avoidance scenarios).

How quickly should an AI investment pay back?

A well-scoped first SMB AI project should pay back in 2 to 4 months. If the pitched payback is 12 months or more, the project is too large, too speculative, or mis-scoped. Start smaller.

Should I include soft benefits in an AI ROI model?

Yes, but keep them in a separate bucket from the hard numbers. CFOs trust ROI models that cleanly separate quantified savings from qualitative benefits like customer satisfaction and employee morale.