AI ROI: a model that survives the CFO
A real ROI template you can hand to finance — with assumptions you can defend in a board meeting.
Most AI ROI decks die in the CFO's office because they hand-wave the assumptions. This is a real template you can defend in a board meeting — with the numbers that matter and the ones to leave out.
The four-line model
That's it. Four lines. If you can't fill these in with defensible numbers, you don't have an ROI case yet.
| Line | Example |
|---|---|
| Hours saved per week | 22 |
| Loaded hourly cost | $42 |
| Tool cost per month | $800 |
| Net monthly benefit | (22 × $42 × 4.33) − $800 = $3,200 |
That's a 4x return on the tool spend. Annualized, $38,400. Now you have something to defend.
How to defend each line
Hours saved. Don't estimate. Measure. Pick the workflow, time-track it for two weeks before deployment, then again two weeks after. Use the delta. Anything self-reported by the team is suspect.
Loaded hourly cost. Not just wage — include payroll tax, benefits, and the realistic overhead allocation (rent, software seats, management time). For most small businesses this is 1.3–1.5× the hourly wage.
Tool cost. Include the obvious (subscription) and the non-obvious (integration work, onboarding time, prompt tuning). Pad by 20% for the first six months.
Net benefit. Multiply hours × cost × 4.33 (weeks per month) minus tool cost.
What the CFO will push back on
- "Are those hours actually freed, or just shifted?" Answer with a redeployment plan: what does that 22 hours/week now do? (Customer outreach, faster close, a project that was backlogged.) If the honest answer is "nothing," reduce the savings by 50%.
- "What about quality?" Track error rates before and after. If AI introduces new errors, subtract the cost of fixing them.
- "What's the ramp?" Most AI deployments hit full value in month 3. Show a ramp: 30% in month 1, 60% in month 2, 100% in month 3.
Numbers to leave out
- "AI productivity research from McKinsey." Your CFO doesn't care.
- "Industry benchmarks." Your business is the benchmark.
- "Strategic positioning." Save it for the strategy deck.
What to bring to the meeting
- The four-line model with your numbers.
- The two-week time study that backs the "hours saved" line.
- A ramp curve for months 1–6.
- One slide showing the redeployment of freed hours.
If your model survives that conversation, the project gets funded. If it doesn't, you didn't have a real ROI case — you had a hope.