ROI of an AI PoC: how to calculate it before spending
The median ROI of a B2B AI project is +160% over 24 months. But 27% fail. Here is the method to formulate the ROI of your AI PoC before launching the project.
ROI of an AI PoC: how to calculate it before spending
By formulating the business value in one sentence, quantifying the cost of doing nothing, and comparing with the project budget. If you can't write the ROI in one sentence before you start, don't launch the project.
What is the average ROI of an AI project in enterprise?
Across 200 B2B AI projects deployed in France between 2022 and 2025: median ROI of +160% over 24 months, average breakeven at 8 months, and 40% of projects exceeding +300% ROI.
But 27% of projects fail (including 11.5% with a negative ROI). The difference between success and failure doesn't come down to technology, it comes down to scoping.
How do you formulate the ROI of an AI project before launching it?
With a three-component formula:
Value created = time saved (hours × hourly cost) + costs avoided (errors, delays, penalties) + additional revenue (new customers, upsell, time-to-market speed).
Project cost = development budget + training + infrastructure + maintenance.
ROI = (Value created − Project cost) / Project cost × 100.
Concrete example: a supply chain diagnostic on a €115M revenue mid-market company identifies €5–10M in value leaks, of which €2–4M is recoverable as quick wins within 6 months. Diagnostic budget: a few tens of thousands of euros. The ROI fits in one sentence.
Why do small budgets have better ROI?
This is the most counter-intuitive paradox in the data. AI projects under €10K have a median ROI of +245%. Those over €100K: +85%.
| Budget | Median ROI | Breakeven |
|---|---|---|
| < €10K | +245% | 6 months |
| €10K–20K | +195% | 7 months |
| €20K–50K | +155% | 8 months |
| €50K–100K | +110% | 9 months |
| > €100K | +85% | 12 months |
Explanation: small budgets force precise scoping (a single use case), fast deployment (no scope creep), and rapid value proof. Large budgets generate political complexity, expectation inflation, and tool-first thinking.
What factors multiply the ROI of an AI project?
Three documented multipliers:
Training. Companies that invest 25%+ of budget in training achieve 2.4× more ROI (+442% vs. +185% without training). Training is the #1 multiplier, and the most underestimated.
Human-in-the-loop. Systems with human validation generate +372% median ROI vs. +268% without. And 4.3× fewer critical incidents. The human overhead (under 2h/day) is more than offset.
Deployment speed. Projects deployed in under 6 weeks have a 5% failure rate. Beyond 24 weeks: 31%. Every additional month of delay increases the failure risk.
How do you track ROI after deployment?
Four metrics to set from day one:
Time saved per user per week, in hours, measured, not estimated. Costs avoided, errors corrected, delays avoided, penalties sidestepped. Additional revenue, if applicable. Adoption rate, the best proxy for perceived value.
Quarterly recalculation, internal publication. Studies show that ROI measurement rigour is itself a success factor: teams that measure outperform those that estimate.
Calculate the ROI of your AI project →
Related article: Succeeding with an AI PoC in enterprise