The three multipliers vendors omit
Adoption: half your licences typically sit unused. Rework: AI output that must be checked or fixed eats 20–40% of the time saved. Realisation: an hour saved is only worth money if it becomes additional output or reduced headcount — otherwise it's just a slightly easier day. Multiply all three and a claimed 6-hour saving becomes 1.5 real hours.
Adoption is the biggest single lever
At 30% adoption you're paying three dollars for every dollar of licence actually used. Before renewing more seats, reclaim unused ones and invest in enablement — a champion per team, worked examples from your own workflows, and time carved out for practice. Doubling adoption doubles ROI at zero extra licence cost.
- • Audit licence usage monthly and reclaim dormant seats.
- • Train on your real workflows, not generic vendor demos.
- • Track a usage metric per team, not just an org-wide total.
- • Start with a 20-seat pilot and measure before rolling out 200.
Measure output, not sentiment
Self-reported time savings are consistently inflated. Pick a hard metric before rollout — tickets closed, pull requests merged, documents produced, cycle time — and compare a using cohort against a non-using one. If output doesn't move, the saved hours never converted into value, whatever the survey says.
FAQ
How do I calculate ROI on an AI tool?
Multiply licences by adoption rate, weekly hours saved, (1 − rework rate), and a realisation factor, then by the fully-loaded hourly cost and 52 weeks. Subtract licence cost, training, and ramp loss. That net figure over total cost is your true ROI.
Why is my AI tool ROI lower than the vendor claimed?
Vendor models assume 100% adoption, zero rework, and that every saved hour becomes billable output. Real deployments typically see 30–60% adoption, 20–40% rework, and 50–70% realisation — which together cut the claimed benefit by roughly 70%.
What's a good ROI for an AI tool?
3–5× (200–400% ROI) in year 1 is a strong result for a productivity tool. Below 1× you're subsidising the vendor. The key question is whether the shortfall is the tool or your adoption rate — the fix is very different.
How many hours does AI actually save per week?
Credible studies and internal audits usually land at 1–4 hours per active user per week for knowledge work, before accounting for verification time. Claims of 10+ hours are almost always self-reported and unverified.
Should I count reduced headcount as the benefit?
Only if you actually plan to reduce or avoid hires. If headcount stays flat, the benefit is extra output or better quality — real, but you must measure it as output, not as payroll savings.
How long before an AI tool pays back?
With good adoption, 3–6 months is typical for a $30/seat tool against knowledge-worker rates. Poor adoption pushes payback past the renewal date, which is why usage audits should happen at month three, not month eleven.
How this calculator is built
Independently maintained
Written by Sam Doshi and the RevenueLab editorial team. We don't sell the data feeds this tool is built on.
Sourced from primary data
Benchmarks come from public AdSense / Stripe / IRS disclosures and reader-submitted data — never third-party "$X per view" claims. Full methodology.
Last editorial review
Reviewed on a rolling quarterly cycle. Dated reviews are published on the methodology record for each calculator.
Editorial standards
See our editorial policy and disclaimer. Results are estimates, not advice.