How the AI ROI calculator works

A tool that will not show its arithmetic is worth less than one that does.

1. 1

   You give the number of people doing a repetitive task, the hours each spends on it weekly, and a loaded hourly cost.

2. 2

   That produces the current annual cost of the work: people × hours × 52 × rate.

3. 3

   You then set the share of that work you believe AI could absorb. This is the number that decides the answer, and it is the one most often set too high.

4. 4

   The saving is the absorbed share of the annual cost, less your estimate of what building and running the automation costs.

What it assumes

• Loaded cost, not salary. Include employer contributions, tooling and overhead — typically 1.25 to 1.4 times base pay.

• Absorbed work does not vanish. It becomes review, exception handling and maintenance. Something in the range of 60–80% absorption is realistic for well-scoped document or ticket work; 95% is not.

• Running cost is not only tokens. Add monitoring, retries, the person who owns it, and the evaluation suite that tells you when quality drifts.

• No ramp-up is modelled. A real deployment takes weeks to reach steady state, so first-year savings are lower than the figure shown.

How to read the result

Treat the output as a sizing signal, not a business case. Its usefulness is comparative: if two candidate processes come out an order of magnitude apart, that gap is real and tells you which to start with. The absolute number depends entirely on your absorption estimate, so try it at a pessimistic value as well — if the case only works at 90% absorption, it does not work.

Questions

What absorption rate should I use?

For structured, high-volume work with clear rules — invoice extraction, ticket triage, first-line support — 60–80% is achievable. For work requiring judgement or negotiation, 20–40% is more honest. If you are unsure, run it at 50% and see whether the case still stands.

Why does the result look lower than vendor claims?

Because it subtracts running cost and does not assume total replacement. Vendor figures usually quote gross time saved on the best-case task, before review, maintenance or the ongoing cost of keeping quality stable.

Does this account for quality changes?

No, and that is a genuine limitation. Automation that is faster but less accurate can cost more than it saves once rework is counted. Measure accuracy on real inputs before trusting any ROI figure, including this one.