An AI ROI Checklist

Questions to ask

Is the scope a specific process with an owner and measured baseline? Are tasks decomposed with review and exception time? Are error effects modelled per scenario? Is the realisation mechanism named and agreed? Are build, run, oversight and maintenance costs all included, with TCO over the horizon? Are NPV and payback computed transparently? Are sensitivity and Monte Carlo ranges shown, with the chance of loss? Is adoption ramped? Are risks priced and mitigated? Is funding staged with go/no-go criteria, and is post-launch measurement planned with a comparison group?

The checklist

Use it before approving or presenting.

[ ] specific process, owner, measured baseline (time, volume, errors)
[ ] task decomposition with review + exception time
[ ] error rates per scenario from pilot data
[ ] realisation mechanism named and agreed with budget owner
[ ] full costs: build, run, oversight, maintenance; TCO over horizon
[ ] transparent NPV / payback model with sourced inputs
[ ] tornado + Monte Carlo; range and chance of loss reported
[ ] adoption S-curve and change-management budget
[ ] risk register with mitigations
[ ] staged funding with go/no-go; comparison group for measurement

Revisit after the pilot

Re-run the whole model with pilot data before asking for the next stage of funding.

त्वरित जाँच: Which item belongs on an AI ROI checklist?

• Ignoring maintenance costs
• Assuming 100% adoption on day one
• A named and agreed mechanism for turning freed time into value
• Presenting only the best case