PRACTICAL FINANCE · ISLAM ALI HASSANIN
Finance AI value and controls lab
Separate time saved from cash realised.
Explore the decision
All amounts use one consistent currency of your choice. Example data only. Inputs stay in this page; nothing is sent to a calculation service. Export your scenario before leaving—unsaved edits are lost on reload or language change.
Edit assumptions and data
View full results and calculations
The worked case and decision context
The question
Does the use case justify a controlled pilot after review effort, costs and risk controls?
Why it matters
Hours released are capacity, not automatically cash savings. Include human review and a documented conversion to cash before presenting ROI.
The decision
Proceed only to a limited pilot when the required controls are evidenced and the first-year case is positive; otherwise address controls or redesign the case. This is a screening rule, not an investment approval.
Worked example
1,000 cases/month,80%adoption,12 manual minutes versus 3 assisted+3 review minutes gives 960 hours/year. At 100/hour, capacity value is 96,000. At 25%cash realisation, benefit 24,000 less 12,000 annual running cost and 10,000 setup leaves 2,000 year-one net benefit. Simple payback is 10 months.
Method and assumptions
Annual hours=monthly cases×12×adoption×(manual−assisted−review minutes)÷60. Capacity value=hours×loaded hourly cost. Cash benefit=capacity value×realisation. Year 1 net=cash benefit−12×monthly cost−setup. ROI=year 1 net÷total year 1 cost. Payback=setup÷positive monthly net recurring benefit.
What this model does not establish
Steady monthly adoption, one currency, undiscounted pre-tax screening. Excludes implementation ramp-up, financing, residual value and speculative quality benefits. Negative hours correctly represent extra work. A zero cash-realisation rate creates no cash savings. Controls are self-reported, not verified by the site; a positive result does not authorise deployment.
Turn the result into an action
- Measure a baseline and define the permitted task.
- Test approved or fictional data against known answers.
- Measure review effort, failures and actual cash conversion.
- Approve a limited pilot with rollback and a named owner.
Review checklist
Keep your work
The working paper is an HTML file containing current inputs, results, assumptions, sources and your action notes. Open it offline and print or save as PDF from your browser. JSON restores the numeric scenario; it does not include action notes. Neither file is an Excel workbook.
Sources and editorial record
- NIST AI RMF: Govern, Map, Measure, Manage; voluntary framework
- Gartner CFO Report Q3 2026: AI value, risk and skills
Source pages accessed 18 September 2026. Research informs topic selection; formulas, example data and decision rules are original to Practical Finance. Source organisations have not endorsed this resource. Arithmetic and input-edge tests completed; no independent external technical review.