1. What are the four multipliers in a benefit model?○ Cost, time, quality and risk○ Volume, coverage, adoption and unit value○ Build, run, change and contingency○ Feasibility, readiness, value and risk
2. Where should the coverage figure come from?○ The vendor's published accuracy claim○ The straightforward pile of your own hundred-case sample○ An industry benchmark for similar projects○ The technical team's target accuracy
3. Why is adoption the term that most determines the outcome?○ Because it is the easiest to measure○ Because it can swing the benefit by a factor of several, more than any further accuracy improvement will — and it is the term a manager can most directly influence○ Because vendors guarantee it contractually○ Because regulators require adoption reporting
4. Through which routes does saved time become financial benefit? Select all that apply.☐ Headcount reduction where the decision is real and agreed☐ Absorbed growth — volume rising without adding people☐ Redeployed output producing something you would otherwise have bought☐ Improved staff morale from removing tedious work
5. What are the three cost buckets?○ Licence, hardware and training○ Build, run and change○ Capital, operating and contingency○ Internal, external and contingency
6. How do generative AI running costs differ from traditional software licences?○ They are always lower because there are no per-seat charges○ They are metered by usage, so success increases cost and the cost per case is largely determined by design decisions○ They are fixed at contract signature○ They decline automatically each year
7. A system handles 80% of 40,000 annual invoices. Exceptions take about twice as long as an average invoice used to. What does this mean for the case?○ Nothing — exception handling is a transitional cost during rollout○ The exception path is a permanent operating cost that can roughly halve the net benefit, and must appear in the run bucket○ The coverage figure should be revised upward○ The project should be abandoned
8. Why should a business case be presented over at least three years?○ Because most finance teams require three-year horizons○ Because year one carries build costs against partial benefit and climbing adoption, so a project judged on year one gets killed just before it starts working — and year two is the honest test○ Because AI models take three years to reach full accuracy○ Because vendor contracts are typically three years
9. Which sections most distinguish a case written by someone experienced?○ Market context, competitor examples and vendor architecture○ What has to be true, the stop condition, and what the project leaves behind○ The executive summary and the appendix○ The risk register and the project plan
10. What does it mean to write a case so it can be checked?○ Having it reviewed by internal audit before submission○ Expressing every claim in measurable form with a date attached, so that in twelve months someone can determine whether it was right○ Including all supporting calculations in an appendix○ Obtaining written sign-off from every affected department