The framework on one page | AI Cost, Use, and Outcomes Framework
Home
Evaluation and decision framework · Higher education

AI Cost, Use, and Outcomes

Central question: For a defined use case, does AI improve outcomes compared with the current approach or another feasible alternative, at acceptable cost and risk? And who benefits or bears the burden?

1 · Start A use case

If we use [AI capability] in [workflow], we expect [outcome] because [mechanism], provided [conditions] are met.

Baseline Comparator Review band
→
2 · MeasureWhat happened?
ACost and investment. Budget, plus staff time for review, correction, transition, and exit.
BUse and integration. Stages, in order, not one adoption rate:Access↓ Use in workflow↓ Quality↓ Performance↓ Mission impact↓ Sustainable value
COutcomes. Efficiency vs. effectiveness. Observed change vs. attributed effect.
→
3 · JudgeGood enough to act on?
Five lenses, each a gate:
Quality
Capability and readiness
Equity and accessibility
Governance and risk
Evidence quality
Pass · Watch · Fail. One fail stops it, however strong the rest. The bar rises with the review band.
→
4 · DecideWhat next?
Continue
Meets pilot requirements
Improve
Time-boxed fix and re-test
Expand
Evidence and capacity support it
Restrict or pause
A gate failed
Retire
Not worth it, or better option
↺Monitor and loop back. Review at the band’s cadence. Re-evaluate on a model change, quality drop, incident, or cost rise.
At every stepWho uses it?
IndividualUse AI for this task, and how?
TeamRedesign, continue, or stop this workflow?
UnitWhich use cases get resources?
InstitutionFund, standardize, restrict, or retire?
Keep these apart AccessvsMeaningful use Time releasedvsRealized savings EfficiencyvsEffectiveness Observed changevsAttributable impact
Review bands · minimum quality Low Self-assessment0.60 Moderate Cross-functional review0.70 High Specialist review0.75 Critical Executive decision0.80 Bands set by the AI Strategic Compass at intake.
Where it fits
AI Strategic Compass→ Campus AI Registry→ This framework
Compass: what to start. Registry: the record. This framework: what happens next.
Measure with the domains. Judge with the lenses. Decide at the right level. Optimize institutional value, not AI adoption.