Executive use case review · Example
Campus-wide AI assistant licenses
Renewal review for the president’s cabinet
Level Institution
Depth 4 · Formal evaluation
Review band High
Presented by Office of the CIO, with the Provost
AI Cost, Use, and Outcomes Framework by Joe Sabado · CC BY-NC-SA 4.0 · Illustrative figures
How the hour runs
Agenda
- 0:00Decision requested5 min
- 0:05Use case and baseline5 min
- 0:10Evidence: Cost, Use, and Outcomes15 min
- 0:25Five checks and risks5 min
- 0:30Options and recommendation10 min
- 0:40Discussion and decision20 min
Decision requested
Recommendation: continue with conditions
Why
Units that trained their staff show real redeployed capacity.
Only 9% use it weekly, so we pay for many unused seats.
Evidence is mostly self-reported; three checks are on watch.
We ask the cabinet to approve
1. Renew at 3,000 seats, down from 6,200.
2. Require each unit to report one outcome measure.
3. Review again in November 2027.
The use case
What we approved, and for whom
Who uses it
6,200 licensed faculty and staff
What AI does
Drafts, summarizes, and answers questions. A person reviews every output.
Why we bought it
Replace scattered unit subscriptions with one governed tool.
Outcome hypothesis: if staff use one approved assistant with training, units will redeploy time to higher-value work and stop buying their own tools.
Before AI
Baseline, fall 2025
14Separate AI subscriptionsBought by units, with no shared data rules
$95KSpent by units each yearOn tools the campus license could replace
NoneShared measure of valueNo unit reported outcomes from AI
Source: procurement records and unit survey, fall 2025.
How we measured
Evaluation design
| Element | What we did |
|---|
| Period | 12 months, October 2025 to September 2026 |
| Cost | Finance records, plus support staff time |
| Use | Vendor usage logs, by role and unit |
| Outcomes | Survey of weekly users, plus documented results from 9 units |
| Limitation | No comparison group; time saved is self-reported |
A · Cost
Total annual cost
| Item | Annual |
|---|
| Licenses, 6,200 seats | $410,000 |
| Training and support, 1.5 FTE | $165,000 |
| Setup and integration | $40,000 |
| Governance and privacy review | $25,000 |
| Unit subscriptions retired | −$95,000 |
| Net annual cost | $545,000 |
$231Per monthly user2,356 people
$977Per weekly user558 people
Adequate
B · Use
From access to real use
Signed in at least once4,030
In a documented workflow186
WeakWeekly use was three times higher in units that ran their own training.
C · Outcomes
What changed
Documented
- Admissions: 1,100 more applicant follow-ups per term
- Library: chat backlog cleared with the same staff
- 2 more units: reports drafted in half the time
Self-reported or negative
- 61% of weekly users report saving time (median 2.5 hours a week)
- 3 low-severity data incidents, all resolved
- Errors reported in policy answers
AdequateReal gains where documented; most units report time saved, not outcomes.
Capacity and savings
Capacity gains and cash savings
Time reduced2.5 hrs a weekSelf-reported by weekly users
Capacity released9 unitsDocumented freed hours
Capacity redeployed4 unitsHours moved to new work, with evidence
Money saved$95,000Retired subscriptions only
Released hours are capacity, not budget savings. Only the $95,000 in retired subscriptions reduced spending.
Five checks
Results of the five checks
| Check | Result | Evidence |
|---|
| Quality | Pass | Audit of 200 outputs: 94% acceptable |
| Capability and readiness | Watch | Training completion 31% |
| Equity and accessibility | Watch | Lower use in student-facing units; one screen reader issue |
| Governance and risk | Pass | 3 low-severity incidents, all resolved |
| Evidence quality | Watch | Self-reported time; no comparison group |
Options
Options considered
| Option | Annual cost | What happens |
|---|
| Renew as is | $545,000 | Keeps paying for about 5,600 seats not used weekly |
| Continue with conditions | $333,000 | 3,000 seats; units report one outcome; review in a year |
| Expand to students | $900,000+ | Not ready: equity and evidence checks on watch |
| Retire | $0 | 4 units lose capacity; units likely re-buy their own tools |
Right-sizing to 3,000 seats saves about $212,000 a year.
If approved
Conditions, owners, and dates
| Action | Owner | By |
|---|
| Renew at 3,000 seats | CIO | June 2027 |
| Name one outcome measure per unit | Unit leads | March 2027 |
| Raise training completion to 60% | Teaching center and HR | May 2027 |
| Fix the screen reader issue with the vendor | IT accessibility | Within 90 days |
| Re-review against these results | AI governance council | November 2027 |
Discussion and decision
Questions for the cabinet
- Is self-reported time enough evidence for renewal, or do we require documented outcomes?
- What would we need to see before offering it to students?
- Which units should report outcomes first?
Decision: ____________________Owner: ____________Next review: ____________
Appendix
Method and data sources
Data sources
- Finance and procurement records
- Vendor usage logs, by role and unit
- Survey of 558 weekly users, 44% response
- Unit reports from 9 units
Ratings and rules
- Ratings use the shared definitions in the decision guide
- Decision follows rule 7: a check on watch
- Review band High: council decides, reviewed at least yearly
AI Cost, Use, and Outcomes Framework by Joe Sabado · CC BY-NC-SA 4.0 · All figures are illustrative.