Staff meeting summaries with an AI assistant
A 12-person communications office uses the campus AI assistant to draft weekly meeting summaries and action items. Staff edit every draft before sharing.
Illustrative example. Figures are invented to show the method.
Outcome hypothesis
If we use an AI assistant in weekly team meeting notes, we expect staff time per summary to drop by half because drafting starts from an edit instead of a blank page, provided summaries pass a quick accuracy check and no confidential data enters the tool.
What was measured
- No incremental budget cost: the license is already covered by a campus agreement
- 30 minutes of training per person
- About 8 minutes of review and correction per summary
- Access: 12 of 12 staff
- Use: used in 9 of 12 staff members’ meetings
- Quality: 85% of drafts accepted after light edits
- Time per summary: 35 to 14 minutes, including review
- Missed action items: no change
- Two misattributed action items found in a 40-summary spot check
Lens gates
| Lens | Result | Finding |
|---|---|---|
| Quality | Pass | Spot check met the team’s bar. Misattributed owners were the main error. |
| Capability and readiness | Watch | Two staff skipped review at first and needed coaching. |
| Equity and accessibility | Pass | Works with the screen reader one staff member uses. |
| Governance and risk | Pass | Approved tool, internal data only, personal accounts not used. |
| Evidence quality | Watch | Four weeks of before-and-after timing. Small sample, no control group. |
What can be claimed
Task time reduced, and about 3.5 hours a week of capacity released across the team.
Financial savings. No expenditure went down and no work was cut.
Results meet the pilot bar at almost no added cost. Keep it as standard practice for the team. No expansion funding is needed.
Next review in two terms, or sooner if the tool changes how it handles data.
Report released hours as capacity, not dollars. Converting them to salary savings would overstate the result.