Most AI conversations in academia start at the wrong altitude: replacement, disruption, the future of the institution. Meanwhile the actual opportunity is sitting in the inbox. Summarising a forty-page consultation. Drafting the first pass of a committee minute. Turning a programme spreadsheet into a web page. Triage, not transformation.
Augmentation has a shape
- It lives inside existing tools. If using AI means leaving the workflow, it won’t be used. The wins come from wiring assistance into email, documents, and the systems of record.
- It drafts; people decide. The reviewer stays human and named.
- It is measured in time. Hours recovered per week is the honest metric — not abstractions about innovation.
Start with one workflow
Pick a single recurring task that is genuinely disliked, instrument it, augment it, and measure. One working pilot converts more sceptics than any strategy paper. The compounding effect is real: the time recovered funds attention for research, teaching, and members — the work the institution exists to do.
Three pilots that pay for themselves
Committee minutes: record the meeting, let the assistant produce a structured draft against your minute template, and have the secretary edit rather than transcribe. An hour becomes fifteen minutes, every meeting. Inbox triage for shared mailboxes: classification and suggested replies for the membership@ and info@ addresses, with humans sending. Document summarisation: consultation responses, grant guidance, and policy updates condensed to a page with references back to the source.
Each pilot has the same anatomy: a measurable baseline (minutes per task before), a tool already inside your data boundary, a named owner, and a four-week review. If the time saved is real, keep it and pick the next workflow. If it isn’t, stop — that result is just as useful.
What to avoid
- Anything member-facing without review — the reputational downside of one wrong automated reply outweighs a month of savings.
- Big-bang programmes. Appetite follows evidence, and evidence comes from small things measured honestly.
- Tools that require copying data out of your environment to be useful. The workflow is the product; a tool outside the workflow won’t be used.
- Counting ‘messages generated’ as success. Count hours returned to the mission.
Augmentation done this way is quiet. Six months in, nobody calls it an AI initiative — it’s just how the minutes get done now. That is what success looks like.
