Training programmes alone won't make your university AI-literate
Most universities have responded to generative AI with the familiar playbook: training sessions, certification programmes, faculty workshops. These are reasonable first steps. But they are unlikely to be sufficient.
The problem runs deeper than attendance numbers. AI capability becomes unevenly distributed across departments. Teams with early adopters pull ahead; those without them fall further behind. A university where hundreds of staff complete an AI short course, but whose curriculum approvals, ethics frameworks and administrative workflows remain untouched, has not become AI-literate.
The strategic question is not "how do we train more people?" but "how do we redesign the institution so AI literacy becomes a natural property of how it operates?"
Five system-level strategies
1. Embed literacy into infrastructure, not programmes
The most durable path runs through processes people already use. When curriculum design templates prompt teams to consider AI-enabled learning objectives, literacy becomes part of course development, not an afterthought. When funding applications include a section on AI-enabled methodology, researchers engage as routine. Infrastructure scales where programmes cannot: a governance template reaches every course team; a voluntary workshop reaches only those who attend.
2. Distribute intelligence rather than centralise training
Central AI teams create bottlenecks and cannot serve the breadth of disciplinary contexts across a university. A more sustainable model distributes capability through departmental champions: staff with formal recognition, protected time and integration into team-level planning who translate institutional strategy into disciplinary practice. Engineering and humanities will adopt AI differently; distributed models respect that variation while maintaining coherence through shared principles.
3. Redesign incentives to signal institutional value
If promotion criteria, workload models and teaching awards do not recognise AI-enhanced practice, adoption remains the province of enthusiasts. Including innovative AI use in promotion portfolios or weighting it in excellence awards reshapes behaviour far more effectively than mandatory workshops. Incentives are the institution's operating system. Change them, and you change what people prioritise.
4. Design invisible learning into everyday work
The most effective AI literacy development often does not look like learning. When staff regularly use AI-generated insights to support student advising, or review AI-generated meeting summaries as routine, they build critical evaluation skills without attending a seminar. This principle-embedding AI literacy in operational routines-mirrors how organisations have scaled data protection awareness and safety compliance. Design AI encounters into daily work, not training calendars.
5. Surface and formalise 'shadow' AI usage
Every university already has significant informal AI activity. Academics draft grant proposals with language models; administrators summarise meetings with AI tools. Rather than suppressing this, strategically mature institutions surface and formalise it. An AI usage audit, framed as capability mapping (asking teams "Which AI tools are you already using effectively, and where would you welcome more support?" rather than "are you compliant with policy?") reveals where expertise exists, where risks concentrate and where formal support would deliver the highest return.
Different entry points for different institutions
For early-stage institutions, the entry point is infrastructure: audit existing processes and embed simple AI prompts into existing forms. A question on AI-enabled methodology in grant application templates, or a short section on AI-supported teaching in curriculum design forms, requires no new systems.
For digitally mature institutions, the priority shifts to redesigning rewards and recognition, and building AI capability across departments through local champions. This demands more confidence but delivers deeper cultural change.
Faculty resistance to AI integration often reflects legitimate concerns rooted in professional standards. Institutional policies on responsible AI use are still being developed, and adoption will be uneven because AI's relevance varies across fields.
Two counterintuitive insights
Formal training can slow adoption when it creates the impression that AI literacy requires certification before engagement. Staff hold back; the training requirement becomes the barrier. Institutions that normalise experimentation, framing early, imperfect use as legitimate, see faster and more sustainable uptake.
Governance ambiguity should be treated as a design constraint, not a reason for delay. Lightweight frameworks that enable responsible experimentation-such as a one-page set of principles on disclosure, accuracy and student data-prove more productive than comprehensive policies that defer action.
The measure of success
Within a decade, AI literacy will likely sit alongside research capability and teaching quality as a defining marker of institutional strength. Universities that treat it as a training line-item risk managing a growing capability deficit. Those that design it into processes, incentives and governance will find that literacy compounds in ways training alone never achieves.
Perhaps the most telling sign of success will be when the concept itself becomes unnecessary. When AI capability is woven so thoroughly into how an institution teaches, researches and operates that it needs no separate label, the transformation is complete. The goal is not a university full of people who have learned about AI. It is a university that has learned to think with it.
For more on implementing AI across educational settings, see AI for Education and Generative AI and LLM.
Your membership also unlocks:
