.jpeg)
AI governance in higher education: Five lessons for university leaders
2 September 2026
Insights
Capabilities
Conferences
AI governance in higher education: Five lessons for university leaders
As universities move from AI experimentation towards institution-wide adoption, the challenge is increasingly one of governance. With Generative AI being accessible to all, knowing where to invest, how to manage risk and how to turn promising pilots into something genuinely useful will be the difference maker between successful adoption, and ineffective pilots.
1. Be deliberate about experimentation
Universities need room to test AI, but experimentation should have a purpose. Speaking in a QS webinar, Jesus Truillo Gomez, Industry Executive for Education & Research, Google Public Sector, said that “experimentation is great, but being intentional about it” is key. He suggested prioritising pilots that create regular value for the people using them, rather than pursuing AI use cases simply because the technology is available.
That is particularly important as institutions face an expanding list of possible applications. A good starting question is not just whether AI can be used, but whether it solves a meaningful problem?
2. Look for value in everyday processes
Some of the clearest returns from AI may come from relatively ordinary university processes. Gomez highlighted process automation as an area where institutions can identify value quickly. “Everything that is a repetitive task is a possibility to get a quick ROI.” Examples discussed included elements of admissions processing and expense management.
For university leaders building the case for AI investment, small improvements across high-volume processes may prove just as valuable as more ambitious headline projects.
3. Build skills for different types of users
A single AI training programme will not meet the needs of an entire university.
General skills, such as prompting and responsible use, can be delivered at scale. But researchers and other advanced users often need much more specialised support. Gomez described a “blend of tailored workshop for power users plus scalable transversal skills” as a strong model for institutions.
The aim should be to give everyone a useful baseline while ensuring that people working with more complex tools, models or computing environments can develop deeper expertise.
4. Make governance an enabler
Governance does not have to mean stopping people from experimenting.
The discussion highlighted a changing relationship between university technology teams and researchers. Gomez noted that researchers increasingly approach central IT asking how they can access different AI models while remaining “secure and compliant” and avoiding mistakes.
At UC Riverside, this has included creating safe environments where staff are actively encouraged to experiment. Speaking on the webinar was Matthew Gunkel, CIO & Associate Vice-Chancellor at UC Riverside, who said “we just want them experimenting in safe sandboxes.” The goal is therefore not simply to control AI adoption, but to make responsible experimentation easier.
5. Build for change
Perhaps the simplest piece of advice from the discussion was also one of the most important. Asked what institutions should prioritise, Gomez's answer was: “Build flexibility.”
AI models, platforms and the needs of university communities are changing quickly. Institutions therefore need governance, infrastructure and skills strategies that can adapt with them. The aim is not to predict exactly what the university's AI environment will look like several years from now. It is to avoid building an approach that only works for the technology available today.
From experimentation to institutional capability
Effective AI governance in higher education is increasingly about creating the conditions in which useful ideas can scale. That means purposeful experimentation, clear routes to value, appropriate skills, governance that enables rather than blocks, and enough flexibility to respond as the technology evolves.
For universities, the next stage of AI adoption may depend less on how many tools they introduce and more on how effectively they create the institutional capability around them.
.webp)
