Jisc’s Call to AI Suppliers: What It Means for Universities Using Microsoft Copilot

AI Adoption in Higher Education is Moving into its Next Phase

Higher education has spent the past few years working through what generative AI means for teaching, assessment, research and the student experience. Alongside those debates, a more operational question has been gathering pace: how do universities turn experimentation into something they can govern and support at institutional scale?

That’s the question behind Jisc’s sector-wide call to AI suppliers, backed by signatories including Universities UK, GuildHE and the Association of Colleges. It asks suppliers to give education and research organisations greater transparency and control over AI costs and usage, alongside clearer commitments on governance, model changes and the flexibility to switch providers or technologies.

Although the call is directed at suppliers, it reflects how AI conversations in higher education are changing. Universities are now deciding which uses deserve sustained investment, what controls need to sit around them and what wider adoption will demand from people, processes and technology.

Those questions are already coming through in our work with universities. Interest in AI is high, but institutions are at very different stages of turning that interest into everyday use. The conversations are increasingly about practicalities: where AI can genuinely help, what needs to be governed, how staff adopt it and how to judge whether an experiment is worth taking further.

Microsoft Copilot sits squarely within that shift. For universities already running much of their day-to-day work through Microsoft 365, it offers a natural route for AI into existing workflows. And as Copilot moves beyond early trials into broader use, the questions change. Universities need to think about the information it draws on, who has access, how usage is governed, what it costs and what value they’re getting in return. Agents add another dimension as individual teams begin applying AI to more specific problems.

Microsoft is also developing more of the controls, reporting and management capabilities organisations need as adoption grows. Copilot controls now span security and governance, management controls, and measurement and reporting, including licensing and metering, and agent lifecycle management. For universities, the practical work lies in deciding how those controls should be applied across the institution

These are all areas we’re working through with universities today, from AI readiness and Microsoft 365 governance to Copilot adoption and practical agent use cases. This article looks at what Jisc’s call means from the university side: what institutions using Microsoft 365 should be thinking about as AI moves beyond experimentation.

AI Governance Now Includes Cost & Value

Early conversations about AI governance have understandably focused on risk: data security, privacy, permissions, compliance and appropriate use. All of those remain fundamental. As adoption grows, however, governance also becomes a question of cost, control and value.

Jisc calls for clearer visibility of AI usage and expenditure, granular controls, budget management and reporting. It also asks suppliers to help institutions understand the factors driving consumption and how AI costs sit alongside existing licences, cloud commitments and other technology investments.

A contained Copilot trial can be relatively easy to manage: decide who gets access, agree what you want to test and learn from how people use it. Wider adoption brings more variables into play. Universities need to understand where Copilot is gaining traction, which use cases are worth supporting, whether licensed users are genuinely adopting it and where further investment will deliver the most value. As agents and AI-enabled workflows emerge across different teams, questions around ownership, governance and cost become part of that picture too.

At institutional scale, technical controls only cover part of the governance challenge. Universities also need enough information to make confident decisions about:

• where Copilot is being adopted and how it’s being used
• which use cases are demonstrating enough value to support further
• how costs are changing as adoption grows
• where ownership and funding sit as AI becomes part of everyday work

Cost and value therefore need to be looked at together. A relatively inexpensive deployment that delivers little benefit isn’t necessarily good value. Equally, a promising use case can become difficult to sustain if its costs increase as adoption grows, or if the institution has no clear way of deciding who should fund and own it.

The data available to make those decisions is improving. Microsoft is expanding the Copilot metrics available through Copilot Analytics, giving organisations a more detailed view of how Copilot is being used across Microsoft 365. Rather than stopping at whether someone is an active user, the newer metrics can reveal more about the types of work Copilot is supporting, whether that’s helping staff work through busy inboxes, drafting and refining documents, translating content, or cleaning and analysing data in Excel. This gives universities a clearer picture of how Copilot is becoming part of everyday work, and which patterns of use are emerging across the institution.

Copilot Adoption Needs Evidence, not just Activity

Better visibility of cost and usage gives universities a much stronger starting point, but it still only tells part of the story. Knowing how regularly Copilot is used and which tasks it supports shows where adoption is taking hold. It doesn’t tell you whether that activity is saving meaningful time, improving the quality of work or solving a problem the university considers important.

As Copilot moves beyond experimentation, that distinction becomes more important. A university might see healthy usage across a group of licensed users, but usage alone doesn’t tell you whether those licences represent good value or whether a particular use case deserves to be expanded.

If Copilot is helping a professional services team find policies more quickly, what was difficult or slow about that process before? If an internal communications team is using it for drafting or information discovery, has repetitive work reduced? If an agent is answering common staff questions, is it helping people find reliable answers while reducing pressure on another team?

The right measure will depend on the use case, so it helps to establish a baseline before the pilot starts and decide what change you expect to see. Without that, it becomes much harder to separate genuine improvement from positive feedback or increased usage.

Not every experiment needs a detailed business case. Universities need room to test ideas, learn from them and stop the ones that aren’t delivering. But as experimentation turns into sustained investment, there needs to be stronger evidence behind the decision to continue or expand it.

As Copilot becomes more established, universities will need to build a clearer picture of the value it’s creating. Where is it truly improving work, which use cases are proving worthwhile and what does the evidence say about where to go next?

Agents will Make Governance Harder, not Easier

Copilot agents give universities a more targeted way to apply AI to specific problems, but they also increase the number of AI use cases an institution needs to understand and govern.

That makes agents an attractive next step for universities moving beyond general Copilot experimentation. Rather than trying to apply AI broadly, teams can concentrate on a defined problem, a specific audience and a controlled source of information. A policy agent, for example, can help staff find and understand guidance held across an intranet, while other agents might be designed around recurring staff questions or research guidance.

These contained use cases can provide a practical route into AI because their purpose is easier to define and their impact easier to assess. But their relative accessibility creates another challenge: different teams and departments can begin developing agents around their own needs, leaving the institution with a growing collection to keep track of.

Each agent needs some basics in place from the outset:

• Purpose: what problem is it there to solve, and for whom?
• Ownership: who’s responsible for it once the initial experiment ends?
• Knowledge: what information does it rely on, and who maintains that information?
• Access: who can use it and what can it access?
• Review: how will its usefulness be assessed, and when should it be updated or retired?

Microsoft is building more structure around this too. Its agent governance guidance covers areas including access and availability, agent inventories, lifecycle management, data access and billing, while Copilot controls provide visibility into agents from deployment through ongoing governance and eventual retirement.

Cost increasingly belongs in the same conversation. Microsoft has introduced an agent cost-management approach covering estimation, tracking, forecasting and control of agent-related expenditure, reflecting the different financial considerations that come with consumption-based AI.

The technology can provide the controls and visibility, but universities still need to decide how they want agents to operate at institutional level: who can create and share them, which use cases the university will formally support, who owns them over time and when they should be reviewed or retired.

Without a shared approach, universities risk creating a new kind of digital sprawl: multiple agents built for different purposes, with varying levels of ownership, oversight and ongoing support.

What Should Universities Using Copilot do Now?

For institutions moving from early experimentation towards wider adoption, five questions provide a useful starting point.

1. What are we actually trying to improve?

Start with the institutional problem or everyday friction rather than the AI capability. Defined use cases make it easier to decide what’s worth testing and what success should look like.

2. Are our Microsoft 365 foundations ready for the use cases we are prioritising?

Review the content, permissions, ownership and governance supporting those use cases. The entire Microsoft 365 estate doesn’t have to be remediated before experimentation begins, but the information AI relies on needs appropriate ownership and oversight.

3. Do we understand both the cost and the value?

Understand what you’re investing, what adoption looks like and what improvement you expect to see. Microsoft’s increasingly detailed Copilot metrics can help build a clearer picture of usage, but that activity still needs to be connected back to the outcome the use case was designed to achieve.

4. Who owns AI once it moves beyond the pilot?

AI can’t sit indefinitely as an experiment owned by a small project group. Universities need technical governance alongside ownership of content, use cases, adoption and ongoing decisions.

5. What happens when experimentation starts to scale?

Consider the next stage early. More users, agents, automations and AI-enabled processes bring additional questions about governance, permissions, cost, ownership and lifecycle management.

For agents, look at how Microsoft’s controls around access, lifecycle and cost can support the university’s wider governance approach.

The Next Phase of AI in Higher Education

Jisc’s call is directed at AI suppliers, and there are encouraging signs of movement on some of the issues it raises. Within Microsoft’s ecosystem, richer Copilot usage metrics, management controls, agent lifecycle governance and approaches to agent cost management are giving organisations more information and control as adoption grows. Those capabilities don’t make the institutional decisions for universities, but they provide better tools for making them.

For universities using Microsoft Copilot, the focus now needs to be on how those capabilities are applied in practice: where AI is useful enough to invest in, what governance needs to sit around it, how adoption and cost are monitored, and what evidence will show whether it is delivering value. As agents become more accessible, that same discipline will need to extend to a growing range of AI use cases across different teams.

For us, this is where the supplier and institutional sides of Jisc’s call meet. Suppliers need to provide the transparency, controls and commercial clarity the sector is asking for. Universities need to use that information to make informed choices about adoption, investment and governance. Through our work with universities on Microsoft 365, AI readiness, Copilot and agents, we’re helping institutions work through that practical middle ground: understanding what the technology makes possible, putting the right structures around its use and building the evidence needed to decide what should go further.

Jisc’s call reflects a broader shift in the sector. AI adoption in higher education is entering a more mature phase, where access to the technology is only part of the picture. Sustainable adoption will depend on universities being able to turn experimentation into practical, well-governed uses of AI that deliver clear value for staff and the institution.