From Experimentation to Operations: Scaling AI Across the Modern University

Key Insights and Briefing for UCISA members based on a sponsored webinar with AppsAnywhere and LabStats on 22^nd October 2026.

Universities across the UK are entering a new phase of AI adoption. The conversation is no longer centred on whether AI should be used, but how institutions can operationalise it responsibly, sustainably and at scale. That was the focus of a recent UCISA-hosted webinar for its members, featuring insights from Aaron Kingsbury, Head of Service Delivery at University of Reading, and John Felushko, Product Manager at LabStats.  

The discussion explored the practical realities of moving from experimentation and pilot projects towards institution-wide adoption, balancing innovation with governance, and ensuring AI investments deliver measurable value. This article summarises the key themes, audience poll results and considerations for higher education leaders navigating the next stage of their AI journey.  

The Current State of AI in Higher Education

AI has become one of the most talked-about topics in higher education. Much of the initial conversation focused on teaching and learning, but increasingly, they're no longer the only questions universities need to answer.

Today, AI is moving beyond the classroom and into the operational fabric of the university itself. It's increasingly being used across recruitment, marketing, student services, research, IT operations, cybersecurity, administration and decision-making.

A recurring theme throughout the discussion was that most universities are still in the early stages of their AI journey, despite widespread activity across the sector. This is validated by Jisc’s 2026 Higher Education Digital Maturity Assessment, with most respondents on Level 2 out of 5.  

According to Aaron Kingsbury, AI adoption is happening through a combination of dedicated AI initiatives and AI capabilities being embedded into existing software platforms. Rather than large, transformational AI programmes, many institutions are seeing organic adoption emerge from individual departments and teams identifying practical opportunities to improve workflows and productivity.  

This expansion creates both opportunity and complexity, particularly as different stakeholder groups have very different expectations and use cases for AI. Students often prioritise accessibility and trusted tools, academics seek support without compromising academic standards, institutional leaders focus on return on investment, while IT leaders remain concerned with governance, security and service delivery.  

Participants were asked where they had seen the greatest AI adoption within their institutions.

Poll Results: Where Is AI Being Adopted?

AREA:                                                 RESPONSE:

ACADEMICS                                      30%

RESEARCH                                        13%

STUDENTS                                        13%

ADMINISTRATIVE TEAMS                10%

TEACHING & LEARNING TEAMS     7%

MARKETING                                      3%

IT SERVICES                                      3%

STUDENT SUPPORT SERVICES      0%

LEADERSHIP                                     0%

ALL OF THE ABOVE                          20%

‍

Results suggest that adoption remains strongest within academic and research communities, but the significant "all of the above" response highlights that AI is becoming increasingly widespread across institutions.

Perhaps most notable was the absence of responses for student support services and leadership, indicating either lower adoption rates or less visibility of AI initiatives within those functions.

What stands out here is the emerging complexity of managing AI use across multiple stakeholders, use cases and tools.

The Challenge of Governance in an Era of Choice

One of the strongest discussion points centred around governance.

Universities are encountering a growing challenge around platform choice. While institutions may standardise around a platform such as Microsoft Copilot, staff and researchers often prefer alternative tools for specific use cases.

At the University of Reading, for example, some departments have identified stronger alignment with tools such as Claude for their particular workflows. This raises an important question, should universities standardise tools or standardise governance?  

The panel suggested that future governance models may need to focus less on controlling specific technologies and more on establishing guardrails that allow safe experimentation and flexibility. This is particularly important as AI capabilities are increasingly being embedded into existing SaaS platforms, often without institutions actively procuring dedicated AI solutions.  

The emergence of "shadow AI" was also highlighted. Data from LabStats shows many institutions are often surprised by the number of AI tools already being used beyond approved technology stacks, creating challenges around oversight, compliance and cost control.  

AI Is Changing User Behaviour Faster Than Previous Technology Waves

A particularly striking observation from the webinar was the speed at which AI is changing technology usage patterns.

Examples shared included:

• Significant reductions in usage of specialist computing resources as AI-powered tools provide alternative workflows.
• Changes to assessment strategies, with some institutions exploring alternatives to traditional computer-based examinations.
• Rapid growth in AI tool adoption across staff and student communities.
• The emergence of "super users" who account for the majority of AI consumption within institutions.  

Unlike previous software trends, AI is not simply adding new applications. It is fundamentally changing how existing systems, services and resources are used.  

Practical Examples from the University of Reading

The webinar also provided several examples of AI implementation.

1. Improving Knowledge Management

Rather than focusing immediately on chatbots, the University of Reading first concentrated on improving the quality of its knowledge base.

Using AI agents within Microsoft Copilot, existing knowledge articles are reviewed, structured and enhanced to support:

• Better self-service experiences
• Future chatbot effectiveness
• Improved service desk deflection
• More accurate agent assistance within ServiceNow

This approach recognised that AI effectiveness depends heavily on the quality of underlying content and data.  

2. Smarter Software Delivery

The university has also leveraged AI capabilities within cybersecurity tools to identify locally installed applications across its estate, uncovering thousands of unmanaged software installations.

These insights help inform software packaging and deployment decisions, reducing manual effort and supporting compliance requirements.  

3. AI-Assisted Service Operations

Network and service management teams are using AI-powered troubleshooting tools to identify and resolve issues more quickly.

By embedding AI into existing operational systems, Reading has improved service responsiveness without necessarily introducing standalone AI platforms.  

A key takeaway was that many organisations are already owning AI capabilities within current software investments but may not yet be fully exploiting them.  

Data: The Foundation of AI Success

One of the clearest messages from the webinar was that successful AI adoption depends on data readiness.

In fact, the University of Reading delayed aspects of its Microsoft Copilot rollout while reviewing data governance and access permissions, to avoid the possibility of AI tools inadvertently exposing inappropriate institutional information through broad sharing permissions that already existed within Microsoft 365.  

The panel repeatedly stressed that AI does not solve poor data management.

Institutions looking to scale AI should first focus on:

• Data quality
• Permissions management
• Information governance
• Content lifecycle management
• Security controls

Without these foundations, AI can amplify existing risks.  

Poll Results: What Concerns you most about wider AI adoption?

CONCERN:                                              RESPONSE:

DATA SECURITY                                        23%

ACADEMIC INTEGRITY                             4%

COST ESCALATION                                  4%

LACK OF MEASURABLE OUTCOMES     4%

GOVERNANCE                                           0%

REGULATORY COMPLIANCE                   0%

ALL OF THE ABOVE                                  65%

‍

The strongest message here is the breadth of concerns.

Nearly two-thirds of respondents selected "all of the above", highlighting that institutions do not see governance, cost, security and value as isolated challenges. Instead, they are interconnected considerations that must be managed together.  

Data security emerged as the single largest individual concern, reinforcing the importance of information governance and access controls.  

Measuring Success: Are We Asking the Right Questions?

Perhaps the most challenging topic discussed was measurement.

Many institutions are investing in AI but have yet to establish clear frameworks for evaluating success.

The panel agreed that measuring outcomes remains difficult because many organisations are still experimenting. Traditional ROI calculations often fail to capture wider benefits such as improved user experiences, reduced friction and better service delivery.

However, one consistent theme emerged: Data should drive AI investment decisions.

Instead of asking where AI could be applied, institutions should ask:

• Where is time currently being lost?
• Which processes create the greatest friction?
• What problems are users trying to solve?
• Where can measurable outcomes be achieved?  
• The audience was asked which metric do you see as the most important when evaluating AI success?

Poll Results: How Should AI Success Be Measured?

METRIC:                                                         RESPONSE:

STUDENT OUTCOMES & EXPERIENCE        40%

RETURN ON INVESTMENT (ROI)                  35%

USER ADOPTION                                            5%

STAFF PRODUCTIVITY GAINS                       5%

COST CONTROL                                             0%

ALL OF THE ABOVE                                        15%

‍

The results reveal an important sector perspective.

While financial returns remain important, respondents placed student outcomes and experience above all other individual metrics. This reflects a broader recognition that AI investments should ultimately support institutional missions and improve the student experience, rather than simply reduce costs. It would be interesting to explore how this fits into existing outcome measurements or whether new methods will need to be formed.

The Next Major Challenge: Cost

As universities move from experimentation to operational deployment, the panel identified cost as a likely future challenge.

Current AI adoption has been fuelled by rapid innovation and substantial investment from technology providers. However, there are questions about the long-term sustainability of AI pricing models, usage-based charging and token consumption. Institutions may find it increasingly difficult to forecast and manage expenditure as AI becomes embedded across services and platforms.  

This creates a growing need for:

• Visibility into AI usage
• Understanding demand patterns
• Monitoring adoption trends
• Linking usage to outcomes
• Managing licensing and consumption costs

Without this visibility, institutions risk overprovisioning AI services in exactly the same way they have historically overprovisioned software, hardware and infrastructure.  

Key Takeaways for University IT Leaders

For many institutions, one of the biggest challenge ahead is embedding AI into day-to-day operations in a way that is sustainable, governed and measurable.

The webinar highlighted five critical priorities:

1. Focus on outcomes, not technology. AI should be treated as an enabler, not the objective.

2. Get data foundations right first. Information governance and permissions are essential prerequisites.

3. Balance governance with flexibility. Institutions need guardrails that support experimentation, not restrict innovation.

4. Measure what matters. Student outcomes, adoption, productivity and ROI all have a role to play.

5. Prepare for increasing cost and complexity. Visibility into usage and demand will become increasingly important as AI matures. Where AppsAnywhere and LabStats can help

Tools like LabStats can help identify which AI tools are being used sent directly to email via insight reports. This can not only help to see whether AI investments are valuable and that the demand meets the reality of what’s really being used but help to monitor and predict costs associated to AI adoption.  

Final Reflection

A powerful observation from the session was that universities may need to stop asking ,"Where can we use AI?" and instead begin asking ,"What are we trying to improve?".

Those institutions that focus on solving real operational challenges, supported by reliable data, effective governance and meaningful measurement, are likely to be the ones that realise the greatest value from AI over the coming years. As higher education moves beyond experimentation, operational excellence may prove to be the defining factor in successful AI adoption.

Where AppsAnywhere and LabStats can help

AppsAnywhere provides equitable access to all your AI software in one centralised platform, providing students with the security and confidence to chose from institution approved tools that can be accessed on any device, anywhere.

LabStats provides real-time usage insight across hardware, software, and lab environments including the use of AI software. Usage data can be used to create visibility into AI tools usage, assess the value of AI investments and match provision to real usage to predict and monitor costs associated to AI adoption.  

Together we are uniquely positioned to help your campus computing meet student demand, to find out more visit www.labstats.com and www.appsanywhere.com.