AI in Higher Education: what we must do next

Chevonne Hobbs reports on the challenges facing our industry as we prepare the next generation to manage AI.

I recently attended the itSMF UK’s online sector spotlight on AI in higher education. The three university case studies presented from Glasgow, Birmingham and Oxford, highlighted their advantages of using AI within their institutions.

These were great presentations, and a clear requirement of where IT Service Managers have an important role to play in all of this. As some of the challenges were presented, the themes were recognisable to ITSM professionals: unclear ownership, fragmented support, variable data quality, changing user expectations, lacking governance, and user adoption.

Stanford University 2026 AI Insights Report

In a recent AI Insights Report from Stanford University; they conducted a survey spanning 15 countries, which showed 80% of university students using generative AI to support their studies, this is double the percentage reported back in 2023.

The report mentions that 56% of students use AI to look up key concepts or subjects. This was followed by researching assignments at 52%, generating initial ideas at 46%, and writing or editing work at 41%.

Students also perceive tangible benefits from using AI. Half reported better understanding of the topics they were learning, 49% said AI helped them complete assignments and 55% believed it helped them learn faster. At the same time, the report highlights legitimate concern, where 55% college students said generative AI had produced a mixed effect on their critical-thinking skills.

This mixture of value and concern is recognisable to any service manager. Users adopt a useful technology faster than the organisation can develop a policy, deliver a suitable AI solution, maintain and support it.

This is where having a strategy for AI is imperative to any organisation, whether you are an institution or a corporate organisation. Building a strategy based on real use cases that drive efficiency from the benefits of implementing AI, also requires a solid AI-Operating Model foundation to support the future state. From an AI-Operating Model, you can start to understand the requirements for your processes, people, data, technology, and supplier dependencies.

University of Glasgow RO:bot chatbot

The University of Glasgow’s Reach Out Chatbot (RO:bot) service brings together support for their IT team and for Student services through a shared front door. Users can search for answers by interacting with RO:bot, using a live chat or request face-to-face assistance without first having to understand which internal team can help with their request.

The University of Glasgow achieved this by providing users with:

• one recognisable point of entry.
• access to cross-departmental knowledge.
• sufficient context to avoid repeatedly asking the same questions.
• transparency between automation and human support.

The objective showed higher satisfaction feedback from IT and Students who said it helped them with less manual navigation that typically led to time delays and frustrations.

University of Birmingham: AI improves operational intelligence.

The University of Birmingham case study focused on bespoke AI capabilities integrated with ServiceNow. The institution has used AI to analyse incident volumes, identify trends and highlight areas where problem records may be missing. It has also explored AI-generated summaries of priority incidents, assessments of the completeness and likely disruption of change records, and transcription of post-incident reviews to produce timelines, actions, and recommendations.

This is a strong AI use case used in ITSM because AI can help by reducing the administrative burden and highlight patterns that genuinely take a long time to analyse with human investigation. However, the human oversight should not be removed completely as an AI model may overlook the context known to an experienced analyst.

The session also emphasised the importance of staff knowing the ethical use of AI, and clearly communicating how the data is being stored, processed, and used.

University of Oxford: multiple products, one service

The University of Oxford case study offered another valuable message: ChatGPT, Microsoft Copilot, Google Gemini, and other platforms each have their strengths. Rather than searching for a single “best” tool, Oxford’s approach has been to create an environment in which staff and students can use multiple approved AI tools safely and securely for different task objectives.

The University has given access to their enterprise agreements, security assessments, guidance, training and communities such as AI ambassadors and specialist interest groups. They also have set-up an AI Competency Centre providing a central point for guidance on the available technology portfolio.

This confirms that different AI models will continue to provide a combination of capability to an organisation, including costs, accessibility, integration, and data protection. AI models current performance will also change frequently, and this is something to be mindful of whilst maintaining a close relationship with your AI model suppliers.

The University’s approach is not to build the AI strategy around one product. It is to build a governed service wrapper around a managed portfolio of AI products.

Key takeaways

These sessions highlighted three main practical use cases I have seen for AI to date:

• Chatbots
• Workflow integration
• Productivity efficiency

For the Use Cases to be successful, the presenters shared the following considerations:

1. Responsible AI is included and embedded within your AI Strategy
2. Personas & User Journey Maps should be used when designing for AI.
3. Data Quality should be a policy standard that is consistently followed.
4. Human Escalation needs to exist, and AI should be included within your RACIs.
5. Multiple AI Tools can be used, when it’s for the right job.

Useful references

If you would like to follow up with some further reading, please use the reference topics mentioned in this article:

Responsible AI – CGI

How to manage the unmanageable – University of Oxford

Trends, patterns, and timelines using AI – University of Birmingham

Reach Out – University of Glasgow

Artificial Intelligence Index Report 2026 – Stanford University