AI Knowledge Base for Universities: 5 Proven Fixes
AI Knowledge Base for Universities: 5 Proven Fixes
University and College Union analysis published in October 2025 put proposed cuts at UK universities at the equivalent of more than 15,000 jobs in a single year, nearly triple the figure recorded just months earlier. Every one of those departures takes something with it that no spreadsheet tracks: the unwritten knowledge of how admissions criteria actually get applied, which funding council rule overrides which policy, and who to ask when a safeguarding case does not fit the standard process. An AI knowledge base for universities exists precisely to capture that knowledge before it walks out the door, not after, and it is what askKIRA was built to do.
The redundancy wave is taking more than jobs
The scale of the disruption is hard to overstate. UCU’s analysis found university leaders had proposed staff reductions amounting to 12,230 jobs, with a further 3,224 roles at risk from £197 million in planned savings calculated on average staff costs, according to reporting by Times Higher Education. Behind each number sits a member of staff who knew, without needing to check a manual, how a particular process actually worked in practice.
Institutions losing senior admissions officers, registry staff and long-serving department administrators are not just losing headcount, they are losing the informal index that made policies usable day to day. A written procedure rarely captures every exception, and it is the exceptions that new or remaining staff struggle with most once the person who used to field those questions has gone. This is precisely the gap a properly built AI knowledge base for universities is designed to close.
Why higher education’s knowledge problem is different
Knowledge in a university does not sit in one place. Admissions criteria live with one team, funding council compliance with another, safeguarding procedures with a third, and course-specific exceptions with individual departments that rarely talk to each other. A registry officer, a finance administrator and a students’ union adviser can all give a different answer to the same question about mitigating circumstances, and each answer might have been correct at a different point in the policy’s history.
That fragmentation was manageable when experienced staff stayed for years and simply remembered the exceptions. It becomes unmanageable the moment restructuring accelerates staff turnover, interim cover becomes normal, and remaining teams inherit responsibilities they were never trained to hold. As we set out in Universities and the AI Divide, the gap this creates is not only about who has access to the latest tools, it is also about who inside an institution can reliably get a straight answer when they need one.
What an AI knowledge base for universities actually needs to do
- Capture institutional knowledge before it leaves. Upload policies, procedures, funding council guidance and department handbooks as they exist today, in whatever format they are in. askKIRA turns them into a searchable knowledge base rather than a folder that only one retiring administrator understood, which is exactly the first job an AI knowledge base for universities needs to do.
- Give every campus and department the same answer. Registry, admissions, student services and finance all draw on the same governed source, so a mitigating circumstances query gets the same answer wherever it is asked, regardless of who is asked or how recently they joined.
- Shorten onboarding for interim and new staff. During a restructure, new or temporary staff can ask a direct question in plain language and receive an answer grounded in actual, approved documentation, rather than waiting for a colleague who may themselves be new.
- Keep sensitive material under proper control. Role-based access means admissions criteria, HR policy and safeguarding guidance stay visible only to the teams who need them, while general student-facing information remains open to everyone who should see it.
- Maintain a version history auditors and regulators can trust. Every update, approval and revision is tracked, so when a funding council or accreditation body asks how a policy was applied on a given date, the answer does not depend on somebody’s memory.
Keeping control while staff turn to public AI anyway
Institutional pressure does not stop staff from finding their own workarounds. When internal policies are hard to search, staff increasingly paste them into public AI chatbots to get a quick summary, which is exactly the kind of unsanctioned use that creates a data protection problem rather than solving one. The ICO’s guidance on AI and data protection is clear that UK GDPR obligations, including lawful basis, data minimisation and accountability, apply in full whenever personal data is processed through an AI system, whichever platform staff choose to use.
A properly governed AI knowledge base for universities gives staff a sanctioned alternative that is quicker than searching a shared drive, so there is less reason to reach for a public tool with student or staff data in the first place. That is the difference between a knowledge platform staff want to use and a policy staff are told to follow but quietly ignore.
Where to start
The strongest pilot for an AI knowledge base for universities is usually the team fielding the most repetitive queries under the most pressure, typically registry or admissions during a restructuring period. Load the current policies and procedures, measure how many questions get answered without escalation in the first month, and compare that to how long the same questions used to take when they depended on one person’s memory. That single comparison usually makes the case for wider rollout without needing to argue the principle again.
An AI knowledge base for universities will not stop institutions having to make hard staffing decisions. What askKIRA does mean is that when someone with fifteen years of institutional memory leaves, what they knew does not have to leave with them.
Comments are closed