How to run an AI agent readiness assessment in a university

Assess one professional services team, usually admissions or student records, well before the next admissions peak. Staff take a test set in higher education on held actions, the provenance of applicants' documents, decisions that belong to academics, and the record. The report shows gaps by concept and decides whether the first agent answers enquiries, checks documents or drafts offers.

What is the assessment for in a university?

Universities handle seasonal floods of contact: applications, offers, enrolment, fees, timetables, results. Agents can answer routine enquiries, check that applications are complete, draft offer and enrolment letters from approved templates, and chase missing documents. Admissions and academic decisions, such as whether an applicant meets entry requirements by exception or whether a student progresses, stay with academic staff.

The assessment checks whether the professional services staff who will review agent work understand where that line falls and what the agent will hold. The director of admissions or the academic registrar usually owns it, with the data protection lead. For the concept, see what a readiness assessment is.

Who should take it?

One team of up to 25:

• Admissions officers, if the first agent supports applications.
• Student records or registry staff, if it supports enrolment and records.
• Team leaders who handle exceptions and complaints.
• One academic admissions tutor as an observer, to confirm the questions on academic judgement match practice.

Academic staff more widely take their own sitting later, if their work will involve agents. Advice on order is in who in a university needs AI agent training first.

The steps

1. Person (director of admissions): chooses the team and, with academic colleagues, lists decisions the agent never makes: admission by exception, credit for prior learning, progression and anything involving a student's circumstances.
2. Person (director of admissions): asks for the test set in higher education.
3. Person (each member): sits the test alone, at least three months before the admissions peak.
4. Person (director of admissions and data protection lead): read the report by concept.
5. Person (director of admissions): chooses the first agent and its depth. A common start is answering routine applicant enquiries at depth 3 from approved answers, with document chasing drafted at depth 2.
6. Person (team leaders): train the weakest concept before the peak.
7. Agent, depth 1: reads last cycle's closed enquiries and applications in shadow, proposing replies and flags.
8. Person (cohort): reassesses before the agent drafts live replies.

What does the test usually find in universities?

Documents of mixed provenance. A transcript received directly from the awarding institution or a verified issuer is signed. A scanned certificate uploaded by the applicant is external. The applicant's own list of grades is declared. Admissions staff know fraud exists, but under peak pressure many treat a clean scan as proof. The test shows whether they read the agent's provenance stamps and know which documents it will hold.

Academic judgement. Professional services staff sometimes expect the agent to apply entry requirements to every case. Many applications fall into grey areas that belong to an academic tutor. The test checks whether staff know a held application in a grey area should go to a tutor, not be approved on a best guess.

Younger applicants and data. Some applicants are under 18, and parents often contact the university on their behalf. A parent's request is declared, and the university's records cannot be shared with them without the right basis. Staff should know the agent will hold such requests.

What does the report change?

It picks the first agent. A team weak on document provenance should start with routine enquiries, not completeness checks on uploaded documents.

It sets the calendar. A weak result shortly before the peak means the agent runs at depth 1 or 2 through the peak, and depth 3 waits for the quieter months.

It clarifies the line with academics. Where the report shows confusion about academic judgement, the never-decide list is refined and built into the agent's bounds.

What evidence should the university keep?

The never-decide list agreed with academic staff.

The report by concept.

The first-agent decision and its depth.

The reassessment before live replies.

What should the director of admissions measure?

• Scores by concept, comparing admissions and registry teams.
• During shadow, applications the agent flagged for a tutor that staff had decided themselves.
• After go-live, enquiries escalated because of a refusal, and whether the refusal was right.

Where does it go wrong in universities?

Running it during the peak. Nobody has time to train, and the results reflect fatigue.

Leaving academics out of scoping. Without their list of decisions, the agent's scope creeps into academic judgement.

Assuming students are the main risk. Most errors come from staff approving held items too quickly at peak, not from students.

Smaller, faster-moving organisations run the same assessment differently; see the fintech version. High-volume enquiry handling is covered in the contact centre version. The readiness assessment report template gives a format, and the topic guide is training teams for AI agents.

Where the Academy fits

Pryme Intelligence Academy offers a free readiness assessment for teams of up to 25, using the Foundations test set in the team's own industry. The Foundations track is open to anyone, students included, and its badge is free with no expiry, which suits universities that want staff and students to share the same basic vocabulary.