Copy the plan below and fill in one row per field. It puts every student services adviser through a shared foundation, certifies the advisers who decide the agent's held replies, schedules the two-week practicum away from enrolment and results peaks, names who signs off each person, and sets renewal before the next academic cycle begins.
What is this plan for, and when should you use it?
A student services agent usually does three kinds of work. It answers routine questions about deadlines, fees, ID cards and timetables at depth 1 or 2. It drafts replies on extensions and mitigating circumstances for an adviser to check and send, which is depth 2. Sometimes it books appointments at depth 3, inside the limits the university sets.
What it must never do alone is handle a welfare disclosure. A message that hints at risk to a student's safety goes to a named adviser at once, whatever the agent's depth. That rule shapes the whole plan: the people who review the agent's drafts must recognise those messages fast and know the route.
Use this plan before the agent's first draft reaches a student, and again whenever the agent gains a capability or a higher depth. The wider reasoning on who to train first is in the guide to training teams for AI agents.
How do you fill in the plan?
Start with the decisions, not the people. List every point where the agent stops and waits for someone: a held reply, a refusal a student challenges, a booking outside the agent's bounds. Then name who handles each one.
Next, give each person a track. Everyone who reads the agent's work needs the shared vocabulary of depth, verdicts and the record. Advisers who approve held replies need a certified reviewer standard, earned through an exam and assessed real decisions. One person, often a digital or systems officer, needs to know how the agent's brief and sources are built.
Then fix dates around the academic calendar. The practicum needs a steady flow of real held cases, but not the flood of enrolment week, when nobody has time to be marked. Build in calibration, too: running calibration sessions for student services teams keeps advisers deciding alike after sign-off. For the method of teaching review itself, see how to train the student services team to review AI agent decisions.
The training plan template
| Field | What to enter |
|---|---|
| 1. Team and service | Name of the team, campuses or channels covered |
| 2. Agent capabilities | Each capability with its depth (for example, answers enquiries: depth 2) |
| 3. Red lines | Work the agent never completes, such as welfare disclosures or outcome decisions |
| 4. People and decisions | Each person and the held actions they will decide |
| 5. Track per person | Shared foundation, certified reviewer, builder or architect |
| 6. Foundation complete by | Date, before any reviewer training starts |
| 7. Reviewer modules window | Start and end dates |
| 8. Written exam date | One date per person, with a fallback |
| 9. Practicum window | Two weeks of real held decisions, outside peak periods |
| 10. Practicum case mix | The case types each reviewer must see, including at least one disclosure route |
| 11. Assessor and sign-off | Who signs off each reviewer, and the date |
| 12. Approval rights granted | Which held actions each person may approve once certified |
| 13. Calibration rhythm | How often, who runs it, where results are kept |
| 14. Renewal date | Twelve months after issue, placed before the next peak |
| 15. Retraining triggers | New capability, new depth, change of role, a missed red line |
What does a filled-in plan look like?
Take an illustrative student services hub with 12 advisers, two team leads and one digital officer. The agent answers enquiries at depth 2 and books appointments at depth 3. Red lines: welfare disclosures, outcomes on mitigating circumstances, and any fee refund.
- Fields 4 and 5: all 15 people take the foundation; the 12 advisers and two leads take the reviewer track; the digital officer takes the builder track.
- Fields 6 to 8: foundation complete in weeks 1 and 2 of a quiet teaching period; reviewer modules in weeks 2 to 5; exam in week 6.
- Fields 9 and 10: practicum in weeks 7 and 8, covering extension requests, booking clashes, a fee question the agent refused, and a disclosure routed to a person.
- Fields 11 and 12: each adviser signed off by an assessor; certified advisers approve held replies, leads approve bookings outside bounds.
- Fields 13 to 15: calibration every four weeks run by a lead; renewal set two months before the next enrolment; retraining whenever the agent gains a capability.
The full timeline runs to about eight weeks, which is in line with the estimate in how long it takes to train student services teams to supervise AI agents.
Where the Academy fits
Pryme Intelligence Academy maps onto fields 5 to 14. Foundations is free and takes about four hours. Operator takes about 14 hours plus a two-week practicum, with a 40-question exam passed at 32 of 40 and at least 95 in every 100 practicum decisions correct, no missed red line, signed off by a Pryme assessor. Certificates are valid 12 months and are checked at source through VerifyDoc.ai.
Questions people ask
Does every adviser need the full reviewer certificate?
Only those who approve the agent's held replies. An adviser who only reads the agent's answers needs the shared foundation. If rotas mean anyone might pick up the held queue, certify everyone on the rota.
What if the practicum would overlap results or clearing?
Move it. A reviewer marked during a surge is marked on speed, not judgement. Run the agent at depth 1 or 2 through the peak and practise on its drafts afterwards.
How does this differ from a plan for patient services?
The structure is the same, but the red lines differ. The training plan template for patient services teams centres on clinical questions; this one centres on welfare disclosures and academic outcomes.