AI student assistant for universities and colleges: eligibility decided by a rule table your registrar edits
How a university configures Kav: repeat semester questions answered from registrar data, eligibility decided by an editable rule table, plus a cost worksheet.
Student questions repeat every semester: scholarships, registration, schedules, discounts. The answers are policy, and policy is exactly where a friendly guess does damage.
A depiction, not a customer. No university is named here. The example is the platform's own illustration, marked example data on the home page.
What a student asks, and what Kav answers
- "Am I eligible for the sibling discount?" Eligibility is a named question answered by a rule table your registrar edits directly. The table is compiled and tested before it is ever published. The answer comes back as a decision with its reason attached, in the example "eligible, based on two enrolled siblings". The model did not decide it and did not write it. Apply it and Ask a person sit beside it (a decision, not a guess).
- Published answers that stay current. A change on a policy page becomes a draft; one named person submits it and a different one approves it before the assistant can read it back (the two-person rule).
- A person when it is needed. An Agent Desk takes over anything that is not a lookup.
Where the money moves
The bold rows are the levers a university leans on most.
| Lever | What Kav does here | What you measure |
|---|---|---|
| Fewer repeat contacts | Semester-cycle questions are answered from the registrar's own data and approved content. | Contacts in registration and results weeks against the rest of the year. |
| Faster when a person steps in | Ask a person hands over with the question and the rule outcome attached. | Handling time on escalations. |
| Out of hours, every channel | Students ask at night and between lectures; the answer is there on the channel they use. | Share of contacts arriving outside office hours. |
| Fewer errors and rework | Eligibility is one rule everyone gets, with its reason, instead of ten agents' readings of a policy. | Appeals and re-decisions that overturn a first answer. |
| Revenue kept | Apply it sits next to an eligible answer, so a discount that applies is claimed rather than missed. | Applications started from a chat. |
| Compliance and audit | Rule changes are compiled and tested before they publish; content is checked by two people. | Effort to show who changed a rule, and when. |
Work out your own number
- Count contacts in one peak week and one quiet week; the gap is your seasonal load.
- Cost per contact = staff hours × loaded hourly cost ÷ contacts handled.
- Value = repeat questions answered without a person × cost per contact.
Work out your break-even from the published prices, or talk to us.