Superintelligence Prep for Private School Admissions and Parent Requests
How a private school can use AI for admissions inquiries, tour scheduling and parent request routing, with student records kept under staff control.

An AI agent for a private school's admissions office answers routine inquiries at any hour, captures the facts staff need to act, offers tour times from a calendar the office controls, and passes every decision that touches a family, a payment or a child's record to a named person. The agent works from an approved set of answers and keeps its memory of each inquiry separate from the official student file. Schools that start with those boundaries get a useful front desk without handing judgment to software.
Key takeaways
- Limit the agent to inquiry intake, tour booking and request routing. Admission decisions, fee exceptions and safeguarding matters stay with staff.
- Capture the same intake fields every time so the admissions team starts from facts, not from a long email thread.
- Offer only tour slots that staff have published and marked open, and release unconfirmed holds on a fixed schedule.
- Route each category of parent request to a named human owner with a backup and a written response window.
- Keep the agent's inquiry memory apart from the student record, and give it only the access its role requires.
What an admissions agent should handle, and what it should not
Most messages that reach an admissions office repeat the same questions: which grade levels are offered, whether there is before-care, how to request a visit, and which documents a typical application includes. An agent can answer these well if the answers come from a short library that the head of admissions or registrar has reviewed. That library is the source of truth. The agent should not improvise policy from a general web search or from memory of last year's handbook.
The boundary matters as much as the answers. Admission decisions, scholarship offers, fee adjustments, exceptions to policy, learning assessments, safeguarding concerns and complaints about staff are outside the agent's role. The agent can acknowledge these, collect the relevant facts, and hand them to a person. It should never tell a family that a place is likely, that a fee can be waived, or that a concern has already been handled.
The same pattern of admissions questions, start dates and placement follow-ups appears in vocational training, which is covered in AI for Trade Schools: Admissions Questions, Cohorts and Placement. The lesson carries over: scope the agent to what the office has already decided to publish.
How should inquiry intake work?
A good intake captures enough to act on and nothing more. Staff should be able to read the summary in under a minute. A workable set of fields looks like this:
- Guardian name and the contact method the family prefers
- The first name of the child and the grade the family is considering
- The preferred start term, described in the school's own terms
- Whether the family wants a tour, an application packet or general information
- How the family heard about the school
- The exact wording of any question the agent could not answer
Keep the form short. Do not ask for dates of birth, health information or household income during intake. Those details belong in the application system that staff control. Collecting them in a chat thread creates a second copy that someone later has to find and delete. Organizations that must limit intake data face the same discipline, which is explored in AI for Behavioral Health Clinics: Intake and Scheduling Limits. After the intake is complete, the agent reads the summary back to the family and tells them when a person will reply, using the response window the office has approved.
Tour scheduling without double-booking
Tour scheduling fails in predictable ways: two families get the same guide, a room is promised twice, or a visit with an accessibility need is booked into a route that cannot accommodate it. The fix is to make the admissions calendar the only source of truth and to give the agent a narrow set of steps.
- Staff publish tour slots with the guide, room and capacity set in advance.
- The agent offers only slots marked open and never proposes a time that is not on the calendar.
- When a family picks a slot, the agent places a short hold with a clear confirmation deadline.
- The confirmation uses approved text for arrival instructions, parking and the check-in contact.
- Unconfirmed holds are released automatically, and the release is noted in the inquiry record.
- After the visit, the outcome (attended, rescheduled or no-show) is recorded for the admissions lead.
Two details cause most scheduling errors. The first is group size: siblings and extra adults change the room and the guide's time. The second is access needs, such as a step-free route or an interpreter. Put both into the booking fields so the front office knows before the family arrives. Rescheduling should follow the same path as booking so the calendar stays accurate.
Illustrative scenario: a family writes on a Sunday evening asking for a visit for a child entering fourth grade, along with an older sibling, and mentions a wheelchair user in the group. The agent offers two open slots with a step-free route and an available guide, holds the one the family selects, and sends the approved confirmation. It flags the access note and group size for the admissions coordinator, who calls on Monday to confirm the route and the room.
Routing parent requests to the right owner
Parent requests vary widely in urgency and in who must act on them. A routing table makes those differences explicit. A simple version looks like this:
- General information from the approved library: the agent answers and logs the question.
- Tour or visit request: the agent books from open slots and confirms.
- Records request, such as enrollment verification: the agent logs the request and routes it to the registrar, and does not send documents itself.
- Schedule change, late pickup or transport question: the agent routes to the front office.
- Fee, scholarship or payment question: the agent acknowledges the question and routes it to the business office.
- Admission decision or appeal: the agent routes to the head of admissions and does not predict an outcome.
- Health concern, safeguarding issue or complaint about staff: the agent stops automated replies, tells the family that a named person will contact them, and alerts the designated staff member at once.
Each category needs a human owner and a named backup. Write down the response window for each category and check it monthly. An agent that routes correctly into a single overloaded inbox has only moved the bottleneck. Hospitality businesses with similar inquiry volume often find that routing rules matter more than reply speed, as described in AI for Bed and Breakfasts: Inquiry Replies and Guest Memory. Review the routing table each term, because staff roles, calendars and reporting lines change.
Keeping student records under staff control
The admissions agent should never be the system of record for a student. Its memory is the inquiry: what the family asked, what was answered, which slot they picked, and who owns the next step. The official file lives in the student information system and the application platform, where staff decide who can read and change each field. Scope the agent's access to those systems by role, using the same permission model the school already applies to front-office staff. A shared administrator login is the wrong design, because it removes the record of who did what.
Inbound messages are untrusted input. A parent's email can contain text that tries to change the agent's behavior, such as a request to reveal another family's details or to skip the routing rules. Treat text from outside the school as data, not as instructions. Keep any action that changes a record behind the same approval rules that apply to human staff, and limit what the agent may return to a family to approved fields. The OWASP guidance on prompt injection describes these controls in detail, and the OWASP authorization guidance covers how to check permissions on every request rather than once at login.
Logs make the system answerable. Record each tool action the agent takes, the staff member who approved any change, and the time it happened, so that questions about who was told what can be answered from the record. Retention and deletion of inquiry data should follow the school's own policy and the data protection rules that apply where the school operates. Staff should confirm those rules with their own counsel. The NIST AI Risk Management Framework is a practical reference for mapping these risks to owners and controls.
Running the agent after launch
Launch is the easy part. The first term shows where the approved library is out of date, which routing category receives more traffic than expected, and which requests families make that the agent should never have been asked to handle. A short review loop catches these early: sample conversations each week during the first term, check handoffs against the routing table, and update the approved answers whenever the school changes a policy.
Be Human designs and operates this kind of setup as a managed service. A first engagement defines the one role the admissions office needs, the memory and tools that role requires, the approval boundaries for each request category, and one observable success criterion, such as how many tour requests reach a confirmed slot without staff retyping details. Ongoing operations can include monitoring, recurring evaluations, integration maintenance, permission reviews, incident response and controlled changes to the answer library. The service is delivered remotely in English and Portuguese.
If you are not yet sure which admissions task should be handed to an agent first, the free Be Human AI diagnostic can help you find the first task worth handing to an agent.
Frequently asked questions
Can an AI agent answer questions about tuition? Fee questions should go to the business office rather than being answered by the agent. The agent can confirm the question, explain that the office will send the current fee information, and record the family's contact details. Quoting figures from memory risks stale or incorrect information, so the approved library should contain only text the business office has cleared for publication.
Should the agent help decide who gets admitted? No. Admission decisions depend on criteria, judgment and records that staff are accountable for. The agent can explain the published steps, list which documents have been received, and tell the family who will contact them and when. It should never predict an outcome or imply that an application is favored.
How long does a first engagement take? The length depends on how many request categories the school handles, how current its calendar and answer library are, and how many systems the agent must read from. The first step is a scoping conversation that defines one role, its boundaries and its success criterion. Only after that does the work on memory, tools and routing begin.
Does the school have to move its student records into a new system? No. The agent works with the systems the school already uses, with access scoped to its role. A first review usually looks at where inquiry data currently lands, such as a shared inbox or a spreadsheet, and decides which copy becomes the single place for it. The student record itself stays where staff already manage it.
Your next step
See where an AI agent fits in your company first.
Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.
Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.


