AI for Language Schools: Placement Tests, Class Changes, Renewals
How a language school can use AI agents for placement test follow-up, class change requests and renewal reminders, with clear rules for level decisions.

AI for a language school works best when it handles the repetitive work around placement tests, class changes and renewals, and leaves level decisions with the people who own the curriculum. An agent can collect test results, apply the school's written rules, draft replies and route exceptions to a coordinator or teacher. The useful question is not whether the agent can talk to students, but which decisions it is allowed to make on its own.
Key takeaways
- Placement follow-up loses students in the gap between the test and the first class, so start there.
- Write the level rules as a short, dated document before connecting any tool to scores or timetables.
- Let the agent recommend, draft and route, and let a named teacher or coordinator approve every level change.
- Treat class change requests as policy and capacity checks, not as conversations to improvise.
- Keep renewal reminders factual and tied to the school's own terms, with no promised discounts.
Where language schools lose students after the placement test
Placement tests usually happen during enrollment, when the school is busy and the prospective student is eager. The result often sits in a spreadsheet or an inbox while the student waits for an email that explains their level, the start date and what to do next. Silence at that point reads as disorganization, and some people enroll somewhere else. The same gap appears mid-term, when a student asks to move up a level, switch to an evening group or pause for work travel. Each request arrives in a different channel, and the answer depends on information spread across the timetable, the teacher's notes and the attendance record.
The operational problem is not the test itself. It is the chain of small follow-ups: confirming the result, explaining it in plain language, offering a seat, checking a requested change against capacity, and keeping the record consistent when the student changes their mind. Clinics face a similar shape of work after an assessment, as described in AI for Speech Therapy Practices: Evaluation Scheduling and Session Notes. An agent with memory of that chain, and clear permissions over each step, can keep it moving. Without those boundaries, an agent tends to invent a level or promise a seat that does not exist.
Write the level rules before any agent touches a score
Every level decision should trace back to a rule that a teacher could explain to a student. Before an agent reads test results, the school should write its current rules in one document, dated and owned by a named academic lead. Keep it short enough to read in one sitting. The rules do not need to be perfect, but they must be explicit enough that two staff members looking at the same result would reach the same answer most of the time.
Use the document to sort every result into one of four categories:
- Clear: the score falls inside a published band and the writing or speaking sample agrees. The agent may recommend the level and draft the offer, and a coordinator confirms before enrollment.
- Borderline: the score sits near a boundary, or the sections disagree. The agent flags the case with the evidence attached and does not recommend a level.
- Out of scope: the student reports a learning need, an accommodation request or a prior certificate that the rules do not cover. The agent hands the case to a person and tells the student that someone will reply.
- Disputed: the student contests the result. The agent records the objection, keeps the original test record unchanged and routes the case to the teacher who reviewed it.
The point of this guide is to take judgment out of the agent's job, not out of the school's. If a category cannot be written down, the school does not yet have a rule, and the agent should not be asked to invent one.
A placement follow-up sequence the agent can run
The sequence below fits most schools that use a written test with teacher review. Adjust the timing to your own enrollment calendar.
- Step 1: Pull the result from the test system and confirm that it matches the booking. The agent reads the record and does not edit scores.
- Step 2: Apply the published rule and label the case as clear, borderline, out of scope or disputed.
- Step 3: For clear cases, draft a message that states the level, the group options and the next step. Send it only through a template the coordinator has already approved for this case type.
- Step 4: For borderline and disputed cases, send the coordinator a short summary with the score breakdown, the sample reference and the one question that needs a decision.
- Step 5: Record the outcome: level confirmed, seat offered, student replied, teacher notified. A missing outcome triggers a reminder to staff, not a second message to the student.
- Step 6: Review the borderline cases once a week. If the same boundary keeps producing disagreement, the rule needs revision, and that decision belongs to the academic lead.
Illustrative scenario, not a real school: a student completes a placement test and scores just under the cut for an intermediate group, but the writing sample reads at the upper edge of that group. The agent marks the case as borderline, attaches both pieces of evidence and tells the student that a teacher will confirm the group within the school's usual reply window. The coordinator asks the teacher to read the sample, the teacher confirms the intermediate group, and the student's record notes the reason for the decision. Trade schools face a similar question when admissions answers depend on cohort placement, as shown in AI for Trade Schools: Admissions Questions, Cohorts and Placement.
Class change requests: check the policy, then the seat
Students ask to change classes for ordinary reasons: a new work shift, a family schedule, a preferred teacher, or a group that feels too fast or too slow. Each request can be run through three checks in order. First, policy: the change window, the notice period and any fee rules, which the agent quotes from the published terms and does not interpret. Second, level fit: whether the target class matches the student's current level under the rules described above. Third, capacity: whether a seat exists and whether moving the student leaves the original group workable.
When all three checks pass and the school has approved an automatic path for that type of change, the agent can confirm the move. Otherwise it prepares the change for a coordinator, with the three checks summarized in a few lines. When a request crosses a boundary, such as a jump of more than one level, the agent should explain the reason and route the request rather than look for a workaround. Waitlists and capacity follow the same logic in other membership businesses, as in AI for Yoga Studios: Memberships, Waitlists and Clear Limits.
Permissions matter here. The agent should be able to change enrollments only for the actions the school has approved, not for every record in the student system. OWASP's authorization guidance describes this kind of narrow, explicit access for software systems in general, and it applies directly to a school's enrollment data.
Renewal reminders that stay factual
Renewals are where a reminder can drift into a sales message. A useful renewal flow tells the student or guardian when the current term ends, which next terms are open, what the school's renewal terms say and how to reply. It does not promise a discount, a waived fee or a reserved seat unless staff have approved that offer in writing for this case. The agent also needs to know when to stop. If a student has paused, moved away or asked not to be contacted, that status must be in the record before the next message goes out.
- The reminder states the term end and the next start in the school's own words.
- Only approved offers appear, and each one has a named owner and an end point.
- Students who asked to pause or to stop contact are excluded before anything is sent.
- Replies that mention payment problems, complaints or cancellation go to a person the same day.
- A monthly review shows which reminders generated confusion, so the wording can be corrected.
How Be Human runs this as a managed service
Be Human starts with one useful role, here most likely the placement and renewal follow-up coordinator. The first engagement defines the memory that role needs, such as the level rules, the term calendar and the approved message templates. It also defines the tools the agent may use, such as read access to test results, a view of the timetable and the messaging channel, along with the approvals each action requires and one observable success criterion. A good criterion is concrete, for example that every placement case ends with a recorded outcome before the next term begins.
After launch, managed operations cover monitoring of failed or stalled actions, recurring evaluations against a fixed set of past cases from the school's own records, permission reviews whenever staff roles or terms change, incident response when a message goes out in error, and controlled updates to rules and templates that the academic lead signs off. This follows the human oversight and risk management practices in the [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework). The service is delivered remotely, in English or Portuguese. If you want to find the first task worth handing to an agent at your school, take the free Be Human AI diagnostic.
Frequently asked questions
Can an AI agent decide a student's language level? It should not make the final call on its own. The agent can apply written rules, recommend a level for clear cases and flag borderline results with the evidence attached. A named teacher or coordinator confirms the level, which keeps accountability where the school wants it.
Which parts of a language school should an agent handle first? Start with placement follow-up, because the work is repetitive, time sensitive and easy to measure. Class change questions that follow published policy come next. Renewal reminders are most useful once the school's terms are clean and the approved offers are listed in one place.
Does the agent need access to our student management system? It needs only the access required for approved tasks, such as reading results and updating the status of a message. Write access to enrollments should be narrow and reviewed regularly, because staff roles and school terms change. Broad administrator access is not necessary to get a useful first version running.
How do we know the agent is doing the job well? Define the success criterion before launch, such as whether every placement case ends with a recorded outcome before the next term. Test the agent against a fixed set of past cases from your own records and repeat that check on a schedule. Review failures with the academic lead, and change rules only through the same approval step you use for the rules themselves.
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.


