Superintelligent Agents for Language Schools: Placement and Attendance

A language school can put AI agents to work now on the repetitive parts of placement and attendance: answering questions about courses and schedules, collecting enrollment documents, sending attendance reminders and flagging learners who need a person to call. The agent should not decide which level a student belongs in. That decision stays with a named teacher, who reviews the evidence the agent gathers and signs off. The goal is fewer hours spent on the same questions and faster follow-up on absences, without handing over judgment that depends on a conversation with the learner.

Key takeaways

• Start with routine work: placement questions, enrollment paperwork, schedule confirmations and absence follow-up have clear inputs and checkable outcomes.
• Keep level decisions with teachers. The agent gathers evidence and drafts a proposal, and a named person approves or changes it.
• Write down what the agent may read and do: the catalog, the calendar, approved policies and only the student fields a task needs.
• Measure before you expand. Track how often teachers change the agent's proposals and how quickly absences reach a human.
• Review permissions on a schedule, so access that made sense in one term does not quietly outlive the role it was built for.

Why placement and attendance come first

Language schools lose time to the same few questions. A returning learner asks which class fits, a new student asks whether they can join mid-term, a prospective learner wants to know what documents enrollment requires, and a parent asks why a student was marked absent. The answers already exist in the school's catalog, calendar and written policies. That makes them a good fit for an agent with narrow memory and a short list of tools.

Attendance has a similar profile. A missed class is an early signal, but it often goes unnoticed until a learner has quietly stopped coming. A consistent follow-up routine is repetitive work, and busy staff tend to postpone it. An agent can run that routine every day while the coordinator focuses on the cases that need a conversation.

The phrase superintelligent agents describes where the tooling is heading. It does not change the first question an owner should ask, which is what the agent is allowed to do on a Monday morning. Begin with tasks where the right answer can be checked against a document or a record.

Which placement questions an agent can answer now

Placement involves two different jobs. Answering questions about placement is routine work. Deciding a learner's level is professional judgment. An agent can handle the first category well when it has access to approved material, in four areas:

• Catalog questions: which courses exist, their sequence and their stated prerequisites.
• Transfer questions: how the school treats prior study at another institution, as stated in its written policy.
• Logistics: term start dates, open seats, placement booking windows and the room or online link for each section.
• Preparation: what a placement conversation covers and what a learner should bring to it.

Each area needs a defined source. If a question depends on an exception, the agent should say that a coordinator will confirm and create a case for that person, rather than improvising an answer. Schools with technical or professional cohorts can apply the same split between cohort questions and individual evaluation, as described in AI for Trade Schools: Admissions Questions, Cohorts and Placement.

Enrollment steps as a checklist

Enrollment is a sequence with clear checkpoints. When setting up the workflow, walk through these steps in order and name the person who owns each exception:

• Capture the inquiry: name, target language, goal, preferred schedule and the channel the learner wants to use.
• Offer the placement path: a short self-assessment form or a booked conversation with a teacher.
• Write the placement evidence to the student file: test results, teacher notes and the level the agent proposes, labeled clearly as a proposal.
• Route the proposal to a named teacher. The agent cannot mark a level as final.
• After approval, send the enrollment packet and the list of required documents, then track which items have come back.
• Check billing status through the school's system with read access only. Any change to amounts, discounts or refunds goes to staff.
• Confirm the first session with its date, room or link and teacher, then schedule the first attendance check.

The steps look simple on paper, and most of the value comes from keeping them in this order. A proposal that arrives before the evidence is filed, or a confirmation sent before billing is checked, creates the kind of error that is expensive to correct with a learner.

Attendance follow-up that learners experience as respectful

Attendance follow-up works best when the rule is written before the first message is sent. Staff define what counts as a pattern for their program, for example a missed session followed by no reply within a set window. The agent reads attendance records, sends a message drafted from approved templates, logs the contact and stops as soon as the learner replies. Messages should be short, written in the learner's preferred language and offer a concrete next step, such as a recording of the missed session or a make-up option.

Illustrative scenario: a learner in an evening conversation course misses two sessions in a row. The agent sends a message in Portuguese offering the recording of the last session and asking whether the schedule still works. The learner replies that a new work shift starts next week. The agent records the reply, notifies the coordinator and pauses automated reminders until the coordinator writes back.

The agent should not give academic or financial advice in these threads. Learners should know what the school records about attendance and why. Schools operating in Brazil should check their notices and retention practices against the guidance of the ANPD, the national data protection authority, before connecting attendance records to any agent.

Where teachers stay in the loop

Level decisions depend on speaking, listening, goals and context that a form does not capture. Placement tests are imperfect, and learners sometimes understate or overstate their experience. As agents become more capable, they will summarize more evidence and draft more polished proposals. That makes the review step more important, because a fluent draft can look more certain than the evidence behind it. Keep the approval step explicit, record who approved each level and review the cases where teachers changed the agent's proposal.

A simple decision guide for the school:

• The agent may answer directly when the answer appears in an approved document and no judgment about the learner is involved.
• The agent may draft a placement proposal when teacher-scored evidence is already in the file.
• Only a named teacher may confirm, change or reverse a level.
• Refunds, exceptions, complaints and changes to contracts go to staff. The agent only creates the case.
• Any new tool connection or permission change is reviewed by the owner before it goes live.

The OWASP authorization guidance is a useful reference for writing these boundaries as explicit permissions rather than assumptions. NIST's AI Risk Management Framework offers a structure for deciding where a person must review an AI output before it affects a learner. Clinical practices with scheduled evaluations face similar handoffs, as shown in AI for Speech Therapy Practices: Evaluation Scheduling and Session Notes.

How Be Human runs this as a managed service

Be Human begins with one defined role, usually an enrollment and attendance assistant for a language school. The first engagement defines the memory the agent may read, the tools it may call, the approval boundaries for level decisions and a success criterion the school can observe, such as whether teachers accept most proposals with only minor edits. Managed operations then cover monitoring, recurring evaluations, permission review, integration maintenance and incident response. The service is delivered remotely, in English and Portuguese.

Evaluation is where most schools need discipline. Build a test set from real questions asked during the last term, with answers approved by staff, and rerun it whenever the catalog or a policy changes. The OpenAI evals guide and Anthropic's writing on building effective agents both recommend testing agents against realistic tasks, not only demonstrations. Track teacher overrides, unanswered inquiries and how often a learner needs a human the same day. Review those results with the owner on a schedule, and pause any capability whose results drift.

The free Be Human AI diagnostic can help you find the first placement or attendance task worth handing to an agent.

Frequently asked questions

Can an AI agent decide which level a student should join? No, and it should not be set up that way. The agent can gather placement evidence, summarize it and propose a level, but a named teacher confirms or changes the decision and the record shows who approved it. This keeps accountability with the school and gives learners a person to speak with when they disagree.

Which tasks should a language school automate first? Start with questions that have a written answer and records that already exist, such as course sequences, term dates, enrollment document lists and attendance reminders. These tasks have clear inputs and outputs, so staff can check the agent's work quickly. Leave anything that involves judgment about a learner until the first workflow has been reviewed and trusted.

What access does the agent need to work safely? It needs read access to the catalog, calendar, written policies and only the student fields required for its task, along with a defined list of actions such as drafting a message or creating a staff case. Write those boundaries down and review them on a schedule. Permissions should be narrower than what the school's software technically allows.

Does attendance follow-up need to be automated? Not necessarily, but automation helps when staff have many learners and missed sessions go unnoticed for too long. The school sets the pattern that triggers a message, the message is short and in the learner's language, and any reply goes to a person. Learners should know what is recorded and why, and the school should check its data practices against local rules, including ANPD guidance in Brazil.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.