
AI Agent for Dance Schools: Class Schedules, Enrollment and Limits
How an AI agent for dance schools answers questions on class times and enrollment, and which class changes need approval from the coordinator.
Want to apply this in your company?
Book a call with Juan →In a dance school, the AI agent worth running today is one with clear limits. It answers questions about class schedules, levels, enrollment documents and performance dates based on what the school has published, and it hands off to the coordinator any request that changes a class, a spot or a rule. Separating lookup from decision is what makes this automation safe for the front desk, for families and for management. In practice, that line is drawn before setup, not after the first mistake.
Key takeaways
- Checking a schedule is a lookup: the agent replies with the current timetable without needing someone at the front desk.
- Changing a class, making up a missed lesson outside the rules and granting an age exception are decisions for the coordinator, not the agent.
- Enrollment should follow a list of documents and steps approved by the school, collecting only what is necessary.
- Performances create requests for costumes, rehearsals and image consent, and each needs a defined channel and deadline.
- Start with a small flow, log every handoff and expand only after reviewing what happened.
Why separate lookup from decision
A dance school receives many repeated messages in the same period: what time the children's class is, whether there is a Wednesday evening class, when the dress rehearsal starts, which document to bring for enrollment. These questions have fixed answers, and answering them quickly is what the agent does well. The problem starts when the same conversation turns into a request to switch classes, an extra spot in a full room or the release of a make-up lesson that is not in the policy. In those cases, whoever responds needs to know the student, the class, the teacher and the room availability. That is the coordinator's judgment.
When this separation is missing, two problems usually appear. The first is that the agent promises a change that has not been approved, which creates frustration and a second round of messages. The second is that the team loses context, because the family repeated everything to the agent, then to the front desk, then to the coordinator. The practical rule is simple: the agent looks up, explains and logs; the person in charge confirms.
Map of lookups and decisions
Before configuring anything, list the questions that arrive today without classifying them yet. Then sort each one into one of the groups below.
- Direct lookup: published schedules, classes offered, address and access, performance dates already announced, the enrollment document list.
- Lookup with verification: open spots in a specific class, a place in a trial class, the status of an enrollment already in progress.
- Coordinator decision: class changes, level changes, make-up lessons outside the policy, age or prerequisite exceptions, any special arrangement.
- Sensitive matters: report of an injury, complaint about a teacher, request to delete data or question about a child's image.
A single topic can fall into more than one category. For example, 'Do you have a spot in the 6 pm class?' is a lookup with verification, while 'Can you move my son to the 5 pm class?' is already a decision. The agent needs to recognize that difference, and the school needs to write it down in a rule that every staff member can read.
Classes and changes: how the agent hands off
Illustrative scenario: on a Tuesday evening, a parent asks whether her daughter can move from the Thursday 4 pm class to the Friday 4 pm class, because the family's routine has changed. The agent checks the schedule, confirms that the Friday class exists and that the student's age group fits it, and explains what happens next. It does not confirm the switch. It records the request with the student's name, current class, requested class, the reason given and the date of the request, and tells the family that the coordinator will reply within the timeframe the school has set.
The next morning, the coordinator sees the request in a queue, checks how full the Friday class is and how the group works, and responds. If she approves, the office records the change. If she declines, the reason is noted and the family receives a standard reply, without the agent having improvised anything. The benefit is not promising everything on the spot. It is never losing the request and making clear who decides.
This design also stops the agent from inventing availability. If the schedule shows no open spot, the reply is 'I can't confirm that right now, I'll pass it to the coordinator,' not a 'yes' based on guesswork.
Enrollment: what to collect and what not to collect
Enrollment is where LGPD, Brazil's data protection law, matters most, because it involves data about children and teenagers, which the law treats with its own rules, and data about guardians. For that reason, the agent's flow should collect only what the school needs to open the record and move the person to the next step. Full identity documents, health information and financial details should not pass through the chat unless the school has a secure way to receive and store them. When those are needed, the agent points to the official channel.
Before going live, the school should answer several questions in writing: who is the legal guardian who starts the enrollment, which documents are required for each class type, how long data is kept and who can access the conversations. These answers become instructions for both the agent and the staff. If the support tool is hosted outside Brazil, it is also worth checking the international data transfer rules under LGPD. For decisions that affect a person, LGPD provides a right to review, and in practice the coordinator is the one who carries out that review.
The agent helps a lot in the interest stage: it explains class types, answers questions about trial classes, reminds people of the document list and books a conversation with the office. Closing the enrollment follows the procedure the school already uses.
Performances: deadlines, costumes and image
Performances generate a surge of messages that builds up close to the date and then disappears. Questions about arrival time, dress rehearsal, costumes, tickets and seating come up again and again. The agent can answer the ones already published, such as date, venue, arrival time and the list of items the student must bring. Tickets, complimentary passes and special arrangements depend on the school and should not be promised by the agent.
Image consent needs its own care. If the school asks for permission to film or photograph students, the text has to be clear about its purpose and the channel where it will be shared, and the agent only tells families where to find the form and how to send questions. When someone asks for a photo to be removed from a publication, the request goes to the coordinator with a defined response deadline.
A short checklist helps the team:
- Publish the date, venue, arrival time and item list far enough ahead for families to plan.
- Define who receives costume requests after the deadline and who approves exceptions.
- Make the image consent form available through an official link, with its purpose described.
- Create a channel for image removal requests and log each one with the date it was received.
- After the event, review the questions that went unanswered and update the published content.
How to put it into operation
The method we use starts by mapping the current process and its baseline: how many questions arrive, through which channel, at what hours and how long they take to answer today. Next, we limit data and actions. The agent gets read access to the schedule, not to payroll, and it does not change anything in the enrollment system without human approval. The pilot runs one real flow, for example schedule lookups and change requests, with logs and an agreed measure defined before it starts. Finally, we hand the operation to the team and decide, based on the pilot, whether the school expands, adjusts or stops.
A simple plan for the first phase can follow this order:
- Pull the most frequent questions from the last few weeks and classify them as direct lookup, verification or decision.
- Write the handoff rule for each decision, with an owner and a response deadline.
- Configure the agent only with the published schedule, the performance calendar and the document list.
- Test with real questions in an environment without real students before opening it to the public.
- Open it to one channel or a limited period, log everything and review incorrect answers at the end of each day.
To understand what an agent does today without overpromising, read what an AI agent does today. The article on pilates studio scheduling with make-up class rules follows a similar logic and is also worth a look.
What to log and when to stop
An agent that leaves no trail is hard to correct. For each conversation, record the question, the category assigned (lookup, verification or decision), the reply given, the handoff made and the outcome confirmed by the coordinator. This record also shows where the agent got it wrong and where the rule was ambiguous. When the school decides to expand, use this data to adjust the instructions, not to widen permissions in the dark.
Before the pilot, define the signals that mean you should stop: an invented answer about a spot, sensitive data collected outside the flow, a change approved without the coordinator, recurring complaints about tone. If any of these appear, the path is to pause, fix the rule and return to a smaller pilot. Stopping is also a valid result.
Frequently asked questions
Can an AI agent confirm a student's class change? It should not. In most schools, a change depends on room availability, the student's age group and level, and the coordinator is the one who assesses that. The agent can explain the request, log it with the necessary details and state the agreed response deadline.
How does the school keep the agent from giving out schedules that have already changed? The published schedule must be the agent's only source, with a visible update date for the team. When the coordinator changes a time, the change has to be made in the source before any communication goes out. If the information is not in the source, the agent should say it cannot confirm and hand the question off.
What data can an enrollment agent collect? Only what is needed to start the record and move to the next step, with the legal guardian identified when the student is a minor. Full identity documents, health data and financial information should go through an official school channel. LGPD gives special attention to data about children and teenagers, so confirm the policy with the legal team or the data protection officer before configuring the flow.
Where should a school start if it has never used AI agents? Start with schedule lookups and the document list, which have fixed answers and low risk. Keep decisions with the coordinator, log everything and evaluate the pilot against the agreed measure before adding another topic. A first flow that is small and well reviewed teaches more than a large project without rules.