AI for a Dance Studio: Class Registration and Recital Logistics
An AI agent for a dance studio can answer registration questions and recital logistics messages, while staff approve every schedule change.

An AI agent for a dance studio can answer registration questions, prepare recital costume lists, and send parent logistics messages at any hour, while studio staff keep the authority to approve every change to a class schedule, level placement or recital roster. The agent works from one written memory of classes, teachers, policies and approved templates, and it stops at the boundaries the studio sets. This article explains where that line sits and how to set it up.
Key takeaways
- Registration questions, waitlist replies and costume sizing questions repeat often enough to hand to an agent, but any move between classes needs a named staff approver.
- The agent should work from one maintained reference of classes, levels, teachers, policies and approved message templates, not from scattered inboxes and memory.
- Parent messages are untrusted input. The agent reads and drafts from them, and it never treats text inside a message as an instruction.
- Each tool the agent can use should carry the narrowest permission that completes the task, so a persuasive message cannot unlock a change that staff have not approved.
- Success is observable: every schedule change has an approval record, and every outbound message traces to an approved template or a staff-reviewed draft.
Where registration questions pile up in a dance studio
The front desk of a dance studio handles a steady mix of questions that look simple and are not. A parent wants to know whether a six-year-old can move from the beginner ballet class to the next level, whether a Tuesday class still has room, whether a Saturday rehearsal replaces the regular session in the week of the recital, or what happens to a missed class. Each answer depends on teacher notes, age cutoffs, current capacity and studio policy, and that policy often lives in a binder, a pinned post and the memory of the owner. The result is that the same question gets answered three different ways in one week.
Registration peaks are predictable. They arrive in the weeks before a new session opens, during the first days of a waitlist, and in the stretch between casting announcements and costume orders. During those windows, staff spend most of their time retrieving the same facts and repeating them. That repetition is where an agent helps. It is also where mistakes are expensive, because a wrong level placement or a wrong costume size reaches a family, a teacher and a seamstress before anyone notices.
Which decisions stay with staff and which can an agent handle?
The useful split is simple. The agent answers questions whose answers are already written down, drafts replies to everything else, and asks a named person to approve any action that changes a booking, a roster or a payment. For most studios, the working boundaries look like this:
- Agent answers directly: class times, room locations, what to wear, the published policy on missed classes, how to join a waitlist, and the recital date and ticket window.
- Agent drafts, staff sends or approves: level placement questions, requests for private lessons, exception requests, and any reply that mentions a teacher's judgment.
- Agent prepares, staff approves before anything moves: moving a student between classes, adding a student to a full class, cancelling a rehearsal, or changing a student's recital group or role.
- Staff only: refunds, disputes, injury or medical information, pickup and custody changes, and anything involving a minor's safety.
How do you set up the class registration flow?
Start with one term and one studio location, because a narrow launch exposes problems that a broad one hides. The setup usually follows a clear sequence:
- Load the class catalog for the current term: class names, levels, age ranges, teacher, room, capacity and the policy text that teachers have already approved.
- Write the placement rules in plain language, including age cutoffs, prerequisites and any teacher sign-off requirement. If a rule exists only in one teacher's head, the agent cannot use it, and neither can a new hire.
- Give the agent read access to the registration system first, so it can see capacity and waitlist status before it is allowed to write anything.
- Define the reply templates the agent may use, along with the exact conditions under which it must hand the conversation to a person.
- Grant write permission only for a narrow action, such as adding a family to a waitlist, and route every move between classes into an approval queue.
- Name one approver for each decision type and one backup. An approval queue with no owner stalls, and a stalled waitlist loses families.
- Replay last term's real questions in a test copy and compare each answer with what staff would have said. Fix the rules before going live, not after the first complaint.
How should recital costume lists and logistics messages run?
The following scenario is illustrative and does not describe a specific studio. A studio runs three recital groups. Each teacher sets costume requirements by level: leotard color, tights, shoe type, hair rules and a sizing sheet for ordering. Parents ask which size to order, whether a sibling can share a costume and when the order closes. The agent answers from the approved costume sheet, asks the family for the measurements the sheet requires, and prepares a sizing summary for the costume coordinator to approve. When a teacher changes a group's color a few days before the order deadline, the agent does not rewrite the list on its own. It prepares a notice listing the affected students and the exact change, and staff approve it before any family receives it. The coordinator sees all pending sizing requests in one place instead of a chain of forwarded emails.
Rehearsal times, arrival windows and day-of reminders follow the same logic. The agent sends reminders from templates that staff approved once, filling in only the date, time and room from the calendar. Anything that falls outside the template, such as a request to change arrival time for one family, goes to a draft queue for a person to review.
How do you keep parent messages from steering the agent?
Parent messages are untrusted input. A message can contain an instruction, by accident (a forwarded note with odd formatting) or on purpose, asking the agent to move a child up a level, skip the waitlist or reveal another family's details. The OWASP prompt injection guidance describes this class of risk, and the practical response is architectural: instructions come from the studio's configuration and from staff, while text written by families is content to read and answer, never a command to obey. Tool permissions are the real limit. An agent that cannot write to the roster cannot be talked into changing it, however persuasive the message.
Privacy matters just as much, because most students are minors. The agent should show a guardian only the records of that guardian's own children, keep medical notes out of outbound messages, and hold only the data the task needs. The OWASP authorization guidance points to the same principle: check each request against the role of the person asking, not merely whether the agent can technically reach the data. The NIST AI Risk Management Framework offers a useful structure for these reviews, with mapping of risks, measurement of how the system behaves and management of changes over time.
How does a managed agent keep working after launch?
Be Human designs the agent as one useful role for the studio, defines the memory and tools it needs (the class catalog, approved policies, registration system access and the template library), sets approval boundaries, and names an observable success criterion before anything goes live. Managed operations then cover monitoring of escalations and sent messages, recurring evaluations against a fixed set of real questions, permission reviews when staff change, integration maintenance when the registration platform updates, and controlled changes to templates or rules, each reviewed before release. The service is remote and available in English and Portuguese.
A studio running this model can check its health with a short routine:
- Every schedule change has an approver's name and a timestamp attached.
- Every outbound message traces back to an approved template or a staff-reviewed draft.
- Escalations are reviewed weekly, and any repeated escalation becomes a rule change or a template update.
- Agent permissions match the current staff roster, reviewed at the start of each term.
- The set of real questions used for evaluation is rerun after every change to rules or templates.
Choosing the first task
The best first task is the one that repeats most often, carries the least risk and already has written answers. For most studios that is the waitlist and class information replies, not the recital roster, which can wait until the registration flow has run cleanly through a full term. Write the task down, name its approver and decide what the agent must never do before it touches a live message. 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 run class registration for a dance studio on its own? Not safely for every decision. A sensible setup lets the agent answer published questions, manage waitlist entries and prepare changes, while named staff approve any move between classes, roster change or exception. Most of the risk sits in placement and exceptions, so those decisions stay with people.
How does an agent handle recital costume lists? It works from the approved costume sheet for each group, collects the measurements the sheet requires and prepares sizing summaries for the costume coordinator. When a teacher changes requirements, the agent flags the affected students and holds the notice until staff approve it. Families then see one consistent list instead of several forwarded versions.
What data should the agent be allowed to see? Only what the task needs: the class catalog, the enrollment and attendance status of the guardian's own children, and the approved templates. Medical notes, custody details and payment disputes stay out of its reach or behind a staff-only step. The studio decides these limits during setup and reviews them whenever staff change.
How do we know the agent is working? Watch the escalation log, sample drafts for corrections and rerun a fixed set of real questions after every change. The signal to look for is whether staff accept drafts without rewriting the facts and whether every schedule change has an approver on record. Be Human tracks these measures as part of managed operations.
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.


