Booking agent replaced 3 hours of daily admin at a yoga studio

The problem: not a booking problem, an admin problem

The studio owner — we'll call her K. — didn't have a problem selling classes. Her studio ran at around 80% capacity and she had a waitlist system through Mindbody. The problem was everything around the bookings: the messages from members asking to swap to a different class, the waitlist notifications that needed to go out when someone cancelled, the 24-hour reminder emails, the follow-ups when a member hadn't shown up in three weeks.

None of these tasks required judgment. They followed predictable patterns. But they added up to roughly three hours of her time every day — time she was spending at 7am and 9pm, between classes, on her phone.

She had already looked at automations through Mindbody's built-in tools and a Zapier integration someone suggested. They handled a slice of it. But anything that required reading a message and responding contextually — "can I switch Tuesday's 6pm to Wednesday's 8am instead?" — still landed in her inbox.

The agent: what it actually does

The agent K. deployed handles four distinct workflows that previously required her attention:

1. Rescheduling requests

When a member sends a reschedule request via email or the studio's contact form, the agent reads the message, checks Mindbody's availability for the requested alternative, and responds with either a confirmed switch or a list of available alternatives. If the member's preferred class is full, the agent offers the waitlist and confirms once a spot opens.

The agent uses a simple template but fills it with the actual class name, time, instructor, and whether the credits carry over. K. reviewed the first 50 responses before trusting it to run unsupervised. She found two errors in that batch — both involved edge cases around multi-class packages she'd set up incorrectly in Mindbody, not agent errors.

2. Waitlist management

When a cancellation comes in, the agent checks who's first on the waitlist, sends them a notification with a 2-hour claim window, and if they don't respond, moves to the next person. K. previously did this manually, often missing the window or doing it too late for the next person to make it to class.

3. Reminder sequences

Standard 24-hour and 2-hour reminders, with class-specific details filled in from Mindbody's API. Nothing novel here — this was possible before — but the agent handles exceptions: if a class has a substitute instructor, it flags that in the reminder automatically.

4. Re-engagement for inactive members

Members who haven't booked in 21 days get a short message (not a promotional push — just a note saying they haven't seen them in a while, asking if everything is okay and whether they'd like help finding a class that fits their schedule). K. wrote this one herself. The agent sends it and tracks replies for her review.

How the setup actually worked

K. used Relevance AI, a no-code platform for building AI agents that can connect to external APIs. The Mindbody integration was the most time-consuming part — not because it was technically complex, but because Mindbody's API documentation is inconsistent in places, and it took a few days of back-and-forth to get the authentication and data formatting right.

The full setup took 11 days. Days 1–3 were reading documentation and mapping out the workflows she wanted to automate. Days 4–7 were building and testing the Mindbody connection. Days 8–10 were writing and refining the agent's response templates. Day 11 was a supervised run where she watched every response go out before approving it.

K. has no technical background. She describes herself as "comfortable with spreadsheets and not much else." The main skill she needed was being precise about what she wanted the agent to do in each scenario — which, she notes, is also the main skill involved in training a new human assistant.

Results, honestly reported

After five months in production, K. estimates the agent handles around 85% of the booking-related messages she used to manage manually. The remaining 15% — requests that are genuinely unusual, complaints, or questions about pricing — still come to her, but they're now pre-tagged by the agent with a suggested priority level.

Her daily admin time dropped from 2.5–3 hours to roughly 20–30 minutes. She's used that time to start offering private sessions, which now account for around 12% of her monthly revenue.

The re-engagement messages have a 34% response rate. She didn't track this before, so there's no comparison — but she says it feels higher than when she sent them manually and "remembered to do it about half the time."

What to try first if this sounds relevant to you

Before building anything, map out the five most repetitive communication tasks in your business that follow a predictable pattern. Not "answer customer questions" — that's too broad. Something like: "When someone emails to reschedule, I read it, check availability, and reply with the same three options." If you can write out the decision tree on paper, you can probably turn it into an agent.

Start with one workflow, not five. K.'s instinct was to automate everything at once. She was talked out of it by someone who'd done this before. The value of starting with one is that you learn where the edge cases are before they compound across multiple automations.

The Mindbody + Relevance AI combination K. used has a learning curve but is well-documented. If you're not on Mindbody, the same principles apply to any booking system with an API — most do.