How a Yoga & Pilates Studio Can Tame Booking and Cancellation Replies with Claude Code

You finish the last class, lock up the studio, and open your phone. Three DMs on Instagram, five messages on your booking app, two auto-emails from the reservation site. “Is there still a spot in tomorrow’s morning yoga?” “Something came up, I can’t make today’s Pilates.” “I want to book a trial class, how much is it?” Your shirt is still damp, and there you are, sitting on the steps outside the studio, thumb-typing replies one at a time.

That was the nightly routine for a Pilates instructor I help run a small studio. She never cut a single corner on the quality of her classes, but the replies could only happen late at night, and some nights she didn’t notice a message until the next morning. That one missed message was a trial student who never converted into a member.

When I actually looked at the inbox, nine out of ten messages followed the same pattern. Checking availability, canceling, asking about pricing or what to bring, requesting a make-up class. When the content is that predictable, the first draft can go to a machine. The only thing a human has to decide is whether to hit send. This article is about how far you can lighten the load of “class info plus booking and cancellation replies” at a yoga or Pilates studio using Claude Code, written exactly the way I actually tried it.

Key takeaways

• Let Claude Code write the draft for booking, cancellation, and general-question replies, and let the instructor just press send. The nightly reply marathon shrinks to about 15 minutes.
• Put your class info (schedule, pricing, what to bring, cancellation policy) into one text file. The AI’s replies get dramatically more accurate the moment you do.
• Never paste a client’s name, contact details, or health notes straight into the AI. Decide your “hide the personal data” routine first.
• Anything involving money or disputes (make-ups, refunds, double bookings) always gets a human check.
• I’ve included a copy-paste prompt template and a check script that catches typos in your class schedule.

Who this is for: studios that look like this

I’m picturing a small-to-mid yoga or Pilates studio where the instructor is also the owner-operator. One to three staff, bookings scattered across a reservation site (Mindbody, Acuity, that kind of thing) plus Instagram DMs and texts, and a price list that lives half in your head and half in a notebook.

You don’t have a call center or a dedicated booking-system person like the big chains do. The person teaching the class is also the front desk, the bookkeeper, and the marketing department. So the biggest pain point is simple: replies eat into the time you’d rather spend preparing the actual class.

Which is exactly why AI helps here. The words you choose for each client matter, but nine times out of ten there’s a template underneath. Hand the templated work to a machine, and spend your saved time on the judgment call: “this person deserves an extra personal line.” That’s the split we’re going to build.

The studio workflow, from inquiry to post-class

First, let’s write out the current flow. You can only see where the rework happens once it’s on paper.

1. An inquiry comes in (auto-email from the booking site / text / Instagram DM)
2. You check availability and pricing (open the booking site in another tab)
3. You type a reply (writing a similar message by hand every single time)
4. A trial or full booking comes in (you note down the client’s info)
5. You send a reminder the day before (what to bring, location, cancellation policy)
6. You handle same-day cancellations and late arrivals
7. After class you send a thank-you and a note about next time

The steps that eat the most time and cause the most rework are 2, 3, and 5. You check availability while composing a message, then retype nearly the same thing from scratch. The day-before reminder is where both “forgot to send” and “reused the wrong template” tend to happen.

Common rework and headaches

Here are the things that actually went wrong on the ground. You’ll probably recognize a few.

• Quoting the wrong price: mixing up the $25 trial fee with the class-pack price, then having to correct it later and shaking the client’s confidence.
• Forgetting to mention the cancellation policy: not saying “free until 6pm the day before,” then arguing over a same-day fee.
• Missing a double booking: promising the same slot to one person via DM and another via the website.
• Leaving out what to bring: forgetting to mention grip socks for Pilates or whether mats are provided, then scrambling on the day.
• Slow replies: not seeing a trial-class DM until the next morning, by which time they’ve drifted to another studio.

The ones AI can reduce are the price misquotes, the forgotten cancellation policy, and the missing “what to bring.” Put the studio’s correct information in one place, and only let replies be built from that source, and the memory-based mistakes simply vanish.

What to delegate to Claude Code vs. what you decide yourself

This is the most important part. Delegate everything and you’ll have an accident. We draw a line.

Task	Delegate to AI	Human always decides
Availability-check reply draft	Yes, writes the message	A human eyeballs the actual open slot
Pricing / what-to-bring / policy info	Yes, builds from saved info	Applying exceptions or promo pricing
Day-before reminder text	Yes, fills the template	Timing and the final send
Cancellation acknowledgment	Yes, writes apology + policy	Whether to charge the cancellation fee
Make-up / refund handling	Drafts a candidate only	Any decision that moves money
Adding / editing client records	No	A human types it in by hand

The principle is simple. The AI writes the words; the human owns anything where money or a promise to a client moves. A human always presses send. Hold that one line and your late-night replies get easier without your problems getting worse.

If you’re still working out where to draw that line, the getting-started guide for non-engineers and the Claude Code getting-started guide make the boundary easier to grasp.

Use case 1: Drafting an availability-check reply

This is the most common one: “Is there still room in tomorrow’s morning yoga?” The instructor checks the booking site and hands only that result to the AI. The key is that the AI never decides availability on its own.

Run it as a checklist.

• Eyeballed the open slot for that class on the booking site
• Gave the AI only “available / full” plus the class time
• Spent one second re-checking the price and time in the draft
• Pressed send yourself

That alone removes the time spent writing from scratch. Two to three minutes per reply shrinks to about thirty seconds.

Use case 2: Apology replies for cancellations and late arrivals

Same-day cancellation messages are quietly draining to write, because you don’t want to come off cold or too soft. Teach the AI your studio’s cancellation policy and have it produce an apology that follows the policy.

The thing the human decides here is just one question: “Do we charge the fee this time, or waive it as a goodwill gesture?” The AI writes the words; the instructor makes the call. If a regular is out sick, you can still choose to waive it, and that warmth stays a human decision.

Use case 3: Day-before reminders for trial bookings

The reminder you send the day before a trial is nearly identical every time. Date, location, what to bring, cancellation policy, “wear something you can move in.” Turn that into a template and swap in only the name and time.

As numbered steps:

1. Prepare tomorrow’s trial-booking list (names and times only)
2. Run them through the reminder-template prompt one at a time
3. Eyeball whether the name and time in each draft are correct
4. Send from your booking app or messaging tool

No more forgotten reminders, and no more missing “what to bring” notes.

A copy-paste prompt template

First, put your studio info into one text file. Handing this over with every prompt is the whole trick. When the information is scattered, the AI fills the gaps with guesses. For how to keep this in a project file, the CLAUDE.md best practices are a good reference.

Here’s a studio-info template (rewrite it for your own studio):

# Studio basics
Name: Sunrise Yoga & Pilates Studio
Trial fee: $25 (tax incl.)
Class packs: 5 classes $150 / 10 classes $280
Class length: 60 min each
Cancellation policy: Free until 6pm the day before. After that, 50% of the class fee.
What to bring: comfortable clothes, water, a towel. Mats are free to borrow.
Pilates note: grip socks recommended (sold at the front desk for $8)
Location note: 3rd floor. Turn right out of the elevator.

Next, the prompt for drafting replies. The key line is the explicit “only use availability that a human has confirmed.”

You are the front-desk staff of Sunrise Studio.
Write a reply to the client using ONLY the [Studio info] below as your basis.
For anything not in the info (availability, exception pricing, etc.),
do not decide it yourself; write "Let me check on that."
Tone is polite, at most one or two emoji, 40 to 70 words.

[Studio info]
(paste the template above here)

[Client message]
Is there still room in tomorrow morning's yoga? I'd also like to know the price.

[Availability confirmed by a human]
Tomorrow's 7:00am morning yoga has space.

Write one draft reply.

Follow this prompt and the AI won’t invent availability. It uses the registered numbers for pricing exactly as written. The memory-based misquote disappears right here. If you want to sharpen your prompts further, take a look at advanced prompt engineering.

Privacy and security notes

Do not skip this part, because you’re handling client information.

• Don’t paste names, phone numbers, addresses, or health notes straight into the AI. What the draft needs is “is there space in tomorrow’s 7am yoga,” not the client’s real name. Build the draft with “the client” as a placeholder, and a human inserts the name right before sending.
• Mask before sending to a cloud AI. When context genuinely needs a name, swap in “Client A” and “Client B.”
• Don’t put bank-account or door-code numbers in the studio-info template. Stop at pricing and what-to-bring.
• Don’t paste a whole screenshot of a DM or email. Another client’s info can be lurking in the background.

Yoga and Pilates studios ask about health information (pregnancy, back pain, medical history) a lot. Treat that information with extra care. For just drafting a reply, you almost never need to hand the health information itself to the AI. For the basics of handling personal data, it’s worth a read through the FTC’s guidance for small businesses on protecting personal information.

A runnable check script: catching schedule typos

When you type out a schedule by hand, you get the occasional fat-fingered end time or an overlapping slot. Instead of catching those by eye, let a script flag them. It runs anywhere you have Node.js. Put the schedule in JSON, and it mechanically checks whether start is before end and whether the same time block is booked twice.

// check-schedule.mjs
// Usage: node check-schedule.mjs
const schedule = [
  { day: "Mon", start: "07:00", end: "08:00", name: "Morning Yoga" },
  { day: "Mon", start: "10:30", end: "11:30", name: "Pilates Basics" },
  { day: "Mon", start: "11:00", end: "12:00", name: "Relax Yoga" }, // deliberate overlap
];

const toMin = (t) => {
  const [h, m] = t.split(":").map(Number);
  return h * 60 + m;
};

const errors = [];
for (const s of schedule) {
  if (toMin(s.start) >= toMin(s.end)) {
    errors.push(`${s.day} ${s.name}: start (${s.start}) is at or after end (${s.end})`);
  }
}

for (let i = 0; i < schedule.length; i++) {
  for (let j = i + 1; j < schedule.length; j++) {
    const a = schedule[i];
    const b = schedule[j];
    if (a.day !== b.day) continue;
    const overlap = toMin(a.start) < toMin(b.end) && toMin(b.start) < toMin(a.end);
    if (overlap) {
      errors.push(`${a.day}: "${a.name}" and "${b.name}" overlap in time`);
    }
  }
}

if (errors.length === 0) {
  console.log("OK: no problems found in the schedule");
} else {
  console.log("Needs fixing:");
  for (const e of errors) console.log(" - " + e);
  process.exitCode = 1;
}

Run it with node check-schedule.mjs as is, and it points out that the 11:00 Relax Yoga (deliberately added) overlaps the 10:30 Pilates. Run your schedule through this every time you update it, and the input typos that cause double bookings get stopped before they go live. Ask Claude Code to “rewrite this schedule array with my own studio’s timetable,” and you can hand off filling in the array too.

What changed before and after

Here’s the rough before-and-after at the studio I help.

• Before: 40 to 60 minutes on replies every night. Trial DMs going unanswered until the next morning twice a week. Two or three price or cancellation-policy misquotes a month.
• After: replies in about 15 minutes. Because the draft already exists, you can answer in spare moments. Misquotes are near zero thanks to referencing the template.

A rough sense of the ROI

The numbers shift with your studio’s size, but here’s the way to think about it.

If reply work drops from 45 minutes a day to 15, that’s 30 minutes a day, roughly 15 hours a month freed up. Value the instructor’s time at $35 an hour and that’s about $525 worth of time back each month. If that time lets you add a single class slot, you recoup it even faster.

The bigger win is the memberships you used to lose because trial DMs went unanswered too long. Even one trial-to-member conversion a month means a class pack of $150 and up starts stacking. Preventing missed leads often beats the time savings.

If you want to standardize this across a company or multiple staff, you need a system that prevents knowledge from getting stuck in one person’s head. For team rollout and operational design, that’s what training and consulting covers. If you’d rather try it solo first, start with the free PDF and learning materials.

FAQ

Q. I’m worried the AI will quote the wrong price. A. Hand over the studio-info template with every prompt and instruct it to “write ‘Let me check on that’ for anything not in the info,” and the AI won’t invent numbers. One human glance at the end is enough.

Q. Can I do this even if I’m not good with computers? A. You don’t have to touch the check script. Just pasting the prompt and generating a reply is enough to get value. I’d start with the getting-started guide for non-engineers to get a feel for it.

Q. Is it okay to put a client’s name into the AI? A. I recommend a no-names workflow. Build the draft with “the client” as a placeholder and insert the name right before sending. You almost never need to hand over health information either.

Q. Can it connect to my booking site automatically? A. The scope of this article stops at “drafting the reply.” A human hands over only the availability they’ve confirmed. Automatic integration carries heavy responsibility when something breaks, so add it slowly once you’re comfortable. For ideas on streamlining further, see productivity tips.

Q. How long until I see results? A. The day you make the studio-info template, replies already get easier. You should feel the drop in misquotes within one to two weeks.

What I found when I actually tried it

At the Pilates studio I help, I ran it for two weeks first: “AI drafts only, the instructor sends.” I wanted to confirm two things. Does reply time actually go down, and do the price and cancellation-policy misquotes drop?

Reply time, which used to be 40-plus minutes batched at night, settled into roughly 15 minutes split across spare moments. The single biggest factor was making the studio-info template up front. Just having it made the “made-up pricing” disappear from the AI’s replies.

The check script, for its part, found one real overlap of two slots in a Friday schedule. That’s a slot I’d have missed reviewing by eye. Stopping a double-booking in the bud before it went live was quiet but meaningful.

On the other hand, I stopped sending AI drafts as-is for anything involving make-ups or refunds. When money moves, every client’s situation is different, and unless a human sets the tone it can read as cold. The AI writes the words, the human makes the money call. Drawing that one line is how we landed on today’s setup: easier, without more trouble.