Now You Can Ask an AI Assistant About Your Own Studio. Here's What It's Actually Good For.

A studio owner I work with asked me a simple question last month: "Is my Tuesday 7pm class actually worth keeping?"

It sounds like a question software should answer in two seconds. It isn't. To answer it honestly you need attendance over the last few months, the revenue that class brings in, whether the people in it also come to other classes, and what else you could put in that slot. Those numbers live in four different screens. She had the tabs open. She still couldn't tell me.

That gap — between the data you already have and a question you can actually answer with it — is the thing I've been chipping at for two years. Last week we shipped something that closes a lot of it. You can now connect an AI assistant like Claude or ChatGPT directly to your studio's data and ask it questions in plain English.

I want to be careful about how I describe this, because "we added AI" is the most worn-out sentence in software right now. So let me tell you what this is, show you what it's good at, and be honest about where it falls over.

What almost every "AI feature" actually is#

When a booking platform says it added AI, it usually means one of two things. Either there's now a chatbot trained on the help docs that can tell you how to issue a refund, or there's a little "summarize" button that writes a paragraph about numbers you were already looking at.

Both are fine. Neither is interesting. They're a layer on top of the same screens you already had to go hunting through.

What we built is the other direction. Instead of putting an AI inside our app, we let you point your own AI at your studio's data. You connect the assistant you already use, like Claude or ChatGPT, and it gets read access to your numbers. Then you ask it whatever you want, the same way you'd ask a sharp assistant who happened to have your whole studio memorized.

The difference matters because the assistant isn't limited to one screen. It can pull from your schedule, your revenue, your attendance, your membership list, and your campaign data in a single answer. That's the part a dashboard structurally can't do. A dashboard shows you each of those things in its own box. It can't reach across them, because reaching across them is your job. Now it doesn't have to be.

What you can actually ask it#

Here's where it stops being abstract. These are real questions owners ask me, and the thing they have in common is that none of them live on a single screen. That's exactly why they're hard to answer today and easy to answer now.

"Which of my classes makes the most money per booking?"

Your dashboard shows revenue in one chart and attendance in another. It doesn't divide one by the other, because that's not a number anybody pre-built a widget for. The assistant does it on the spot:

Class	Revenue (90 days)	Bookings	Per booking
Saturday Workshop	$4,200	140	$30.00
6am Power Vinyasa	$5,500	275	$20.00
Lunch Express (30 min)	$3,300	330	$10.00
Community Flow (donation)	$1,800	360	$5.00

Look at that for a second. Community Flow fills the most spots and brings in the least money. It's a marketing cost wearing a class schedule's clothes, and that's fine as long as you know it. Your Saturday Workshop is quietly your best earner per head. Lunch Express is the volume engine. None of this is in any single chart. The assistant ranked it in one answer, and now the Tuesday-7pm question has actual math behind it instead of a gut feeling.

"Where am I leaving money on the table in my schedule?"

This one needs three things at once: your upcoming schedule, how full each class is, and who's stuck on a waitlist. The assistant holds all three together and finds the mismatch. It'll point out that your Saturday 9am has run a waitlist three weeks straight while the 10:30 right after it sits half empty, so you're turning people away and then running a near-empty room an hour later. Or that one instructor's Thursday class waitlists every single week, which is about the clearest signal you'll ever get that it's time to add a second session. Capacity problems hide because the waitlist lives on one screen and the empty seats live on another. Put them in the same answer and the fix is usually staring back at you.

"What changed in the last month, and should I be worried about anything?"

This is the question I think it's best at, because it's the one nobody has time to ask. The assistant can compare this month to last across revenue, bookings, and attendance, and tell you what moved. In practice it tends to surface the thing you weren't watching — your no-show rate creeping up on evening classes, say, or one instructor's classes filling slower than they used to. A dashboard will show you a line that's trending down. The assistant tells you which line, when it started, and what you might check.

"Are my members actually using their memberships?"

You might have 60 active members and feel good about it. Ask the assistant who hasn't booked in the last 30 days, and the good feeling gets more specific: 18 of those 60 have gone quiet. That's not a number, it's a list of people about to cancel, and you found them while there's still time to send a "we miss you" email instead of finding out when the charge fails.

"Did my spring promo actually do anything?"

Promo codes are easy to launch and easy to forget to measure. Ask, and you get the unsentimental version: your SPRING10 code got 38 clicks, 3 bookings, and $54 in revenue. That's not the answer anyone hopes for, but it's the one that stops you from running the same dead promo next quarter out of habit. The assistant pulls clicks, bookings, and revenue for each campaign and lays them side by side, which is the whole difference between "the promo felt busy" and "the promo converted at eight percent." If you want to go a level deeper on which channels actually walk students through the door, that's its own rabbit hole, and I wrote about it in how to know which marketing is actually working.

And because it's a conversation, you don't stop at the first answer. You ask "okay, why is Tuesday so dead?" and it goes and looks. That follow-up loop is the whole point. You're not building a report. You're thinking out loud with something that can check the books while you talk.

The quality of the answer tracks the quality of the question, though. Give it a time window. Tell it what you're actually trying to decide. "Should I move my Tuesday 7pm or kill it?" gets you somewhere "how's Tuesday doing" never will, because now it knows what it's helping you choose, and it can reason toward that instead of just reciting numbers back at you.

What one session actually surfaces#

To make all of that concrete, I pointed the assistant at a real dance studio's account — around 60 members, the usual mix of kids' classes, adult drop-ins, and a couple of premium workshops — and asked it the questions above, one after another. Here's everything it came back with, in one place:

You ask	What comes back
Which class earns the most per booking?	Saturday Master Class — $32 a booking, four times what Open Studio brings in. The master class is the quiet earner; the drop-in is volume, not money.
Which class is secretly a loss leader?	First Class Free fills the room and earns $0 — a marketing cost wearing a class schedule's clothes. Fine, as long as you meant it.
Are members using their memberships?	60 active, but 18 haven't booked in over a month — a list of people about to cancel while there's still time to do something about it.
Did the spring promo do anything?	The spring code: 250 clicks → 20 bookings → 8%. The summer one managed 2%. One's worth running again.
Who's on the Saturday waitlist?	Friday Teen Hip-Hop, 7pm: 7 waiting, two weeks running — about the clearest "add a second section" signal you'll get.
What changed this month — anything to worry about?	Revenue down a little ($1,950 vs $2,230 last month), and two things sitting underneath it: evening no-shows up from 5% to 22%, and Adriana's Advanced Ballet filling slower than it used to — 74% down to 40%.

That last row is the one that earns its keep. Nobody had flagged Adriana's class; it wasn't failing, just quietly cooling — the kind of slide you only notice once it's a problem. Sitting next to the no-show creep in a single answer, it's a conversation you have with her this week instead of a number you explain to yourself next quarter.

How you turn it on#

This part takes about a minute. In your dashboard, go to Settings → AI. You'll see step-by-step instructions for whichever assistant you use — Claude.ai, Claude Desktop, or ChatGPT.

The short version: you copy one connection address (https://www.studiobase.org/api/mcp) into your AI client, and the client walks you through a quick authorization screen. You approve the connection the same way you'd approve any app connecting to an account, and if you run more than one studio, you pick which one to connect. That's it. From then on, the assistant can answer questions about that studio.

Under the hood it runs on the Model Context Protocol, the emerging open standard for letting AI assistants talk to outside tools. It's the same plumbing those assistants use to connect to everything else. You don't need to know any of that to use it. I mention it only because if your AI client supports MCP, it'll work, and more of them support it every month.

What it won't do, on purpose#

I'm suspicious of any feature description that doesn't name a limit, so here are the real ones.

It can look, but it can't touch. The connection is read-only. The assistant can tell you 18 members have gone quiet; it can't email them for you. It surfaces the thing to do. You still do it. That's deliberate. I'm not handing an AI the keys to message your students or move your money. But it does mean this is an analysis tool, not an autopilot.

It's only as smart as your data is honest. If you don't check students in, your attendance numbers are fiction, and the assistant will confidently reason about fiction. Garbage in, confident garbage out. This rewards studios that keep clean records and quietly punishes the ones that don't.

It can be wrong in a very convincing voice. The numbers it pulls are real. They come straight from your data, not from the model's imagination. But the story it tells about those numbers is its best guess, and its best guess can be wrong. Treat it like a clever intern, not an oracle. When it says "your Tuesday class is dying," check the figure it's pointing at before you cancel anything.

It can't see your students' private details, and neither should it. This was a hard line for me. The assistant gets aggregate numbers and, when you search members, their names and visit stats. It never gets email addresses, phone numbers, or anything about how someone pays. Your students' contact and payment information never leaves your studio, even when you're the one asking. So no, you can't use this to pull up someone's phone number — and that's a feature, not a gap.

Why I think this is the right shape#

The honest reason I built it this way: I don't think the answer to "AI in studio software" is a smarter chatbot that lives in my app. You already have a good assistant. You're probably already paying for one. What you didn't have was a way to let it see the one set of numbers that actually runs your business.

So instead of competing with the AI you already use, StudioBase just hands it the keys to read your books. Your tool, your questions, your data. We stay out of the middle.

It's new, and it'll get sharper as people use it and tell me where it's dumb. But the Tuesday-7pm owner texted me a screenshot last week of a five-line answer that took her assistant about ten seconds to produce, with the note: "I've wanted this number for a year." That's the bar I care about.

StudioBase now lets you connect your AI assistant to your studio in about a minute. Open Settings → AI to set it up, or start a 14-day free trial first, no credit card required.

Related: this is the same instinct behind the rest of the platform — give a small studio owner the answer instead of the spreadsheet. I wrote about where that came from in I built a studio platform for one studio, and about the numbers worth watching in the first place in the studio metrics that actually matter.