The Saturday-night lead
"Trial class boleh?" comes in at 9pm. Your front desk sees it at 10am. By then she's booked at the studio down the road.
The four empty ones are almost pure profit. We go after them — off your existing WhatsApp, on your Rezerv schedule.
We map the bookings you're missing this week — free, no commitment. Or just talk to us →
Three moments you already know — every one of them a reformer that stays empty, in a class you are paying for either way. At the studio on this page the median package the AI closed was RM322.92. That is what one recovered "let me think about it" is worth.
"Trial class boleh?" comes in at 9pm. Your front desk sees it at 10am. By then she's booked at the studio down the road.
Someone asked your price three days ago and went silent. Nobody followed up. That was a member.
A client has three classes left, expiring Friday. Rezerv knows. Nobody told her. The credits vanish — and so does the renewal.
19.8% of their customer messages land outside 10am–10pm, and 19.7% did before us too. The demand was always there. The only thing that changed is whether anything answers it.
One client — a Pilates and yoga studio with two outlets in the Klang Valley, live 27 June 2026 — five weeks, and their own Rezerv reports. Here is all of it.
Quoted in the chat, paid in the chat, granted into Rezerv by the AI itself. No human touched the sale. This is the floor, and it is the number we'd judge us on.
31 packages in July, none of them touched by a person. For 18 of 31, the buyer was new to Rezerv and the AI created the account too.
77.2% of that revenue — 22 buyers, RM10,234.06 — came after a follow-up. Median 26 messages and 1.68 days to payment; 11 of 31 bought the same day.
Money paid by customers the AI had actually talked to on WhatsApp, from launch to 31 July. The conversation happened, then the payment happened — a link, not a cause, and not money the AI closed.
174 paying customers and 215 transactions, launch → 31 Jul 2026. Over the slightly wider window launch → 2 Aug 2026, the desk handled 1,466 conversations, 1,048 of them genuine new leads.
70 brand-new customers who had never bought before, worth RM46,264.76 — measured launch 27 Jun → 2 Aug 2026, so it overlaps the RM122,446 above rather than adding to it.
It closed RM13,248 — the card headed “AI alone” — and we can show you the thread behind every ringgit of this one.
Every ringgit the studio collected in July 2026 — every class, every channel, everything the staff sold at the counter too. Not ours, theirs: we did not make them this money.
Their best month since their Rezerv records begin: +25.4% on their best month before us (Jun 2025, best pre-Jawab month), and +44.0% on the month right before we went live.
Earned revenue set a record too — RM76,389.90, past the RM73,022.61 they earned in June 2026.
Every ringgit the studio collected in July 2026 — every class, every channel, everything the staff sold at the counter too. Not ours, theirs: we did not make them this money.
Their best month since their Rezerv records begin: +25.4% on their best month before us (Jun 2025, best pre-Jawab month), and +44.0% on the month right before we went live.
Earned revenue set a record too — RM76,389.90, past the RM73,022.61 they earned in June 2026.
Money paid by customers the AI had actually talked to on WhatsApp, from launch to 31 July. The conversation happened, then the payment happened — a link, not a cause, and not money the AI closed.
174 paying customers and 215 transactions, launch → 31 Jul 2026. Over the slightly wider window launch → 2 Aug 2026, the desk handled 1,466 conversations, 1,048 of them genuine new leads.
70 brand-new customers who had never bought before, worth RM46,264.76 — measured launch 27 Jun → 2 Aug 2026, so it overlaps the RM122,446 above rather than adding to it.
It closed RM13,248 — the card headed “AI alone” — and we can show you the thread behind every ringgit of this one.
Quoted in the chat, paid in the chat, granted into Rezerv by the AI itself. No human touched the sale. This is the floor, and it is the number we'd judge us on.
31 packages in July, none of them touched by a person. For 18 of 31, the buyer was new to Rezerv and the AI created the account too.
77.2% of that revenue — 22 buyers, RM10,234.06 — came after a follow-up. Median 26 messages and 1.68 days to payment; 11 of 31 bought the same day.
Three measurements over three different windows, never added together: the money the AI closed by itself sits inside the money that moved through conversations it was part of, so summing them would count it twice. All of it comes out of the booking system's own accounting and our database. The largest is July 2026 — one closed month at one studio, whose own promotions, instructors and season were working on that month too. The smallest, 31 closes out of 1,466 conversations, is 2.1%: a floor from five weeks of an immature system. It is what happened there — not a rate, and not a result, that we’re promising you.
Real threads from the studio, first message to last. Where the transcript skips a run of routine messages it says so on the spot, and the studio's own staff turns are left out — bar one, the front-desk reply you can see in the after-hours thread. Her first name is a stand-in, not her real one; the AI introduces itself here as Jawab in place of the persona name it uses in the studio's chat; the studio, its outlets, its instructors and its own package names are masked out. Everything around them is the message as it was sent.
She walked past the signboard and messaged knowing nothing about Pilates. She paid RM214.92 thirty-eight minutes later.
•••• is a name we removed — the studio, its outlets, its instructors, its own package names. Her first name is a stand-in, not her real one, and the AI's name reads Jawab here in place of the persona name it uses in the studio's chat. Everything around them is the message as it was sent.
The moment she said no to the expensive option, it stopped selling it and closed the cheaper one.
It answers in English and Manglish, Bahasa Malaysia or Mandarin, picking from what she typed — and your team can pin a customer to one language.
Between the AI writing a reply and your customer reading it, a stack of ordinary code checks it — not another AI being asked nicely. A reply that fails is blocked before it sends.
Illustration. Not wired to your booking system, so it answered from memory. The 6pm was full, and a member is now double-booked.
Illustration. It checked Rezerv, saw 6pm was full, and pointed her at 7pm — no double-book.
Every catch is stamped on the reply it fired on and kept, so we can see which guard tripped, how often, and go fix the cause.
One more sits in the AI’s instructions rather than in the code above: when the facts run out, it is told to say it will check with the team instead of guessing.
No form to re-key tomorrow morning. Creating her Rezerv account and granting the package are one switch you turn on — and can turn off again.
Any banking app, DuitNow QR or card link. More than RM1 short of the total it quoted her, and it does not proceed. If she takes the card link, your team gets a ping the moment the link goes out, asking them to check that payment off when it lands.
It reads Rezerv before it writes, every time — so if your staff already activated her manually, it sees that and doesn't grant her twice.
The actual PDF out of Rezerv, not a picture of one, plus her password-and-first-booking guide. Your team gets the ping: paid, activated, who.
Built for the group case too: if three friends pay in one transfer, it handles it as one sale and provisions each of them separately. It has not been needed at this studio yet.
Your booking system already holds the facts a generic chatbot is blind to. We wire into them — so the WhatsApp desk works off what's actually true in your studio, not what an AI guessed.
At the studio on this page, its Rezerv package grants succeeded 96.9% of the time (July 2026, 31 of 32 grants), median 2.55 seconds.
Booking, cancelling and freezing stay with your team — the AI passes those over and says so.
77.2% of everything the AI closed at the studio came after a follow-up — 22 of the 31 buyers, RM10,234.06. Day two, day four, day fourteen. That is the part a busy front desk never gets to.
For every follow-up it sent, it held back or postponed about 1.8.
23.3% of the ones it did send got a reply within 48 hours.
A gate checks before every chase. Whether it still makes sense — time passed, how often she has already gone quiet, whether a human has taken over — and then a second pass re-reads the last 30 messages and asks again.
And it stops on its own. The sales chase stops when she pays, opts out, or your team replies from the dashboard. Activating starts a different one, and that stops the moment she books her first class.
In her birthday month it composes a poster with her name on it and sends it, in your studio's own artwork — another switch you turn on, and can turn off. Chases run inside an 8am–11pm window, never overnight, and each one is written in her language.
And the chase is yours, not ours. Tell us in the group and we'll change any step — whether it runs at all, when it fires, and what rides with it.
The dashboard installs on your phone like an app: you can read every chat the desk has ever had, pause the AI on any one of them, and send us a change from the same screen. That is the route we build for. If you would rather just type it in the WhatsApp group we already share, that works too — no portal, no ticket form.
Illustration of the flow.
Every request becomes a tracked ticket on our side. You never have to chase us for a status.
It's the trial lead you didn't reply to by Tuesday.
We'll show you exactly where WhatsApp is leaking bookings this week — then run the desk that answers, follows up and sends her to your Rezerv timetable to book.
We read your recent chat history. Packages, trial offers, schedule, handoffs, and the way your best front-desk person turns a “how much?” into a regular. Then we write her voice into it — no setup work on your side.
We bring your whole prior WhatsApp history across. Then we run the live channel: it replies, follows up and routes the edge cases — and from the dashboard you can take over mid-sentence on any thread.
Then we watch it every day. Anger, a billing dispute or “can I talk to a person” stops the AI on that chat and pings your team with the thread. During the day it nudges them again a quarter of an hour later, and if nobody has picked it up by the half hour the AI takes the thread back rather than leaving her sitting on read. A pregnancy or injury question works the other way on purpose — she gets a careful answer and your team gets a note, because going quiet on her mid-question would be worse. And if your phone drops offline, that raises its own alert.
Takes about 20 minutes. You keep the audit either way.
Built for reformer and mat Pilates, boutique fitness, yoga, barre, group fitness. Also running on the same desk: a beauty and wellness chain, a gym chain, and a B2B SaaS company.