An EMS chain resolves 72% of member requests on WhatsApp with AI.

The majority of fitness businesses have a white-label app now, and this chain is no exception. They pay for a white-label Virtuagym app where members can schedule sessions and check basic things about their membership. But sometimes something goes off. Members do not have access, or they simply cannot figure out where to find the information they need. We did not expect how common this situation was. About half of the member requests that reach the chain on WhatsApp are things the app already does: booking, moving or cancelling a session, checking credits or prices.

The clearest example is a request for the next open slot in a particular city. When people want to train in the city tonight, they do not always care about the specific location. Instead of going to the app, they prefer texting on WhatsApp because it is faster and it is something they got used to.

Imagine being a studio operator and stopping a session, again and again, to open six calendars for one reply. Someone has to spend time answering those messages by hand. That is not the most productive work.

But the questions do not stop at six. A third of member messages arrive after the desk closes, and as the 24-hour clock below shows, the evening stretch is as busy as a late morning. Some of these late senders are prospects deciding whether to visit, since about a fifth of all fast requests are about trial sessions. Replying right away matters when people are active and making a buying decision, but expecting staff to answer from home every night is not a sustainable way to run the studios.

Then there is the issue of brand consistency across six locations. Anyone who cancels or moves a session with less than 48 hours’ notice loses the credit. Enforcing a rule like that means making sure all your employees are actually on the same page. Humans being humans, that is not always possible.

When we started in August 2025, we analyzed member requests for the past several months, and we noticed a clear divide. Some messages needed a trainer’s judgment, while the rest just needed a fast answer from the schedule or studio knowledge. The second group made up 74% of the total volume, and the breakdown looked like this.

We connected a WhatsApp assistant directly to the Virtuagym schedule. First, it read the chain’s website to absorb everything the business had ever written about itself and learn its brand tone, and then the owners sat down with us to fill in the unwritten rules. They explained how the 48-hour cancellation policy works, which details a trainer needs before a trial session, and exactly what to warn a member about.

Now, when a member asks for the nearest open slot in the city, the assistant checks availability across the studios in the city, then replies with the open times. When a cancellation comes in, the assistant checks it against the 48-hour rule the owners want enforced at every studio. The first time someone cancels late it issues a warning and lets it go, but after that, it cancels the session and takes the credit. Of course, some things belong with a person, so if a member asks whether they can train with a specific injury, wants to cancel their subscription, or brings up something off-topic, the assistant steps back and hands the conversation to a trainer.

To keep the assistant from inventing facts, we set up a second AI agent on the backend. The member-facing assistant writes the reply based on the website, the special rules the owners described, the schedule and the memberships from Virtuagym, and the second agent reads that draft back with a red pen, checking every claim against the sources. Nothing goes out if it cannot point to the exact rule or schedule slot.

Of course, the trust was not there from the very beginning. In the first months the team still answered most messages themselves. Over time we kept adding to what the assistant knows, and more and more questions were handled without anyone stepping in: in its first full quarter the assistant took part in about one in four of the conversations that got a reply, and by this summer in six in ten.

We went back and checked how the assistant worked out for this chain. In Q3 2026, 72% of the conversations the assistant answered ended without a trainer stepping in, and its replies came within minutes, where a person’s took three hours on average.

When members asked for a session, the assistant found the nearest open slots across the studios in the city directly from the live Virtuagym schedule, instead of a trainer checking the locations manually. It enforced the 48-hour cancellation rule on its own, and the staff no longer had to verify notice times by hand. Medical questions, subscription cancellations and off-topic messages bypassed the automation and went straight to a trainer, so the team only spent their time on the conversations that actually needed them.

Trial sessions changed too. Everyone who books a trial now gets a reminder before the session and a follow-up about the payment, so more of the people who book a trial actually show up.

Now that the assistant handles the routine schedule requests, members get a fast answer for the next open slot in the city on the channel they already use, and trainers only step in when a conversation actually needs them. Enforcing the rule for 48 hours notice consistently across six locations was always difficult to guarantee manually. The assistant applies the policy exactly the same way at every studio, so it gives the warning and it takes the credit. Staff no longer verify notice times, and the owners know the standard holds.

The next step is the phone. An AI receptionist can answer calls the way the assistant already answers WhatsApp. There is also a second angle: once AI has good data to work from, it can help make business decisions, and that is where Regymo Copilot comes in.