Where AI fits in GoStudent’s customer work

At GoStudent, Achilleas Kasimidis and his team set out to automate the first sales call, taking a new enquiry through to a trial booking. In a separate project, he wanted to free account managers from routine WhatsApp admin so they could spend more time on sales and retention.

In our interview, the online tutoring provider’s Global Director of Revenue Operations discussed both projects, including the WhatsApp assistant he built himself without a coding background. He prefers to get a pilot running, learn from it, and test again, without having every answer first.

The caller needed to qualify prospects

GoStudent was handling a high volume of inbound enquiries and wanted to contact prospects quickly. Its sales process involved two human calls. People signed up to express interest in tutoring, then an agent called to understand their needs, match them with a tutor, and book a free trial. After the trial, a second call aimed to turn the prospect into a paying customer.

The team handed the first of those calls to an AI caller. A human still made the closing call.

In the first few weeks of testing, the AI booked almost four times as many trial sessions as the human agents it was being tested against, Achilleas says. The team was excited. But conversion from trial to customer fell sharply, and the total number of paying customers stayed the same. More trials were costing more without bringing in more customers.

“Our sales agents had a clear mandate to close a customer,” Achilleas says. The human team’s incentives were tied to revenue. “The sales AI caller had a clear mandate to book a trial session.” The AI was qualifying prospects less strongly than the humans, so more people reached a trial.

GoStudent began retraining the caller. “So mandate is not just booking a trial, mandate is also pre qualifying, going through specific milestone steps before you obtain the yes and you move forward the customer to the next step.” The team also added qualification steps at sign-up, before the AI call, to learn more about prospects. Achilleas says fewer people went through the funnel, but more became paying customers. The changes took a couple of months, with tests of different flows, agent types, and styles.

For Achilleas, that trade-off depends on the service being sold. Education is a large investment for a family. Deciding to trust a tutoring provider takes time, as does explaining the service and nurturing that interest. GoStudent also invests in filtering tutors and in the quality of the service itself. “Not everything is the same, not all marketplaces sell the same product.”

The assistant freed time for customer conversations

Account managers use WhatsApp to chat with tutors, prospects, and customers, alongside central support channels. Customers asked about payment details, account access, and their next session. Tutors wanted to reschedule or cancel. Achilleas says account managers were spending almost two hours of an eight-hour day on this WhatsApp admin.

The WhatsApp assistant scans the inbox and handles almost all of this routine admin, Achilleas says. At aibl’s growthLIVE in June, he demonstrated how the assistant identifies incoming enquiries and flags their urgency, helping account managers prioritise replies. In the version he showed, staff still reviewed and sent messages themselves while the team tested its replies. His aim was to automate most of that sending.

In the interview, he estimated that taking on this admin could free up roughly two hours a day for account managers to call prospects, book trials, and retain customers. He reports that call time, revenue per account manager, and customers per account manager all increased.

Using Google Antigravity, he built the first working version in roughly two to three weeks and describes spending a couple of hours a week improving it. He puts its running cost at under €200 a month. At the time of the interview, around 20–25 account managers were using the assistant. He plans to hand it over to GoStudent’s technical team to make it more scalable and more stable, and extend it to other account managers.

A handful of account managers in the intended markets were involved from design through setup and use. Their feedback covered whether the language was too formal or informal, whether the assistant had understood the enquiry, and whether it had taken the right action.

For account managers handling hundreds of conversations, the prospect of relief was appealing. “I wouldn’t say trust yet, but you can easily get endorsement,” Achilleas says. Involving them in designing and improving the tool helped build trust over time.

Achilleas also built dashboards that let him see message volumes and bottlenecks, and distinguish admin from revenue-related activity within WhatsApp. He built feedback buttons into the tool so its users could tell him whether it was doing what it was supposed to do. “And again, that goes back to the first AI use case that we saw that wasn’t working as expected.”