The Day of Voicemails Is Gone

The Business of a Clinic usually invites operators, providers and investors rather than suppliers, but Jared makes an exception for Thomas Wojtowicz, co-founder of BookedSolid, because the questions he works on come up in almost every episode.

BookedSolid started as booking software after Thomas's co-founder, Oliver, built a booking app for his dad's gym in a week. They soon realised they were years too late for booking software, but one feature, automated scheduling over WhatsApp, was the part everyone wanted. They dropped the rest and built an AI front desk around it.

Thomas is candid about the early product: they went to market, realised it wasn't good enough, and went back to listen to customers. Their first clinic customers came through a practice management system marketplace, and by his account they added voice and signed their first paying customer at the end of 2024.

In his view, most of the progress since then has come less from new models and more from learning how to turn a language model into a dependable agent, especially on phone calls, where latency matters.

Both he and Jared hear clinics say they will build it themselves. Thomas compares it to handing the world's fastest car to someone who can't drive: AI is a tool, and the last stretch of quality is the hard part. Clinics don't build their own practice management systems, he notes, yet AI makes building look easy.

The biggest lesson from implementation is that AI exposes messy set-ups. Unused appointment types, practitioner names inside appointment types, or a separate account for each clinician all carry straight over, and the larger groups he works with tend to have the tidiest systems.

On accountability, Thomas points out that AI is scrutinised far more closely than people, who also make mistakes. He separates small conversational slips, such as getting a day of the week wrong and correcting it, from the serious mistakes clinics worry about.

His approach to the serious ones is not to give the AI access to information it doesn't need in the first place. Jared turns that into a buyer's question: what red lines has a supplier drawn, and how bad could it get if something goes wrong? Buying AI, he suggests, is a bit like hiring someone who needs context and close supervision at first.

Patients have become far more comfortable talking to AI over the last two years, Thomas says, although older patients were wary at first. In his experience the benefit is clearest in large multi-site groups with heavy call volumes. Both agree the right framing is AI and human rather than AI or human.

Patients should always be able to ask for a person, and the aim is to let stretched admin teams do their existing jobs better, so that a clinic can grow without admin costs rising at the same rate.

Thomas also argues that patients now expect access around the clock and will go elsewhere if nobody answers in the evening; out-of-hours calls that would otherwise go unanswered are the easiest place to start. Looking ahead, he expects AI to take on more of the routine admin and outbound messages such as reminders, and eventually triage, which he believes the technology isn't ready for yet.