The Bar For Your Team Went Up

Klarna cut around two thousand jobs on the strength of AI, watched its service levels fall, and then quietly hired people back. Sean opens this episode with that story, and Jared uses it as the clearest available window into what is coming for healthcare. His read is blunt: AI does not replace the human, it augments the human. An ultra high net worth traveller could book their own flight with an agent, but they pay fifty thousand pounds for a travel agent precisely so they do not have to. Just because something can be automated does not mean it should be, and the human layer is becoming the premium layer, not the redundant one.

From there the conversation moves onto the ground clinic owners actually stand on. Jared says he has yet to meet a single healthcare provider who reduced front office headcount after adding an AI receptionist, and argues the framing is wrong to begin with. The question is not how to go from twenty people to zero. It is how to go from twenty to ten while holding or improving service quality, because that is where the P&L moves. But that reframe carries a cost most operators have not priced in: if technology covers half the work, the half of the team still in the room has to be genuinely good. Having a body there to pick up the phone is no longer the job. Jared's conclusion is that AI is a talent strategy, not a technology strategy, and the same discipline applies to buying it. He walks through a rollout where patients recognised the AI receptionist, hung up and called back until they reached a person, a California plastic surgeon who answered the one more booking pays for itself pitch with what about the three people you annoy along the way, and a Sheffield dental clinic with four hundred five star reviews whose all singing octopus got things wrong and could not be corrected, because AI is not deterministic and one clinic was never going to be a priority for a vendor serving hundreds. When you buy intelligent tech, he argues, you are buying certainty as much as capability, and the right test is to buy it the way you write a job description: what is the job I am hiring this tech to do?

The back half is about listening, and about tools built for the wrong job. Single site clinics get from zero to a few million because the decision maker is in the room hearing patients, and at multi site that line of sight disappears. Jared describes a medical aesthetics group where rota volatility cancelled a handful of regular patients' appointments repeatedly, three or four of them left, roughly five to ten thousand pounds of annual value walked out of one site, and head office never heard about it. At another clinic, half the inbound correspondence is people who could not finish booking online, and those are only the keen ones. He closes on tooling: a practice management system is built around the clinical appointment and everything wires through that architecture, which is why one way appointment SMS still exists and why a private GP can put a friend's blood results behind four access codes she no longer remembers. You cannot carve a steak with a butter knife. And changing the knife rarely helps either, because unless you are moving off pen and paper, the answer to will a new PMS change the game is almost always no.