Where this was said
Building the Product: From Human Operators to 98% Automation
At 29:58 · chapter starts 29:40
The technical journey from manual billing clerk to 98% automated agent is essentially a story of the founders automating themselves out of a job. Frédéric Renken explains the approach with characteristic precision: they started by taking over all the work entirely, with no human-handoff to the client. They became the back office. Then, having deeply understood each task, they built automation for it — first the simple, rules-based stuff, then the increasingly complex judgment calls as the models improved. The architecture was built for this: context layer first (access to all patient records, insurance portals, historical data), tools layer second (read/write integrations with every relevant system), intelligence layer third (the models). When reasoning models improved dramatically, the already-built context and tools layers meant Lassie's product got smarter without major re-architecture. The target automation threshold before releasing a product is 95%, not 100% — Renken argues that waiting for perfection would mean never shipping, and the remaining edge cases are learned in production through feedback from the staff who handle the exceptions.
Before writing a line of automating code, Lassie's founders sat in dental offices doing the billing work by hand. They automated away their own jobs — and that ground-level experience is why their product works at 98% accuracy when other AI tools fall short.
Every startup-versus-incumbent battle reduces to one question: can the startup lock in distribution before the incumbent copies the innovation? In dental billing, there is no tech incumbent to race against. The category was served entirely by human labor — and that changes the calculus entirely.
