A pediatric sleep consultant with a full client load and a waiting list. Every parent who needed help required her time directly. The expertise was there. The hours weren't. The goal wasn't to replace her - it was to build something that could carry her methods, her tone, and her judgement to parents who couldn't get to her directly.

Translating human expertise into AI behaviour is harder than it sounds. The gap between "here's what I'd say" and "here's a system that says it correctly, every time, to any parent" is significant. The discovery process had to come first - observing real consultations, unpacking the logic behind advice, identifying the patterns that made guidance land well with stressed, sleep-deprived parents. Only then could the AI be structured around those principles rather than around what a generic language model would default to. The other constraint was the user. This isn't a product for tech-comfortable early adopters. Parents accessing it at 3am, mid-crisis, needed onboarding and flows designed around that reality.

Released on the App Store with a paying user base. The product validated real demand and demonstrated that specialist expertise can be scaled responsibly through a well-designed AI system. The app was later taken down following the co-founder's exit from the project.