A pet-sitting service in eastern France. A high-touch operation with enough regulars and edge cases to outgrow what one person can hold in their head. Repeat clients, real reputation at stake.

He came with his business and a question: where could AI help? On the call I asked how he worked. Within minutes, intro days surfaced — two new dogs can't arrive the same hour without one going wrong. Then memory: owners getting asked the same questions stay after stay because there was no place to write the answer down. The ask wasn't really about AI. It was about keeping the personal touch and dropping the human bookkeeping.

Booking with constraint logic for intro days. Pet-specific memory tied to each reservation. Follow-up sequences triggered by stay events instead of by remembering. A tool that holds the context the operator was holding in his head.

First booking flow live with intro-day constraints in place. Pet notes auto-attached on every reservation. The 'who is this dog again?' moment gone from day one.

// What a Residency looks like in practice.

• ◆A few weeks in, special-diet edge cases surfaced. Added a structured per-pet medical field — no more lost notes in the margins.
• ◆A developer on the team pointed out we already had the anniversary dates — why not surface them automatically? Live a week later.
• ◆I suggested triggering post-stay surveys for long bookings — surfaced a recurring concern about pickup-time confusion that had never made it to him.