PetPilot

Shelter staff and volunteers constantly notice things the system has no field for. A dog peed purple, seemed off after a visitor, wouldn't eat. Today those observations get lost in notes nobody routes, and there's no clear rule for who should be told.

PetPilot watches for problems two ways and sends them to the right people.

First, a camera monitoring system watches the kennels and foster feeds, learns each dog's normal routine, and flags anomalies: missed feeding interactions, unusual activity during quiet hours, prolonged inactivity. Every flag is reviewable by a human before it becomes official, so the AI never overrides the record on its own.

Second, a remarks box lets anyone type a plain-language observation, which is instantly classified (medical urgent, behavior, facility, and more), matched against past incidents, and routed to the right team. Alerts are sent to Slack via Arcade so they reach people where they already are, and the observation lands back in the dog's record tagged and routed.

Together they close the loop: whether a concern comes from a camera or a person, it reaches the team that needs it. The routing layer is the wedge into the bigger problem, since it's the one piece that has to touch every other system (medical records, transfers, quarantine history), which makes it the foundation for a shelter's single source of truth.