Every app has an AI feature now. Most of them are a chatbot bolted onto a corner, answering questions the app already could. When we added Lorem - the AI companion in the Pet Parent App - the goal was the opposite: make it genuinely useful to a pet parent at the moment they need it.
“Is this safe for my dog to eat?” “How often should I bathe a kitten?” “My cat’s acting strange - should I worry?”
These are real, anxious, middle-of-the-night questions. A generic chatbot answers them generically. A useful companion answers them in the context of this pet - its species, age, weight, and history that the app already knows.
The difference between a gimmick and a tool is grounding. Lorem isn’t answering in a vacuum - it has access to the pet’s profile and records. Guidance on diet, training, health, and behaviour lands better when it’s about your pet, not pets in general.
The engineering work is less about the model and more about what context you feed it and what you let it do - and, just as importantly, where it should stop and say “see a vet.”
The Pet Parent App is built with Flutter, so Lorem had to feel native on both iOS and Android from one codebase. The AI integrations live in the Node.js/Express backend, not the client - so we can evolve prompts, context, and safety rules without shipping a new app build every time.
That separation matters: the client stays thin, the intelligence stays server-side and updatable.
The most responsible thing an AI pet companion does is recognise when not to answer. Health questions have a line where the right response is “this needs a real vet.” Building that boundary in isn’t a limitation - it’s what makes the feature trustworthy enough to keep.
AI in a product isn’t about having a chatbot. It’s about meeting a real need, grounded in real data, with clear limits. Get that right and it stops being a feature people try once - and becomes one they rely on.