Scoping a pet-care chat assistant that has to remember

The job is a first-response thread. An owner describes a cat — food, a behaviour change, a symptom — and the next message has to know what was already said.

Teams ask for “a ChatGPT on our site.” That is how you get a box that forgets the timeline two turns in. Consultancy writes the memory and the stop rules first.

What the first assessment is for

We ask who the assistant is for, and what it is not allowed to decide. We ask where the conversation already lives — email, Instagram DMs, a clinic inbox. We ask what “done” looks like: a home-care note, a booking, a clear “see a vet.”

If there is no owner for the policy, volume will only create cleanup. We will say that out loud.

If the stall is repeating the same history every chat, a product with stored context can replace the copy-paste. The Kitty Care case study is that slice: a chat that keeps the thread, switches tone without resetting, and still works when the human desk is not in the room. We will not treat one product’s voice as yours.

This is not a diagnosis machine. If a person still has to judge every case, the model sits behind them.

When we tell you not to build yet

If you cannot write the refusal — emergency, medication, anything a vet must own — do not ship a public chat. If the knowledge is a pile of PDFs nobody has ranked, the first slice is a source map, not a widget.

If the questions are repetitive and the policy is written, we scope history → reply → handoff. That is AI consultancy, then a narrow product or agent if you want us to build the thread.

Bring three real owner messages and the sentence you wish the desk had sent. Book a consult or email hello@jamilglobal.com.