When a pet parent clicks your clinic's intake link, they have a brief, warm conversation with an AI assistant. Here's what happens behind the scenes and what you'll see in your dashboard afterward.

## The conversation (about 3 minutes)

The AI opens by setting expectations:

> *"Hi! I'll ask you six quick questions about your dog — takes about three minutes. At the end I'll give you one simple thing you can do this week to make the visit better. What's your dog's name, and how old are they?"*

It then walks through six required questions, in order. After the first turn, every question is preceded by a **one-sentence warm acknowledgement** tailored to the previous answer — for example *"Puppies are a handful, but they bring so much joy"* before asking about vet-visit fear.

Once the AI learns the dog's name, it uses the name in every subsequent question (*"Got it, is Milo around other dogs often?"*). The system prompt forbids labels like "reactive" or "aggressive" and diagnostic language mid-intake. The tone is explicit: *listening, not prescribing.*

A **progress bar** above the chat ticks up as each of the six signals is captured so the owner can see how close they are to done. The bar flips to "Wrapping up…" once all six are in and hides when the intake is complete.

## The six signals

The AI extracts six structured data points (called "signals") from the conversation. Each has a category, a normalised value, and a confidence score:

1. **Dog age** — coarse bucket + specific age if mentioned
2. **Visit timing** — when the appointment is, how urgent
3. **Vet visit fear** — how anxious the dog gets at the vet
4. **Handling sensitivity** — how the dog reacts to being touched / restrained
5. **Human reactivity** — risk of reactive behaviour toward people
6. **Dog reactivity** — how the dog responds to other dogs

Two optional follow-ups trigger when the AI hears the right cue: **sedation context** (if meds are mentioned) and **bite risk** (if growling or snapping is described).

## What the AI produces

### Behaviour profile

From the six signals, the AI generates a structured behaviour profile — fear level, handling sensitivity, human reactivity, dog reactivity, a readiness stage (not ready / in progress / ready), and any **safety flags** that need special clinic attention (bite risk, sedation context, etc.). The profile is persisted to your database and shown to your team under [Intake Summaries](/docs/understanding-intake-summaries) and inline on the [Today](/docs/todays-schedule) view.

**The owner does not see the structured profile.** During the chat they only get a friendly one-sentence summary like:

> *"Sounds like Milo's pretty relaxed in most situations but gets a little nervous around other dogs. That's common, and we have something that'll help."*

### One recommendation, opt-in framed

Instead of dumping a multi-task plan on the owner, the AI asks an **opt-in question** and waits:

> *"Based on what you told me about Milo, we've got one thing that'll really help your visit. Want to hear about it?"*

If the owner says no, the AI thanks them warmly and ends the intake. If they say yes, the AI proposes **one** curated preparation module from a short vetted list — each with a one-sentence pitch the AI reads aloud, like:

> *"Spend two minutes a day touching Milo's paws with a treat — it makes the exam painless."*

A single **RecommendationCard** renders in the chat with two buttons:

- **Count me in** — accepts the module. It's added to the owner's preparation schedule and visible to you on the Today view's prep indicator.
- **Show me a different one** — the AI picks the next best module from the list and proposes that instead. Up to 3 swaps are allowed; after that, the AI presents the best remaining match and says *"this one's the best fit I've got — give it a try?"*.

Some clinics are still on the earlier delivery shape, where the AI produces a multi-task preparation plan instead of a single recommendation. The behaviour profile is identical either way — only the owner-facing delivery differs. **Your dashboard renders whichever shape the session recorded**, so nothing breaks if your clinic moves between them.

## What you see

All of this lands in your **Intake Summaries** page:

- The behaviour profile with its assessments and safety flags
- The accepted recommendation (or the legacy multi-task plan for pre-flag sessions)
- The full conversation transcript (click to expand)
- Module completion progress as the pet parent works through whatever was accepted

On **Today**, the same prep progress appears inline as `N/M modules` next to each appointment row.

## Anonymous vs. linked clients

Pet parents can complete the intake without creating an account — they're shown as "Anonymous" in your client list. Anonymous clients have the same behaviour profile and preparation schedule as linked ones; they just can't receive email reminders or track progress across devices.

If they later create an account (the save-progress prompt shows up at the end of intake), their anonymous session is linked to the new profile and everything carries over.

## Related

- [Reviewing Clients](/docs/reviewing-clients) — see client stages and preparation progress
- [Understanding Intake Summaries](/docs/understanding-intake-summaries) — deep dive into each intake result
- [Today's Schedule](/docs/todays-schedule) — the operational view with the same data inline per row