Connect an AI assistant
Use ChatGPT or Claude to run your practice, on your own subscription and under your own login.
You can connect the AI app you already pay for - ChatGPT, Claude, or another one that supports connectors - to your own Wattle data. Once it is connected you can ask it things in plain English, including by talking to it:
- "What's on tomorrow?"
- "Book Berlin in with me on Friday at 9:30, half an hour, pentosan injection and review."
- "Which cats over ten haven't been in for a year?"
- "How much is outstanding, and who owes the most?"
- "Start a visit for Mooshka and put the consult and the cerenia on the bill."
- "Mrs Chen paid $120 cash, put it against her oldest invoice."
- "Email the Hendersons their estimate for the dental."
It runs on your AI subscription, not on ours. Wattle does the doing; your app does the thinking. There is nothing extra to pay us and no per-question cost.
Check what it does. An assistant can be confidently wrong. You are responsible for what it writes into your records, exactly as you would be if you had typed it yourself. Everything it does is logged against your name.
Connecting it
Settings → Integrations → AI assistant. You need Connect AI apps and integrations, which the clinic owner and practice managers have by default.
This is being trialled with a small number of clinics first. If you can't see the card, it isn't switched on for your login yet.
- Name it something you will recognise later, like "ChatGPT on my laptop".
- Choose its permissions (see the table below). Start with Read only if you are unsure - you can widen it later without reconnecting.
- Create connection. The link is copied to your clipboard automatically.
- In ChatGPT or Claude, add a connector and paste the link.
- Sign in to Wattle when it asks, and choose Allow.
That is the whole setup. There is no API key to generate, no developer account, and no separate password.
The card says Waiting for the link until the AI app actually connects for the first time. That is normal, not a fault.
Permissions
You choose this when you create the connection, and you can change it any time afterwards from the same screen. Four one-click options:
| Option | What it grants |
|---|---|
| Read only | It can look at everything you can look at. It cannot change anything. |
| No financial access | It can work with the schedule, records and clinical notes. Nothing at all on billing, pricing or reports. |
| Day to day | Schedule, records and clinical notes, plus read-only billing. No settings, no staff. |
| Everything I can do | Everything, up to your own permissions. Not the default, and not the one to pick first. |
Below the four options is a row per area if you want to set them individually.
Things it can never do, whatever you choose
- Write to the controlled-drug register. It can be given permission to read the register, and that is never part of the four options above - it is a separate tick you have to make yourself. It can never write one. A chat transcript is not the register.
- Remove a charge that is already in the register. Taking one off a bill would write a correction into the register, so it refuses and tells you which drug and how much. A vet does that in Wattle, by hand. Every controlled-drug entry is made by a person, always.
- Give itself more access. It cannot change permissions, roles, or its own connection.
- Delete a consult, a staff member, or your clinic.
In Western Australia, South Australia and Tasmania the controlled-drug register returns nothing at all, because in those states the register you keep has to be one a regulator has approved.
Before it changes anything
When you ask it to do something with a consequence, it stops and tells you what that consequence is before anything happens. You will see something like:
This appointment has a consult linked. Deleting will unlink the consult. The clinical notes stay on the patient file.
Nothing has happened at that point. You say yes or no, and only then does it go ahead.
This is Wattle asking, not the AI app being polite. Both ask, and that is deliberate: AI apps let you tick "always allow" for an action, and if that were the only check you would stop being told anything.
How much it asks
Settings → Integrations → AI assistant, under Before it writes anything:
| Setting | What it means |
|---|---|
| Ask before anything major | Deletions, sends, invoice reversals, staff changes and edits to your catalogue or clinic details are read back to you first. The default. |
| Ask about money over a limit | The same, but the money ones only stop at or over the figure you set. |
| Ask only when records are attached | It gets on with it, and everything is logged. |
Routine work never stops to ask, whichever you pick. Booking, writing a note, recording a weight, adding a charge: those just happen. A setting that stopped for those would be clicked through without being read within a day.
Whichever you pick, anything with records attached to it is always read back to you first. That is about knowing what you are agreeing to, not about trust.
It can never do more than you can
Whatever you allow here, the assistant is still limited by your own permissions, and those are checked every single time it asks for something. If a colleague's account cannot see billing, an assistant connected to their account cannot either, no matter which option they picked.
Change someone's permissions in Settings → Roles & Permissions and it takes effect on their assistant's very next question. There is nothing to reconnect.
Turning one off
Settings → Integrations → AI assistant → Revoke. It stops working immediately.
Revoking a connection does not sign you out of Wattle, and signing out of Wattle does not revoke the connection - they are deliberately separate. A connection that has not been used for 90 days expires on its own.
Where to see what it did
Every single question it asks is written down: what it asked for, when, how many records came back, and whether it was refused. Owners and practice managers can see the whole clinic's; everyone can see their own.
Things worth knowing
It cannot take a payment. When you tell it someone paid, it writes that down, exactly as you would. No card is charged and no terminal does anything.
Changing a price never changes an old invoice. Every charge kept the price it was sold at. If you ask it to "update all the prices" it changes the catalogue and tells you so.
It reads your data in the clinic's time zone. "Tomorrow" means tomorrow where the clinic is, not where the assistant thinks it is.
It will tell you when it cannot see everything. If a list is longer than it fetched, it says so rather than answering as though it had the lot.
Some text in your records was written by someone outside the clinic - a booking enquiry, or notes brought across from your old system. Wattle marks that text so the assistant treats it as something to report to you, never as an instruction to follow.
Your data goes to the AI provider you chose. That is the point of connecting it, and it is a disclosure you are making deliberately. Consumer plans at some providers may allow your content to be used for training theirs; check your own plan if that matters to you.