Summary

Ask Grounded Scribe is the AI clinical assistant built into every session and client page. The button opens a dialog where you can ask documentation questions, summarise progress, draft correspondence, look up MBS items, or generate personalised handouts.

This guide focuses on how to use it well — for the full feature pitch and a side-by-side capability list, see the feature page. For tier limits, see pricing.

The two things to know up-front:

1. Where you ask from matters — the assistant sees different context on a session page vs a client overview page.

2. It refuses to invent detail — every clinical claim is traceable to a specific session, with the date cited inline.

Where to Ask From (and Why It Matters)

The Ask Grounded Scribe button appears in two places. The context it sees depends on which one you click — and that materially changes the kind of answer you'll get.

From inside an open session (you're viewing or recording a specific appointment): the assistant sees the current session's transcript and generated note in full, plus a light backdrop of the client's recent history.

> Best for: *"Help me write today's note", "What MBS item fits what we just did?", "Did I cover the homework review?"*

From the client overview page (the client's main profile): the assistant sees a comprehensive longitudinal view — finalised notes from up to 50 past sessions, recent assessment scores, the active treatment plan, and presenting issues.

> Best for: *"Summarise her progress over the past year", "Draft a referral letter covering treatment to date", "What themes have come up most often?", "How has she changed since the start of treatment?"*

A small badge — *"Analysing N past sessions..."* — appears under the loading spinner when the assistant is reading across a longer history. If you ask a longitudinal question from a session page (where only a light backdrop is loaded), the answer will feel thin; switch to the client overview and ask again.

How It Stays Grounded

The grounding protocol is strict and deliberate. When the assistant draws on past sessions, it follows five rules:

Every clinical claim must trace to a specific session in the client's record.
Dates are cited inline — for example, *"she discontinued medication [12 Sep 2025]"*. You can verify any claim in seconds by looking up that date.
If a fact isn't in the notes, it says so — *"I don't see X in the available history"* rather than inferring or filling gaps.
Most recent sessions are weighted as the current clinical picture; older sessions are background.
If a client has more sessions than fit in working memory, the assistant tells you how many were omitted (*"[N earlier sessions omitted due to length budget]"*) rather than silently truncating.

This is conservative on purpose. The aim is that you can trust what it tells you, and you can always verify.

Generating Client Handouts (Core and Above)

On Core, Pro, and Max plans, you can also ask the assistant to generate personalised handouts for clients:

*"Create a grounding-techniques handout for her"*
*"Write a session recap she can take home"*
*"Generate a homework sheet on values clarification"*
*"Make a kid-friendly worksheet for managing big feelings"*

The handout opens in an editable preview. From there you can download as PDF, save to the client's file, assign as homework, or save the template for reuse on future clients.

Handouts draw on the same client context the rest of the assistant uses — the topics you've worked on, the techniques discussed, age-appropriate language for child clients — so they feel personalised rather than generic.

Safety-plan exception: the assistant will not generate safety plans for suicide or self-harm. Crisis steps are too important to risk AI inaccuracy. If you ask for one, you'll be redirected to the Stanley & Brown Safety Planning Intervention and the Australian crisis lines (Lifeline 13 11 14, Beyond Blue 1300 22 4636, 13YARN 13 92 76, Kids Helpline 1800 55 1800).

What It Will Refuse to Do

Ask Grounded Scribe is not a medical device and does not diagnose. By design, it will not:

Recommend specific medications or doses
Make diagnostic statements it's not asked to summarise (it can describe diagnostic codes already on a client's record, but won't invent new ones)
Generate safety plans (see above)
Provide clinical management recommendations based on billing codes alone

These aren't limitations — they're guardrails. You remain the responsible clinician for every decision. The assistant's outputs are drafts and synthesis, never final clinical judgement.

How a "Conversation" Works

A conversation is a single dialog with up to 10 follow-up messages. Only the first message of a new conversation counts against your daily limit; follow-ups within the same conversation don't deduct.

That means refining your question is cheap: *"now make it shorter", "what about her medication history?", "rewrite this as a referral letter"* — all free against the cap.

Tier limits are listed on the pricing page. Limits reset at midnight AEST.

Prompting Tips That Help

Be specific about scope. *"Summarise the last four sessions"* beats *"what's happening with her"*. The narrower the scope, the more focused the answer.

Ask for the format you want. *"In dot points", "as a referral letter", "in two paragraphs"* — the assistant respects formatting requests reliably.

Ask follow-ups rather than starting over. Follow-ups keep the same context and don't cost a conversation, so iterating is faster and cheaper than starting fresh.

Use it for the awkward synthesis tasks. Reading across many sessions, drafting clinical correspondence, compressing 12 months of work into a paragraph — these are where it saves the most time.

Cite back when verifying. When the assistant gives you a date in brackets, treat that as a checkable claim. If it cites a date and your gut says "that doesn't sound right", look up the session.

Privacy and Data Handling

Conversations stay inside your Grounded Scribe account. No client information is shared with third parties for AI training. Every conversation is recorded in your audit log (showing what was asked, when, and which sessions informed the answer) for compliance and traceability.

The service runs on Australian-hosted infrastructure and is operated under the Australian Privacy Principles (Privacy Act 1988). For solo practitioners and small clinics this means your client data does not leave Australia in the course of normal operation.

Troubleshooting

"You've reached your daily limit" — Conversation caps reset at midnight AEST. To lift the cap, upgrade from your billing settings.

The summary mentions a session that doesn't look right — Click through to the cited date and check the original note. If the assistant is misattributing detail across sessions, send us feedback via the in-app feedback button so we can investigate.

It says *"I don't see X in the available history"* for something I know is documented — The assistant only reads finalised notes (not drafts, not transcripts still being processed). Check the session's status; if the note is finalised and the assistant still can't see it, contact support.

Responses feel generic or thin — You're probably asking a longitudinal question from a session page. Switch to the client overview and ask again; the assistant has the full record from there.

The badge says "Analysing 12 past sessions" but I've had 30 with this client — Only finalised notes count toward the longitudinal context. Sessions where the note is still in draft, or where you recorded but never generated a note, are excluded. Generate the missing notes and ask again.

---

Get started with Grounded Scribe — Ask Grounded Scribe is available on every plan, including the Free tier.

---

Using Ask Grounded Scribe Effectively