Introduction
For administrators and template owners
This article is your starting point for configuring AI Session Notes in Motivity Clinical. It covers everything you need to do before your first AI-generated narrative: enabling the feature on a template, understanding permissions, and knowing exactly which fields the AI can see and how they work. If you haven't set up AI on a template yet, start here.
Once setup is complete, see Best Practices for Creating AI Session Note Templates for guidance on designing templates that consistently produce high-quality narratives.
Overview
The AI Session Notes Narrative Generator is a feature that uses artificial intelligence to help clinical staff generate high-quality, objective session narratives faster and with less manual effort. Instead of writing a full narrative from scratch, note authors provide their observations and session details, and the AI transforms that input into a polished clinical note ready for review.
Successfully incorporating this feature involves a few key steps:
Creating quality AI prompts
Setting up the tool in your templates
Testing and reviewing the output of the prompts
Executing an organized rollout across staff
Continuing to monitor output
We have tools to help you with each of these important tasks! This article specifically reviews creating AI prompts and how to set this tool up in your session note templates. Please see the Related Articles section at the end of this article for more information on successfully implementing this tool.
Before You Begin
Check the template type. AI-enabled text boxes are only available on session note type templates. They aren't available on general note or progress report templates. If you don't see the Enable AI option, this is almost always why.
Decide how many AI fields you actually want. There's no limit to how many text boxes can be AI-enabled per template, but each one is a separate generation and a separate narrative the author has to review before signing. Three AI fields on a template means three drafts to review on every note. Start with one — usually the main session narrative — and add more only when you have a reason.
Know what the AI can see. The AI reads the current note: imported data, checkboxes, static text, and whatever the author types in the author prompt. It does not read client profiles, prior sessions, or anything outside the note. If you want something in the narrative, it has to be in the note.
Know that we have a best practices article. We strongly recommend you read through Best Practices for Creating AI Session Note Templates to understand best practices when using the AI Session Note Narrative Generator.
Know that we have template examples. Motivity has created AI Session Note templates you can use as a model or reference. The 97153 template has been field-tested and refined; the other note types have been internally tested and are an excellent starting point.
To find the templates:
Select Organizations
Your organization's name
Select Note/Report Template Libraries
ABA Payor Note/Report Templates
Each AI template is marked with a pink "AI" label
Permissions
Two permissions control access:
Manage Note/Report Templates — required to configure AI on templates
Use Session Note AI (under Notes) — required for note authors to see and use the generation tool
Without Use Session Note AI, the AI text box appears as a standard manual text box. The author sees no Generate button and no AI interface.
Important: Do not turn on the Use Session Note AI permission until you have completed pilot testing of your templates. For more information, see the [Rolling Out AI Session Notes] article (coming soon).
Enabling AI on a Text Box
Navigate to the session note template you want to configure, or create a new one.
Select the text box element where you want the AI to generate content.
Check Enable AI. Two new fields will appear — AI Prompt Instructions and Note Author Prompt — both of which are covered in detail in the next section.
Repeat for any additional text boxes.
The Three Fields — And Who Sees What
Understanding which text is visible to whom, and which text reaches the AI, is the most important thing to get straight before you write anything.
Field | Who Sees It | Editable by Author | Sent to the AI | Appears in PDF download |
AI Prompt Instructions | Administrators only — hidden from note authors | No | Yes, every generation | No |
Note Author Prompt | Note authors | Yes | Yes | No |
Generated Narrative | Note authors, reviewers | Yes | Yes, as context on regeneration | Yes |
Two implications worth keeping in mind:
Authors can't see your AI Prompt Instructions. Keep in mind that note authors are working blind — they can see the narrative the AI produced, but not the instructions that shaped it. If the output ever seems unexpected or confusing, they'll have no way to understand why. Communicate anything authors need to know through training or the Note Author Prompt instead.
Pre-filled Note Author Prompts are sent to the AI. If you input initial text in the Note Author Prompt field (e.g. “Use objective language when describing behaviors.”), that input becomes part of what the AI reads. If you would like to provide guidance to your team members, keep the initial text short, or add longer instructions to a hidden/internal field instead.
Writing AI Prompt Instructions
Your prompt instructions are the foundation of every narrative your AI generates. Before writing anything from scratch, we strongly recommend starting with AI Session Notes: Suggested Prompt Bank — it has ready-to-use prompts for 97153, 97155, 97156, and 97151 that you can copy directly into your template and customize from there.
When building or refining your prompts, focus on these four things:
1. State the Service and Who Delivered It
Tell the AI explicitly what service is being documented and who delivered it. Without this, narratives can attribute activity to the wrong person and will require manual correction on every single note
For a 97153 note: instruct the AI to describe only the direct service delivered — not supervision or protocol modification
For a 97155 note: the reverse applies
2. Tell It What the Note Already Covers
The AI reads the entire note — including static text, imported fields, and checkbox lists. If your template already captures goals, data, and interventions and your prompt asks for the same things, you'll get duplicate content.
Tell the AI what the template already covers and instruct it not to repeat it. Then direct it toward what the narrative is actually for: how the session unfolded, how the client responded, and what that responding looked like. This is the single most effective change for producing shorter, more useful narratives.
3. Ask for Reporting, Not Judgment
The AI should report what happened — not draw clinical conclusions. Consider adding the following instructions to your prompt to keep the output objective:
Don't assert that a client is progressing, regressing, or generalizing a skill — report the data and the clinician's selection
Don't compare performance to a baseline unless the comparison was entered
Don't supply a rationale for a clinical decision — report the clinician's rationale as entered, or describe the decision without one
Don't state that a barrier caused a change in performance
This matters most on 97155 and 97151 notes, where the prompt is otherwise asking for interpretation.
4. Set Your Organization's Standards — Once
Details like word count, how narratives should open, preferred phrasing, and how to handle quoted phrases belong here — not left to individual note authors. If these aren't defined upfront, authors add them manually on every note, inconsistently.
What to Avoid When Creating AI Prompts
Avoid restating content the template already captures.
e.g., "State the start and end time of the session" — this is already captured in the template.
Avoid writing instructions about a single client or session.
e.g., "Include details about Andy's response to new interventions" — this is too specific and won't apply to other notes.
Avoid asking for things the AI can't do.
e.g., "Compare the current data found in the note to data from yesterday's sessions" — the AI can't access prior sessions, look up a treatment plan, or make a treatment decision.
As you can see, there are many important considerations when creating your AI prompts. Check out AI Session Notes: Suggested Prompt Bank for a ready-to-use starting point.
Using Hidden Sections to Inform the AI
Hidden or internal note sections are read by the AI along with the rest of the note, but don't appear when the note is exported or printed. That makes them useful for structured input that shapes the narrative without appearing twice in the finished record.
A common pattern: put the setting events checkboxes, barriers selections, and response-to-intervention selections in a hidden section, and let the narrative field carry the prose that results. The author's selections inform the draft; the exported note shows the narrative rather than the raw checkboxes.
Two things to get right:
Tell the AI how to treat the hidden content. It should describe the author's selections as observations from the session, rather than as clinical conclusions — and where a selection conflicts with the collected data, it should report the data.
Avoid having hidden and visible sections cover the same information. This creates duplicate content and may result in long, redundant narratives.
For more on how to create these sections, see Hidden/Internal Note Sections.
Related Articles
AI Session Notes: Suggested Prompt Bank — Ready-to-use prompt templates organized by session type
Best Practices for Creating AI Session Note Templates — Guidance on structuring templates for the best AI output
[Rolling Out AI Session Notes] — How to stage and execute a rollout across your staff (coming soon)
[Using AI Session Notes as an RBT or BT] — A guide for note authors using the generation tool (coming soon)
[Reviewing AI-Assisted Notes as a Supervisor] — How supervisors should approach reviewing AI-generated content (coming soon)
Hidden/Internal Note Sections — How to create and configure hidden sections in your templates
Last Updated: 9/21/26 by Tatum Winslow