Best Practices for Creating AI Session Note Templates

The success of your AI Session Notes rollout heavily depends on your note templates. Simply toggling AI on in your existing templates without additional setup will likely produce underwhelming results.

This article is for administrators and template owners who have already enabled AI on their session note templates and are ready to go deeper. While getting AI turned on is straightforward, getting it to produce narratives that are accurate, useful, and efficient to review takes intentional template design. This article walks you through the decisions that make the biggest difference: how to structure hidden sections, how to write and refine your prompt instructions, and how to manage changes over time.

If you haven't configured AI on your templates yet, start with Setting Up AI Session Notes first.

When adding AI to your existing templates, consider starting with one session note type at a time while you learn to refine your template and prompts. To help you get started, 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:

While the Motivity templates are a strong starting point, your existing templates have been designed to meet your specific payor requirements — so you may need to adjust accordingly.

There are two aspects of your template you may need to add, change, or enhance:

Both are covered in detail below.

The AI can only write about what's on the note. That sounds obvious, but it has a consequence worth considering: your session data records what the client did, not what happened. A target shows 6 out of 10. The data doesn't say whether the client was fighting a cold, whether a fire drill interrupted the session, whether the first three trials were rough and the last seven were clean, or whether you switched prompt levels halfway through. A narrative built from numbers alone will be nearly useless — it will restate the data in sentences.

Hidden sections are where you collect the context that makes those numbers mean something. Checkboxes for setting events, dropdowns for how the client responded, a field for barriers, a space for the author's observations. The AI reads all of it and uses it to explain the session rather than just report it.

The reason to make them hidden rather than visible is that the finished note shouldn't show its own scaffolding, and it's also redundant. A reviewer or a payer reads the narrative, not the checklist behind it. Hidden sections are read by the AI and used to shape the writing, but they don't appear when the note is printed or exported — so you get the structure without cluttering the record.

There's a second benefit that's easy to miss. Structured fields are faster for your staff than free text. Checking three boxes takes a few seconds; writing the same information takes a few minutes and comes out differently from every author. Giving people well-designed options to choose from will provide greater consistency among team members.

See the screenshot below for an example of how you might use a hidden note section. For more information on how to create these sections, see Hidden/Internal Note Sections.

To learn more about how to write effective AI prompts, see the Setting Up AI Session Notes article.

One underutilized resource: AI assistants like Claude or ChatGPT are highly capable at helping write prompt instructions. Keep a log of changes you make to your prompts, and save a copy of the template before making significant changes so you can track the impact on your notes over time.

Writing a good prompt instruction is itself a writing task, and you can use an AI assistant — Claude, ChatGPT, or similar — to help. Used well, this turns a blank page into a draft you refine. Used carelessly, it produces confident-sounding instructions that don't match your template.

Follow these steps to get the most out of an AI assistant:

1. Give it the material

The AI assistant cannot see your Motivity template, so you'll need to provide the context yourself. Paste in:

2. Be specific about what you're asking for

Broad requests produce broad output. Useful asks include:

3. Ask it to explain its reasoning

Request that it explain each change rather than just handing back a rewrite. This helps you understand what changed and why — and makes it easier to catch anything that doesn't fit your needs.

4. Always verify the output

The assistant doesn't know your payors, your state regulations, your organization's documentation standards, or what your supervisors actually want. It will sometimes produce instructions that sound authoritative but are wrong for your context. Before deploying a prompt it helped write:

Treat the assistant as a fast first drafter with no knowledge of your organization. The clinical judgment stays yours.

Prompt instructions rarely land perfectly on the first attempt — refining them over time is a normal and expected part of the process. The challenge is that without a record of what changed and when, it becomes very difficult to understand why your notes suddenly look different, or to recover a version that was working well.

We recommend two simple habits to keep you in control of your prompt changes:

Keep a change log.

Each time you update a prompt instruction, note what you changed, why you changed it, and the date. This doesn't need to be elaborate — even a shared document or spreadsheet with a few columns works well. Over time, this log becomes a valuable reference for understanding what drives the quality and consistency of your narratives.

Save a copy of the template before making significant changes.

Before making major edits to a template's prompt instructions, duplicate the template and store it as a backup. This gives you a safe version to compare against — or revert to — if the changes don't produce the results you expected. Think of it the same way you'd version-control any important document.

Last Updated: 9/21/26 by Tatum Winslow