When messaging patients, Medbridge can suggest AI-generated reply options to help you respond efficiently. This article explains how these suggestions work, what the safety guardrails do, and how to share feedback on suggestion quality.
Important
AI suggestions are a starting point, not a clinical decision. The disclaimer "This content is generated by artificial intelligence and is not a substitute for clinical judgment" is displayed in the messaging panel at all times. Always review and edit suggestions before sending.
How AI suggestions work
When you open a message thread with a patient from the messaging panel, an AI suggestions section appears below the conversation.
Based on the patient's most recent message, the system generates two response options:
- Friendly - an empathetic, conversational tone appropriate for emotional check-ins or general support.
- Clinical - a more direct, clinically-focused tone appropriate for program guidance and follow-up.
To use a suggestion, click on it. The text will populate in the message compose field, where you can edit it before sending.
Note
Suggestions are generated per episode. When you open a message thread, you'll first select which episode the conversation is related to before suggestions are shown.
Safety guardrails
The AI is designed to keep conversations clinically focused. If a patient sends a message that is off-topic or unrelated to their care program, the suggested responses will gently redirect them back to their assigned exercises or program activities rather than engaging with the off-topic content.
For example, if a patient mentions something personal or asks about an activity outside their program, the AI suggestion will acknowledge the patient while steering the response toward their care goals and it will not generate replies that veer into unrelated topics.
Guardrails will also kick in if the patient seems to be in a high risk situation such as falling during an exercise:
When AI suggestions are unavailable
In some cases, the AI suggestions panel will display the message: "AI suggestions are not available for this patient. Prompt the patient to log in and accept the terms."
This happens when the patient has not yet logged in to the Medbridge app or has not accepted the terms required to enable AI features. In these situations, you can still message the patient manually by simply typing your message in the compose field and send as usual. To resolve this, ask your patient to log in to the app and complete the terms acceptance step.
Giving feedback on suggestions
If a suggestion isn't accurate, relevant, or appropriately toned, you can share feedback to help improve the feature. To do so, click Give feedback or the thumbs up/thumbs down arrow in the AI suggestions panel.
A dialog will appear with a text field where you can describe what you liked or disliked about the suggested messages (up to 1,000 characters). Click Submit when finished, or Close to exit without submitting.
Tip
Specific feedback such as noting that a suggestion was too casual for the clinical context, or that it missed important context from the patient's message is the most useful for improving future suggestions.
Hiding AI suggestions
If you prefer not to see AI suggestions for a given conversation, click Hide in the suggestions panel.
This will collapse the section for that session. You can show it again at any time from the same area.