Admins can now choose which voice their assistant uses in Voice Mode. This is especially useful for multilingual setups — no more relying on a hardcoded default.
Flows can now be set to apply only to specific courses. This means you can create different assistant behaviors for different subjects — a math course can feel very different to a writing course.

Students and staff can now raise support tickets directly in Microsoft Dynamics 365 through the assistant.
The assistant can now create IT help desk tickets in Ivanti, making it easier to route support requests without leaving the chat.
📄 How do I integrate LearnWise with Ivanti ISTM?
When a connected integration stops working, admins now get a clear alert in the admin panel so the issue can be spotted and addressed quickly.
The assistant can now pull in transcripts from videos hosted in the Panopto LTI, even when those videos aren't embedded directly in a course.
When adding a website as a knowledge source, you can now have the assistant wait a few seconds after each page loads before reading it. Use this for sites whose content loads dynamically (via JavaScript) — if imported pages come back blank or incomplete, turn it on. For normal static pages, leave it at zero.
Suggested questions are now generated faster and with better quality, with automatic fallbacks to keep things running smoothly.
If the chat drops connection while the assistant is generating a long response, it now recovers automatically instead of appearing to time out with no answer shown.
Notifications feel snappier and more responsive, even on slower connections.
When setting up Zendesk escalations, fields that can't actually be edited are no longer shown as options — reducing confusion during configuration.
Word files that previously failed to process due to formatting issues will now fall back to an alternative parser, improving reliability of document uploads.
Improved reliability when fetching data from Canvas, reducing the chance of authentication errors under heavy use.
Fixed an issue where numbers in AI responses were being incorrectly interpreted as mathematical expressions, causing them to display in an unexpected format.