2025.8.1

Date de publication : 11 septembre 2025

Conversational Agents in Autopilot for Everyone

Conversational Agents are now in Public Preview. These are agents designed for specific tasks or domains, and you can access them directly in Autopilot for Everyone. The Specialized Autopilots dropdown list displays available Conversational Agents, making them easy to find and use.

Integration of Medical Record Summarization (MRS) tasks from Action Center

You can now handle Medical Record Summarization (MRS) tasks from Action Center in a dedicated interface within Autopilot for Everyone. Opening an MRS task launches a three-pane view:

• Agent summary with citations – Select citations to jump to the source.
• Document viewer – Displays the source file and navigates to the cited page.
• Chat pane – Integrated Autopilot chat with context from the summary.

This interface lets you review, edit, and complete complex summarization tasks in one workspace.

Refer to Configuring Medical Record Summarization tasks for details.

Tabular query support in Context Grounding

Autopilot for Everyone now understands and retrieves information from structured tabular data within your Context Grounding indexes.

Simply ask questions in natural language, then receive precise answers from tables and spreadsheets in your Context Grounding indexes.

To enable the tabular query support, turn on the corresponding toggle in Autopilot for Everyone Context grounding index section.

For more information, refer to Using tabular query with Context Grounding.

Agents, Maestro processes, and API Workflows as tools

Autopilot for Everyone supports agents, Maestro processes, and API Workflows as tools. This enables execution of complex, multi-step automations and direct interaction with external systems from the chat interface.

Be aware that agents need an unattended account assigned to their folder in order to run.

Bring Your Own (BYO) Like-for-Like model support

For enhanced governance and control, organizations can now use their own subscriptions for the LLM models that Autopilot for Everyone uses: Claude 3.5 Sonnet v1 and GPT-4o mini.

LLM call and prompt configurations

As an administrator, you can now fine-tune Autopilot for Everyone core behavior with two new settings:

• Context Grounding search phrase hints, which customizes how queries are rewritten for Context Grounding searches.
• Request classification prompt hints, which runs two sequential LLM calls: first classify with the Request Classification Prompt, then process the result with System Prompt Hints.

Enhanced user experience

We've applied several improvements to how you interact with Autopilot for Everyone:

• When you start a new chat while using a Specialized Autopilot, the interface stays on that Autopilot.
• Your conversation history is now grouped by date (Today, Yesterday, and so on) for easier navigation.
• Autopilot for Everyone in Assistant Web supports high-contrast theme for improved accessibility and readability.
• Autopilot for Everyone in Assistant desktop now provides a shortcut to Settings directly from the Assistant user interface.
• You can delete previous chats in bulk.

Support for split-style automation properties

We added support for split-style automation properties to give you greater flexibility when creating automations.

Pre-response generation limit updates

We reduced the pre-response generation limit from six to three. This improves initial response time and relevance.

Correction d'un bug

Autopilot for Everyone would incorrectly handle required automation parameters with predefined options during initialization.