Synthesized response in Microsoft Teams conversations

Conversational Integration with Microsoft Teams is designed to provide new and rich user experience with synthesized responses using Large Language Models (LLM) based results summarized in a single response for users to see information in a conversational way, and avoid browsing through carousels.

With synthesized responses implemented in Microsoft Teams conversations, the user experience is increased in several ways for the response to contain the most relevant items and information, such as:

• Returning multiple Genius Results and topics in carousel format.
• Providing unified search across topics and catalog items.
• Providing a multi-Knowledge Base Q&A pipeline that enables multiple snippets from multiple Knowledge Base articles to be passed to the LLM as the context for answer generation.

The overall synthesized response helps users experience a conversational flow that understands query intent, searches across records of various types, and summarizes results in a unified, easy-to-consume response.

When you enable the synthesized response picker on Microsoft Teams, users who start a conversation with a Virtual Agent can view a synthesized response picker in response to a query. The response contains catalog items and topics followed by the citation links. For example, if you input a command laptop in your conversations, you’ll see the responses in a synthesized format.

Mid-topic switching is also possible during a conversation with synthesized responses and user can continue with their original request or switch the conversation. When the user selects the View other options button they get the list of available Knowledge Base articles and catalogs, which the user can select and go through the details.