A fundamental component of conversational AI is Natural Language Understanding (NLU). As human beings, when we speak to someone, we ask questions in our own language using the words we know. A key element of conversational experiences for a user is the ability to ask questions naturally, using vocabulary they understand, rather than a special language or syntax they would need to learn.

When a person asks a question, the one being addressed understands the intent behind the question and acts accordingly. This discovered or assumed intent can result in a response, a comment, or even another question. What makes these interactions complex is matching the intent to what the user actually meant. This is made possible through NLU technology—Natural Language Understanding—which enables a conversational AI application to understand the natural language used by humans, recognize intents in questions, and then act accordingly.

In traditional conversational AI development, developers had to create, train, test, and deploy a separate NLU layer, in addition to the logic, user interface, and other integration layers to manage follow-up behaviors. In Microsoft Copilot Studio, our goal is to democratize conversational AI and make it easier for organizations to get started, manage, deploy, and evolve agents by offering both low-code and pro-code creation experiences.

Copilot Studio comes ready-to-use with generative AI features, such as generative responses (which you saw in the previous task), as well as its own pre-trained NLU capabilities. This means a creator does not need to provision or use an external NLU service and can immediately start creating topics simply by entering trigger phrases into a topic. Additionally, standard and custom entities can be detected during intent recognition or throughout a conversation, by extracting the subject from the user’s sentence, which is then stored and used as a variable.

Furthermore, when creating conversational experiences with multiple topics authored in Copilot Studio, creators should be careful to limit topic overlap to avoid ambiguity scenarios that may arise when the NLU identifies two or more intents with a high score from the user’s question. This forces the NLU to ask the user to clarify the intended intent among the detected topics (“Did you mean X or Y?”). Microsoft Copilot Studio offers features such as topic overlap detection to assist agent creators in their journey, as well as the ability to disable the display of certain topics in the “Did you mean” experience.

To learn more about NLU in Microsoft Copilot Studio, including details about the underlying model and topic overlap management capabilities, see Enable advanced AI features in Microsoft Copilot Studio.

Next unit: Check your knowledge