Exercise 02: Authoring 101

Microsoft Copilot Studio uses natural language understanding (NLU) to interpret what a user is saying and to try to match a user utterance with an existing topic. For example, if the user says, “I tried to use my gift card, but it doesn’t work”, the copilot knows to route the user to the topic that’s related to gift cards not working, even if that exact phrase isn’t listed as a trigger phrase. This concept can also be referred to as intent recognition.

NLU can also help the copilot identify entities in a user’s input. An entity represents a key information you’re trying to identify and extract from a sentence. This can be a phone number, a zip code, a city, a case ID, a person’s name, etc. You can also define your own entities. Your copilot can recognize the relevant information from a user input and then save it for later use. For example, if the user types, “I want fifty red coffee machines” the AI can understand that:

• “Fifty” is the number “50,” and it’s also the number of products to purchase.

• “Red” is a color and is the color of the products to purchase.

• “Coffee machine” refers to the product that the person wants to purchase. In Microsoft Copilot Studio, some subjects (such as numbers and colors) have already been taught to the AI for the copilot. The copilot author needs to specify other subjects, such as the fact that “coffee machine” is a product, as demonstrated in this lab. The two types of entities are:

• Prebuilt – Represent the most-used information, such as age, color, number, and name. Microsoft Copilot Studio can recognize these entities automatically. These entities are shown within the Entities list in Microsoft Copilot Studio.

• Custom – Are the entities that you create. While the prebuilt entities cover commonly used information types, you’ll occasionally need to teach the copilot’s natural language understanding model some domain-specific knowledge. For instance, you might need to create a closed list entity for the list of your product types. Or you may want to configure entities based on a regular expression to recognize specific strings following a naming pattern combing letters, numbers and special characters, for example “SR-00001” for a support request ID. Smart matching and synonyms for closed list entities can make your copilot more intuitive:

• Smart matching – Provides you with the flexibility to let the copilot match the user’s input to an entity that’s a near match but not perfect. Specifically, it lets the copilot autocorrect misspellings and expands the matching logic semantically, such as automatically matching “softball” to “baseball.” You can turn off this feature if you need a match to be perfect, such as if the entity contains model numbers or error codes.

• Synonyms – Allow you to recognize that something that the user has typed matches an option that you provided. For example, for “free shipping”, you can add “complimentary shipping” as a synonym. For “expedited shipping”, you can add “two-day shipping” or “overnight shipping” as synonyms. If the user types any of these phrases, they’re matched appropriately.

The following tasks show you how to create a custom entity and use it within the topic that you already created.