Power Platform Copilot Studio Topics Entities

A chat assistant succeeds when its conversations feel natural. Topics define what the assistant can discuss, and entities help it pick out key details such as dates, emails, and product names.

What a Topic Is

A topic is one conversation path. Each topic has trigger phrases that start it and a chain of nodes that guide the exchange. A topic called Store Hours reacts to phrases like "when do you open" and replies with a schedule.

User types: "What time do you open on Sunday?"
        |
        v
Copilot matches trigger phrases --> Topic: Store Hours
        |
        v
[Message] "We open at 10 AM on Sundays."
        |
        v
[Question] "Need directions too?"  --> Yes / No

Trigger Phrases

Add five to ten short phrases that people might type. Use different wording rather than small variations of one sentence. Copilot Studio uses language understanding, so the assistant recognizes similar phrases without an exact match.

Node Types

Node	Purpose
Send a message	Displays text, images, or cards
Ask a question	Collects an answer and saves it in a variable
Condition	Branches based on a value
Set a variable	Stores or calculates a value
Call an action	Runs a Power Automate flow or connector
Redirect	Jumps to another topic
End conversation	Closes the chat and can request feedback

System Topics

Every assistant ships with system topics. Greeting welcomes users. Escalate hands the chat to a human agent. Start Over resets the conversation. Fallback catches messages the assistant cannot match. Review these topics early and adjust the wording to match your brand voice.

Entities

An entity is a category of information the assistant can recognize inside a sentence. When a user types "Book a table for 4 on Friday", entities capture the number 4 and the date Friday without extra questions.

Prebuilt Entities

Prebuilt entities include number, date and time, email, phone number, money, city, and person name. Pick the entity type on a Question node and the assistant extracts the right value automatically.

Closed List Entities

A closed list holds fixed choices with synonyms. A Size entity might hold Small, Medium, and Large, and the word "regular" maps to Medium.

Entity: PizzaSize
  Small   <-- "mini", "personal"
  Medium  <-- "regular", "normal"
  Large   <-- "big", "family"

Regular Expression Entities

A regular expression entity matches a pattern such as an order number like ORD-12345. It suits structured codes that follow strict rules.

Variables

Variables store values during a conversation. Topic variables live inside one topic. Global variables travel across topics. System variables hold facts such as the user name or the conversation start time. Reference a variable in a message with curly braces in the editor, and the assistant fills in the value at run time.

Layman Example: A Hotel Front Desk

A front desk clerk listens for key words. "Checkout" sends a guest to the billing counter. "Taxi" sends a guest to the transport desk. Each request is a topic. While listening, the clerk notes the room number and the time without asking twice. Entities perform that same listening skill for the assistant.

Design Tips for Better Conversations

• Keep messages short, one idea per bubble.
• Offer buttons for common choices so users tap instead of typing.
• Confirm important details before acting on them.
• Give users an exit to a human agent at any time.
• Review the unmatched messages in analytics and create new topics for them.

Testing Topics

Use the test panel on the right side of the authoring canvas. Type sample phrases, watch which topic fires, and track the path highlighted on the canvas. Change trigger phrases when the wrong topic answers.

Key Takeaways

Topics organize the conversation, trigger phrases start them, and entities capture details. Clear topics and smart entities create assistants that feel helpful instead of robotic.