Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
[This article is prerelease documentation and is subject to change.]
Customer Intent Agent discovers intent from historical data and maintains an up-to-date intent library, along with attributes to support intent determination, interview, and fulfillment, leading to higher deflection rates.
Note
Copilot Studio bot is renamed as Copilot agent (agent or AI agent). Human agent is now renamed as customer service representative (service representative or representative). You might come across references to the old and new terms while we update the product UI, documentation, and training content.
Customer Intent Agent uses the determined intent and interview responses to improve the quality of generated answers for both Copilot agents and customer service representatives (service representatives or representatives).
You can configure Copilot agents with Customer Intent Agent to provide updated solutions for customer issues.
Important
- This is a preview feature.
- Preview features aren’t meant for production use and might have restricted functionality. These features are subject to supplemental terms of use, and are available before an official release so that customers can get early access and provide feedback.
Prerequisites
Before you proceed, make sure you:
- Enable Customer Intent Agent.
- Set up intent discovery and add intents to the intent library.
- To use intents with Copilot agents, enable the Use in AI Agents checkbox for that specific intent.
- Connect your Copilot agent to knowledge sources and a workstream in omnichannel.
- Assign the Environment Maker role to makers.
- Get the administrator role to add connection references.
How it works
- An administrator sets up intent, uses historical data to build the intent library, and enables the library for Copilot agents and representatives.
- The Copilot agent detects intent from the customer, asks follow-up questions related to the intent, and provides a solution based on customer-provided information from the connected Copilot agent knowledge source.
- If the issue is escalated, it is transferred to representatives. The Customer Intent Agent persists the intent and the interview questions generated by the Copilot agents. The service representative goes through unanswered questions with customer and gets an updated solution.
- The application also uses this conversation to detect any updates and discover new intent attributes that can be used in future interviews.
- In the next customer interaction with a Copilot agent, the agent can deal with the customer issue with the newly generated knowledge and intent, and provide a tailored response without escalating to any representative.
Set up Copilot agents for intent management
- In Copilot Service admin center, navigate to the Customer Intent Agent (preview) page > Intent-based suggestions section.
- Select Enable for chatbots and then select Manage. The Bots page appears.
- Select the agent connected to your omnichannel environment for agent transfers.
- Select Add intent-based features, which takes you to Manage component collections (preview) in Microsoft Copilot Studio.
- In Manage component collections (preview), search for and select Intent-based suggestions (preview).
- Select Add agent. You can also disconnect a Copilot agent to stop agent transfers for that Copilot agent by selecting Remove from agent.
- Select Publish. Once added, a series of Intent-based suggestions topics are added to your agent.
Add connection references
Add connection references for each new organization once, even if you add multiple agents to the organization. Learn more in Add connection references to a solution.
- Sign in to make.powerapps.com and select your environment.
- Go to Solutions > Default Solution > Objects > Connection References.
- Search for the IA.Cr.IAPluginDVConnector connection reference.
- Create a connection to Microsoft Dataverse, and ensure that the user has administrator privileges.
- Go to Microsoft Copilot Studio, complete the user authentication setup, and publish the agent.
User authentication
By default, users must be authenticated to access the Microsoft Dataverse instance to connect to the intent library and get an intent response.
You can switch to Copilot author authentication if users connecting to Copilot Studio don't have access to the Dataverse instance.
- In Microsoft Copilot Studio, open the Intent-based suggestions – Plugin topic.
- Select the ellipsis next to Connect action > Properties.
- Select the End user authentication dropdown and change it from User authentication to Copilot author authentication.
Connect to intent-based suggestions
To use Customer Intent Agent, you need to connect your Copilot agent to a knowledge source in Microsoft Copilot Studio. Learn more in Add knowledge to an agent.
- In your existing topic flow, where you want to use Customer Intent Agent, add the following node:
- Topic management > Go to another topic > Intent-based suggestions – Main.
- For a default agent, add this node to either the Conversational boosting or Fallback system topic that triggers from the On Unknown Intent topic.
- Add another node: Advanced > Create generative answers, and select SolutionQuery for Input. Adding a SolutionQuery helps generate information from the determined intent and answered questions to search knowledge sources for a solution response. Ensure you edit your data sources and check all the sources you want to search against for an answer.
- Add another node: Add a condition, and set Answer variable to is not Blank.
- To add logic to complete the conversation flow after providing an answer, add another node: Topic management > Go to another topic > End current topic.
Connect to your knowledge base
Connect the knowledge articles in your Dynamics 365 instance to your Copilot agent. Learn more in Integrate knowledge management in Dynamics 365 with Copilot agents.
Make sure that you add the new flow you created in Power Automate, called Generate answer from Dataverse knowledge articles for Power Virtual Agent. Follow the steps provided in the Connect to intent-based suggestions section, and replace the Create generative answers node with the new flow.
Intent-based suggestions output variables
This section lists the output variables from Customer Intent Agent that provide information about the status of intent determination and interview. Depending on your use case, you can customize these variables for specific situations.
| Variable | Scenario |
|---|---|
| Global.IntentInterviewComplete | If set to true, ensures that the intent is determined and all the interview questions were answered. |
| Global.IntentInterviewIncomplete | If set to true, indicates that the intent is determined but not all interview questions were answered. You can still query knowledge based on the information gathered, but you might choose a different behavior. |
| Global.IntentUndetected | If set to true, indicates that a known intent from the intent library wasn't detected, or was detected but rejected by the customer. |
| Global.IntentDetectedWithoutAttributes | If set to true, indicates that an intent has been successfully detected, but there are no associated interview questions to ask the user. A solutionQuery is still generated and available for use. |
| Global.IntentError | If set to true, indicates that there was an error in detecting intent. The issue could be transitory like a time-out, but if it's consistent, we recommend that you disable intent-based suggestions and contact support. |
Variables in the Intent-based suggestions - Configuration topic
The Intent-based suggestions - Configuration is a noncustomizable topic that contains a list of variables that can be overwritten to provide the desired behavior. Here's a list of variables in the topic.
| Variable | Default | Description |
|---|---|---|
| Global.EnableIntentConfirmation | False | If enabled, agent asks the customer to confirm the intent before moving on to the intent attributes questions. |
| Global.EnableIntentRephrase | False | If enabled, and if intent can't be determined or intent isn't confirmed by customer, customers are asked to rephrase their issue. |
| Global.RephraseCount | 1 | Determines how many times customers are asked to rephrase the issue. |
| Global.MaxRepetitionCount | 2 | Sets the maximum number of times the agent repeats an interview question with the same attributes. If the user doesn't provide a valid response on the first attempt, the agent asks the question again. If the user still doesn't answer, the agent moves to the next question. |
| Global.EnableAskAllQuestions | False | If enabled, asks all intent-attribute questions before ending the loop, regardless of whether the customer answers the question in the next turn. If disabled, as soon as the customer doesn’t answer an intent attribute question, ends the loop (a solution query is still generated and generated answers are still initiated). |
Override default variable
To override any of the variables for the intent-based suggestions:
- Go to the System > Conversation start topic.
- Select Add node (+) and Variable management > Set a variable value.
- Select the variable that you want to change and update the value.
- Save and publish the agent.
Learn more in Work with variables.
Updates to Customer Intent Agent
Customer Intent Agents might receive periodic updates that update the corresponding component collection. To get access and receive these updates, you need to manually publish your Copilot agent.