Use cases are the mechanism by which zero-training AI agents understand what a customer is asking about and connect them with the right dialogue.

Use cases leverage generative AI to power AI agents without the time-consuming activity of training expressions. This zero-training approach simplifies setup and maintenance while enhancing the AI agent's understanding and performance.

This article contains the following topics:

  • Creating a use case
  • Categorizing a use case
  • Deactivating a use case
  • Deleting a use case
  • Resolving conflicts between use cases
  • Restoring a previous version of a use case
  • Best practices for creating use cases

Creating a use case

An AI agent can start supporting customers with as little as one use case.
Tip: Before creating your first use case, see Best practices for creating use cases.

To create a use case

  1. In the top-right corner, use the AI agent drop-down field to select the AI agent you want to configure use cases for.
  2. In the main menu on the left, select Content > Use cases.
  3. Click Add Use Case.
  4. Fill out the following fields:
    • Name: Enter a short, descriptive name for the use case. For example, “Order status.”
    • Visitor request reason: Enter a short description that identifies why a customer is reaching out. For example, “Customer is asking for the status of their order.”
  5. Click Confirm.

Categorizing a use case

After you create a use case, you can categorize it. Categorizing makes use case management easier, but it doesn’t affect how AI agents connect use cases with a customer’s message (only use case names and descriptions do that).

To categorize a use case

  1. In the top-right corner, use the AI agent drop-down field to select the AI agent you want to configure use cases for.
  2. In the main menu on the left, select Content > Use cases.
  3. Click the use case you want to categorize.

    The use case details page opens.

  4. In the Category field, enter a free-text category for the use case.
  5. Click Save.

On the Use cases tab, a use case’s category appears in the Category column. You can also filter use cases using the Category drop-down at the top of the page.

Deactivating a use case

Deactivating a use case allows you to retain its configuration (nothing is deleted), but your AI agent won’t use it to connect customers with a dialogue until you reactivate the use case.

To deactivate a use case

  1. In the top-right corner, use the AI agent drop-down field to select the AI agent you’re configuring use cases for.
  2. In the main menu on the left, select Content > Use cases.
  3. From the list, find the use case you want to deactivate, click the options () menu, and select Deactivate.

    Your use case is immediately deactivated. You can reactivate it by clicking the options () menu and selecting Activate.

Deleting a use case

If you’re confident that you don’t need a use case anymore, you can delete it. As with deactivating a use case, deleting a use case ensures that your AI agent will no longer use that use case to connect customers with a dialogue. Deleting a use case is permanent and cannot be undone.
Important: Deleting a use case also removes the associated dialogues, intent, and expression data, if there were any. This means that you won’t be able to revert the use case back to an expression, if that were to become necessary. In addition, names of deleted use cases no longer appear in reports.

To delete a use case

  1. In the top-right corner, use the AI agent drop-down field to select the AI agent you’re configuring use cases for.
  2. In the main menu on the left, select Content > Use cases.
  3. From the list, find the use case you want to delete, click the options () menu, and select Delete. Your use case is immediately deleted.

Resolving conflicts between use cases

Overlapping or duplicate use cases can confuse the AI model, leading to poor performance. The Conflict Resolution tool helps you identify and fix these overlaps, ensuring the AI model functions smoothly and provides a good experience.

There are two types of use case conflicts:

  • Confusion: Two or more use cases partially overlap, causing the AI model to conflate them and trigger the wrong dialogue.
  • Duplication: Two or more use cases fully overlap, causing the AI model to interpret them as identical.

Both types of conflicts can be resolved with the Conflict Resolution tool.

To resolve a conflict between use cases

  1. In the top-right corner, use the AI agent drop-down field to select the AI agent you’re configuring use cases for.
  2. In the main menu on the left, select Content > Use cases.
  3. From the list, find a use case that has an entry in the Conflict column and click Review Confusion or Review Duplication (depending on what kind of conflict exists).

    Tip: When you hover over a use case with an entry in the Conflict column, other conflicting use cases are highlighted.
  4. In the window that appears, review the details of the conflict and the suggested edits, if available.

    • If the use cases overlap or are wrongly identified as duplicates of each other, resolve the conflict by modifying the name and description of one or both of the use cases. You can:
      • Accept the proposed changes as is by clicking Accept Suggestions.
      • Click the pencil icon to modify the proposed changes before accepting them.
      • Deselect the use case you don’t want to modify by toggling the Selected button to Not Selected.
    • If the use cases actually are duplicates of each other, resolve the conflict by clicking Cancel and deactivating the use case you no longer need.

Restoring a previous version of a use case

Any updates made to a use case’s name or description are tracked and can be rolled back if necessary, making it safer to test changes or recover from errors.

To restore a previous version of a use case

  1. In the top-right corner, use the AI agent drop-down field to select the AI agent you’re configuring use cases for.
  2. In the main menu on the left, select Content > Use cases.
  3. From the list, click the name of the use case you want to restore a previous version of.
  4. In either the Name or Visitor request reason field, click the Show version history () icon.

    The Version History panel appears on the right, showing you the changes made to the use case’s name or description. The panel shows which user made the changes and when, and which version of the use case is the current version. You can expand the version history entries to see the content of the Name or Visitor request reason field for each saved change.

  5. (Optional) Click the Type drop-down field, select the types of use case changes you want to see (name, description, or both), and click Apply.
  6. Find the version of the use case name or description you want to restore and click the Restore () icon.

  7. Click Save.

Best practices for creating use cases

A use case’s name and description provide the necessary context for an AI agent to understand the customer’s messages and connect them with the right dialogues to help them. As such, it’s important to write clear and specific use case names and descriptions in order to get the best performance possible from your AI agent.

Best practices for use case names

When creating use case names, keep the following best practices in mind:

  • Write use case names in English.
  • Keep names short, ideally 3-5 words.
    • Do this: Username or password change
    • Don’t do this: How do I change my username or password?
  • Choose names that are as clear as possible.
  • Avoid any punctuation except commas.
    • Do this: Issues with Klarna
    • Don’t do this: Issues - Klarna

Best practices for use case descriptions

When creating use case descriptions, keep the following best practices in mind:

  • Write use case descriptions in English.
  • Start all of your use cases in a consistent way (for example, “Customer wants to…”).
  • Avoid any punctuation except commas.
    • Do this: Customer is looking for advice on a product or service
    • Don’t do this: Product / service advice
  • Avoid including keywords in the description.
    • Do this: Customer wants to schedule a collection for their return
    • Don’t do this: Customer wants to schedule a collection for their return. Keywords: pick-up, collection
  • Write descriptions that are concise (around 120 characters, including spaces) yet detailed enough to connect the customer to the right dialogue.
  • Avoid acronyms, or write them acronyms to provide additional context.
    • Do this: Customer requests do it yourself instructions
    • Don’t do this: Customer requests DIY instructions
  • Create distinct use cases and ensure they don’t overlap. Overlapping use cases cause confusion and worse results in the AI model.
  • Don’t create use cases for small talk or for recognizing a specific language.

Use case examples for different industries

The tables below collect well-performing use case names and descriptions for various industries. Feel free to use these in your own AI agent.

Use case name Description Insights and tips
General industry
Refund status Customer asks for an update on the status of their refund  
Login issues Customer can’t log in due to incorrect password or username In this use case, “login issues” and “password reset” have been merged together due to semantic similarities with customers having trouble logging in.
Update account information Customer asks how to update personal or account information Most often, customers want to update either their personal details (name, phone) or account information (delivery, billing address). Due to semantic similarities, these topics have been merged.
Payment methods Customer wants to know what payment methods are supported  
Email confirmation Customer didn’t receive an email confirmation for their order This use case can be adjusted according to your use case (for example, order, booking, etc.).
E-commerce industry
Order status Customer wants to know the status of their order  
Lost order Customer hasn’t received their order despite delivery  
Sizing enquiries Customer asks about product sizing This can be further iterated to be more product-category-specific, like shoe or clothing sizing.
Incorrect item received Customer claims they received an item they did not order  
Return request Customer wants to return their order, or part of their order  
Travel industry
Cancel booking Customer requests to cancel their booking The use case name and description are unified and not overly lengthy. It clearly expresses the user's desire to cancel their booking.
Changes to booking Customer wants to change the date, destination or departure for their booking. This use case covers three different changes: arrival, departure, and date for travel. Customers can change these details themselves depending on their ticket type, whereas a name change would require agent attention, which is why that’s covered in a different use case (below).
Change the name on booking Customer needs to change the name on their booking As mentioned above, this use case has been excluded from the more general “Changes to booking” use case as they’re semantically different enough, and we were able to write clear names and descriptions without strong confusion.
Recommendations for activities Customer wants to know what activities are available at their location  
Financial services industry
Status of transfer Customer needs information on the status of their money transfer Customer has concerns about whether the recipient has received a money transfer, and wants to check the status of it. This use case doesn't cover lost or unsuccessful transfers, as these are covered in a separate use case (below).
Unsuccessful transfer Customer is concerned about a money transfer that didn’t reach the recipient As mentioned above, this use case covers instances where a money transfer has been made, but it never made it to the recipient account.
Card activation issues Customer expresses issues with credit card activation  
Failed payment Customer wants to understand why their card payment failed  
Powered by Zendesk