A use case represents an AI-powered application or workflow being monitored. Creating one sets up how that application's traffic is tracked and governed, and it brings together every interaction type in the workflow under a single monitoring view.

Most of the form is straightforward identifying information. The decision that warrants the most thought is the Profile Flow, which is where you control how each type of interaction in the workflow is evaluated. Use cases are created from the Inventory section.

Create a Use Case

The Create Use Case form is a single-page form.

1. Navigate to Inventory > Use Cases and open the Create Use Case form.
2. Enter a Name for the use case. This field is required.
3. Optionally enter a Description.
4. Select a Type. The Type selection affects reporting only. Selecting Agent adds the Agentic Analytics charts to the use case's analysis in Observe. It does not change the event types available in the Profile Flow or how the use case's traffic is monitored and evaluated:
5. 1. Chatbot: Conversational AI that responds to user messages and answers questions in a dialogue format.
   2. Agent: Autonomous AI that takes actions, uses tools, and completes multi-step tasks on behalf of users.
6. Select a Default Profile from the dropdown. A default profile is required. The default profile serves as the fallback for any event type not explicitly mapped in the Profile Flow.
7. Configure the Profile Flow. The Profile Flow controls which profile evaluates each type of interaction in the workflow. It works on a default-plus-overrides model: every event type routes to the default profile unless you assign it a different one, so you only need to configure the exceptions.
8. 1. To override an event type, expand it and select a profile. This lets different interactions within the same use case be governed by different policies, such as applying stricter data handling to user-facing events than to internal reasoning steps.
   2. The event types are:
   3. 1. Agent Initialized: One agent creates and configures another at runtime, handing off the task context it needs to begin work.
      2. User to Agent: A user sends input to an agent, such as a request or instruction.
      3. User to LLM: A user sends input directly to a language model and receives a response, with no agent involved.
      4. Agent to User: An agent returns output or a response to a user.
      5. Agent to LLM: An agent sends a prompt to a language model as part of its reasoning or processing.
      6. Agent to Tool: An agent calls an MCP server, external tool, or function to perform an action or retrieve information.
      7. Agent to Memory: An agent reads from or writes to a persistent memory store.
      8. Agent to Agent: An agent communicates with another agent, such as delegating a task or passing along a result.
9. Click Create Use Case to complete the form.

Edit a Use Case

After creation, the use case has a detail page showing the creation date, last updated date, and the user who last updated it. you edit a use case directly on this page rather than through a separate form: select a field to edit it, and text fields show and edit pencil when you hover over them. The type, default profile, and Profile Flow mappings can all be changed here.

Edits are not applied as you make them. After you change something, a save control appears, and your changes take effect once you save.