Xurrent AI Agent example
Use an AI Agent to help users find knowledge, inspect requests and submit request updates through conversation. Xurrent remains the system of record; Autom Mate maps the agent's selected skill to the configured Xurrent API action.
Initial capabilities
The documented template includes Get request, Get resolved requests, Add note to request, Create request and knowledge search. Enable only the skills your first service journey needs. Start with knowledge and request lookup, then introduce a controlled creation or note operation.
Prepare credentials and permissions
Configure the model and Xurrent credentials in Vault. The source template describes OpenAI or Azure OpenAI, with an API key and, for Azure OpenAI, the appropriate endpoint and model/deployment selection. Use the provider configuration supported by your installed agent setup.
For Xurrent, the source describes API-key or OAuth authentication. The OAuth client-credentials configuration documented in the integration guide uses API OAuth Service URL, API URL, Account Name, Client ID and Client Secret. Verify the correct Xurrent account and permissions for the chosen operations. Request creation and notes need write access; knowledge and lookup need read access. See the Xurrent library.
Configure the experience
Create or import the Xurrent template in AI Agent Composer.
Select the model and Xurrent credentials. Review each skill's purpose, input contract and output mapping.
Configure Knowledge → Applications → Xurrent for the knowledge content the agent should retrieve.
Explain in the agent instructions when it should search knowledge, ask a clarifying question or call a request skill.
Map the required fields of Create request explicitly. If your request depends on a service template or additional form fields, collect and validate those values before calling the write action.
Test in Playground, then enable the selected channel. For Teams, use the channel setup guide.
Validate the full exchange
Try a knowledge question, a known request, a request the user must not access, a missing required field and a failed API response. For a write, inspect the resulting Xurrent record and return its reference only after the action reports success. Do not say a service was fulfilled merely because its request was created.
If users should access only their own requests, map the channel identity to the relevant Xurrent identity and test the record filters. Treat identity mapping as part of your implementation, not a property of natural-language conversation.
Add fulfilment separately
You can extend the agent using additional skills and Hyperflows. Request intake, Xurrent approval and downstream execution are separate steps. Design the event that releases approved work, the operation it can execute and how completion is written back before enabling that extension. The Service Automation model is useful when this becomes a repeatable service.
Was this helpful?

