AI Agent Studio v.33 release

Important: This offering is available on Automation 360 Cloud only, and will be available for use with the Enterprise Platform license. For details about the Enterprise Platform license, see Enterprise Platform.

Review what's new and limitations in AI Agent Studio for the v.33 release.

Note: Generative AI models can produce errors and/or misrepresent the information they generate. It is advisable to verify the accuracy, reliability, and completeness of the content generated by the AI model.

What's new

AI Agent Studio is our newest generative AI-powered automation offering that enables organizations to effortlessly build intelligent applications by addressing challenges faced around generative AI foundational model selection, visibility into the data exchanged with the model, and accuracy of the response received.

See AI Agent Studio.

AI Tools

With this feature, you can easily experiment with a range of foundational models from vendors such as Amazon Bedrock, Google Vertex AI, Azure OpenAI, OpenAI and others to gauge their performance within automations, and choose the one that best meets your needs. Build and share reusable Prompt Templates to empower developers with easy integration of automations with foundational models, all through an experience embedded within the Automation 360 Control Room.

See AI Tools
AI governance

Use this feature to drive stronger governance to ensure only approved models are used in automations to meet compliance and regulatory requirements. Uphold the integrity of data exchanged with the model during automation execution through robust monitoring and auditing.

See AI governance
Generative AI Prompt Template package v2.0.6

Use Generative AI Prompt Template package > Execute action in your automations to use the predefined Prompt Templates connected to foundational models.

See AI Skills package.

Limitations

When creating or editing a Model connection (in Chrome version 127), the drop-down menu is not functioning as expected. For more information, see A360.33 | AI Studio Model Connection Dropdowns Not Working in Chrome v127.

Workaround: Use keyboard navigation (arrow keys, Enter, and Tab) to select the desired value in the drop-down.

When you test a Prompt Template, or run an automation executing a Prompt Template, the AI governance > Event log captures and displays the system generated SYSTEM_PROMPT event type. This should not be displayed. We support logging and displaying only these four event types: Send_prompt, Response_received, Model_connection, and Trigger_prompt_template.

In the AI governance > AI prompt log and AI governance > Event log tabs, the system captures and displays the internal serviceUser token (UUID) in the User ID field, and not the bot RunAs UserId as expected.

While testing a prompt input, the response does not render in the same format as the prompt input. This prompt response displays empty logs.

The AI governance > AI prompt log incorrectly displays logs for connecting, disconnecting, and authenticating actions for Generative AI Packages. The logs should display foundational model interaction data only.

For the AI governance feature, when the Automation Admin enables Administration > Settings > AI Data Management > Data logging settings, the Control Room displays a notification asking the Automation Admin to install a compatible Bot Agent version to ensure the audit logs for all model interactions get logged successfully.
Note: We recommend using Bot Agent version 22.60.10 and above.

See Assign roles and permissions to enable AI governance.

In the Manage > Model connections navigation, when you edit an existing Model connection or create a new one, in the Invite roles section, currently, there is no search filter. So the user has to manually search through the list of roles using the scroll bar.

In Manage > Model connections list screen, users could occasionally see a blank page when the server is refreshing the page. To resolve this, users can manually refresh the page to display the content. We are actively working on a fix.

When you are testing a Model connection or Prompt Template on your machine, but are connected to a remote device for running the Bot Agent, you could see the following error messages:
  • "Another desktop operation is in progress. Cannot process new request"
  • "An internal server error occurred, please stop the operation and try again"

Scenario 1: Click on Model connection > Test connection button.

Scenario 2: Click on Create Prompt Template > Get response button.