AI Agent Studio v.35 release

Review what's new and changed, and the fixes and limitations in AI Agent Studio for the v.35 release.

What's new

Pre-built AI Skill templates

We now offer a faster and more efficient way to develop AI Skills through pre-built templates. These pre-built templates are tailored to specific use cases, streamlining the AI Skill creation process.

Create AI Skills from a template
Convert an AI Skill to a template

You can now save your AI Skills as templates and reuse them for similar use cases.

Create template from an AI Skill
Support for RAG capability in AI Agent Studio to create grounded Model connections using Automation Anywhere Grounded by Enterprise Knowledge

Create Grounded by Enterprise Knowledge Model connections using the native RAG (retrieval-augmented generation) capability from Automation Anywhere to generate accurate and contextually relevant information referenced from enterprise knowledge bases.

Create Grounded Model connections with Automation Anywhere Enterprise Knowledge
Custom models in AI Agent Studio

Automation 360 introduces new APIs that enable you to connect to custom generative AI models. These APIs allow you to define models that are not natively provided by AI Agent Studio. After a model is created using the API, you can then create Model connections and use the model within AI Skills in AI Agent Studio.

Custom model definitions

What's changed

AI governance enhancements

  • Governance logs display accurate model names: Earlier, in the AI governance session detail logs, a system-generated model name was displayed. We have enhanced this so the logs now display user-friendly model names. This is in-parity with the way the model names are displayed when selecting a model from the Model connection wizard from the Create Model connection > Connection settings > Choose a model drop-down list.

    Relevant model names displayed
  • Navigation update for AI governance: The left-pane navigation now displays AI > Governance. This has changed from AI > AI governance.

    AI governance brand update to Governance
  • Exported logs via CSV display the updated AI Skill branding: In the Governance > AI prompt log screen, when you use the Export items to CSV feature, the exported CSV logs now export with the new AI Skills branding, which was earlier called Prompt Templates.

    CSV exports with AI Skillbranding
  • Populating the correct user name in the Governance audit logs: The AI governance logs now display the actual names of users who have executed the automations, in the User column. Earlier a system-generated name displayed that was not useful for reference in the logs.

    Actual User name display in logs
  • Folder path of AI Skills: The AI governance logs now display the specific folder path of AI Skills. When an AI Skill is integrated with an automation (parent or child), the folder path displayed is the same as the folder path of the parent automation. This enhancement provides increased visibility and traceability, aiding in security, compliance, and troubleshooting efforts.

    AI governance logs folder path

When creating a Model connection, even if there is a mismatch with the model name in the Azure OpenAI console and the model name configured in AI Agent Studio, you can continue creating the Model connection as long as the Deployment ID is common.

When creating a Model connection using the Azure OpenAI models, while setting up authentication parameters in the Test connection phase, ensure you enter the exact Deployment ID as it appears in the Azure OpenAI console. The model used in Azure deployment must correspond to the model you select from the Choose a model drop-down list in the Connection settings phase.

For example: If the you choose GPT-4 from the Choose a model drop-down list, and the mapped name for the GPT-4 model in the Azure OpenAI console is deploy-gpt-4, then you would have to enter that exact name in the Deployment ID field for the validation to be successful.

In Automation 360 v.33 release, because of this discrepancy, you would not be able to proceed any further and would be blocked from creating the Model connection. We have modified this behavior for the AI Skills package version 3.0.5, to let you continue creating the Model connection in spite of the discrepancy. You would see the warning message in the back-end stating the discrepancy, which could result in the model response metrics to vary.

For example: If you choose GPT-4 from the Choose a model drop-down list, and use the Deployment ID name for GPT-3.5 from the Azure OpenAI console, the model call in the back-end would call the GPT-3.5 model instead of GPT-4, which could result in differences in model-response.
Expanded AI Skills input size

AI Skills input field now supports a significantly increased input size of up to 10,000 characters. This enhancement allows for more complex prompts and richer interactions with your AI Agents, enabling you to build more powerful and versatile automations.

Fix

In the AI Skill editor, when you tried updating the title of an existing AI Skill without an associated Model connection, Prompt-text, or defined Prompt Input, the system would display an error message stating a validation failure and failed to save the update. This limitation is now fixed, and the AI Skill edited title gets updated, and the change gets saved.

Limitations from previous releases

Bot Agent 22.60.10 and later dependency for generating audit logs

For the AI governance feature, when automation administrators enable the Administration > Settings > AI Data Management > Data logging settings, they see a notification requesting them to install a compatible Bot Agent version to ensure the audit logs for all model interactions are logged successfully in the Control Room.

The GRC Leads, Automation Leads, and the Professional Developers accessing the data logs in AI > AI governance > AI prompt logs and AI > AI governance > Event logs in the Control Room also receive the same notification.
Note: We recommend using Bot Agent version 22.60.10 and later.

Assign roles and permissions to enable AI governance

The Search field in the AI governance > AI prompt log and AI governance > Event log tabs list-view pages is not able to process entries with special characters such as the following:

+ = && || > < ! , { } [ ] ^ ~ :