Scale your automations responsibly with generative AI across all areas with the AI Agent Studio product suite consisting of AI Tools and AI Governance.

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.

Surfaced directly within your Control Room, AI Agent Studio gives you the tools and governance you need to empower your automation developers to successfully build AI-powered agents. AI Agents combine the power of foundational, generative AI models with the ability to execute and take actions such as creating a case in Salesforce or providing a product replacement recommendation.

These AI-powered agents enable fully autonomous automations by executing a variety of tasks across different enterprise systems within a comprehensive business process. Our AI Agents are ready to support you and take you beyond by connecting you to agentic-processes, orchestrated across enterprise systems.
  • AI Agent Studio is built to be open so you can connect to models across all the major cloud platforms like Amazon Bedrock, Azure OpenAI, and Google Vertex AI while also being able to evaluate the performance of individual models as you create and fine-tune AI Skills. AI Skills can be re-used across automations and empower any automation developer to harness the power of generative AI models without any prior experience.
  • You can also drive stronger AI Governance by ensuring only approved models are used in automations to meet compliance and regulatory requirements while getting better insights into the prompts and responses exchanged between these foundational models through a new set of AI Governance logs.
    Note:

    AI Agent Studio visibility and access

    • To ensure visibility and access to the AI menu (containing AI Agent Studio) and its functionalities in the Control Room, you must have specific role permissions. These permissions are defined in Roles and permissions.
    • This offering requires the Enterprise Platform license. For information about supported version for this feature, see Enterprise Platform.
    • However, explicit permissions assignment is required for users, including administrators to view and interact with the AI Agent Studio interface. If the AI menu is missing, verify that your assigned user role includes all necessary permissions.

AI Agent Studio addresses the challenge of creating automations using the power of generative AI with AI Tools and AI Governance.

The following video gives you an overview of AI Agent Studio with a focus on its value offering:

AI Skills

AI Skills are specific, defined capabilities that you impart to an AI Agent. They represent the cognitive functions an agent can perform, such as summarizing content, extracting specific data points, classifying information, or generating responses based on a given context.

See Skills.

Model connections

Model connections refers to the establishment of links between your AI Agent or application and various external artificial intelligence (AI) models. These connections enable your agent to leverage the advanced processing power and intelligence of pre-trained models, including large language models (LLMs) or grounded models, to execute its defined skills.

See Model connections.

AI Guardrails

To ensure the responsible and secure deployment of AI models built within AI Agent Studio, AI Guardrails provide a critical control mechanism. AI Guardrails are a key component in managing these risks by enabling the enforcement of organizational policies and ethical guidelines within your AI deployments. By configuring guardrails, you can proactively address potential issues such as the generation of harmful content, mishandling of sensitive data, or deviations from desired AI behavior, ensuring your AI initiatives align with your governance framework.

Within the AI Guardrails framework, two key mechanisms contribute to responsible AI behavior:

  • Data Masking: This feature allows the system to automatically identify and mask sensitive data (like PII, PHI, and PCI) in both user prompts and AI model responses. By replacing sensitive information with non-sensitive tokens, the system prevents exposure of real data to LLMs while still allowing for accurate processing.
  • Toxicity Monitoring: AI Guardrails can also be configured to block prompts or responses that are deemed toxic, containing hate speech, profanity, or other harmful language. This ensures that interactions with AI Skills within automations remain safe and appropriate.

See AI Guardrails.

AI Governance

Drive stronger AI 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. Ensure users are following best practices when using generative AI within automations, and when receiving model responses.
  • Data analysis is available for users who have an existing SIEM integrated with the Automation 360 Control Room. Users would have control over data logging managed through setting configuration.​
  • Availability of secure and role-based access control to access the model interaction data.
  • Ability to audit and drill-down into every model interaction occurring within automations.

See AI Governance.

Benefits

Empower automators with GenAI capability to build automations quickly:
  • Guarantee appropriate foundational model usage with Model connections.
  • Fine tune prompt and model usage with AI Skills.
  • Easily re-use AI Skills in a Task Bot.
Drive stronger AI governance to log and track by controlling access to approved models, monitor all user prompts and model responses for an automation execution:
  • Audit and monitor model interaction usage within automation executions to ensure compliance of governance policies. Track model interactions in automations to ensure proper usage and performance of models.
  • Manage stronger security for Model connections and prompt data with role-based access control to approved models. All connections to approved models is managed centrally that would be used within automations.
  • Monitor logs to proactively detect risks of sensitive data exposure and prompt use as per best practices. Prompt data and credentials for authorized access to models is stored securely using industry-standard encryption algorithms.

AI Agent Studio overview