Agent Studio and Custom Agents Overview

Overview

The Agent Studio empowers non-technical users to design, deploy, and manage custom agents  fine tuned to your business and processes. With a natural language interface that eliminates the need for coding experience, users can create and publish single-purpose AI agents that allow your teams to perform critical tasks in Kantata in just the way your organization prefers. 

Expertise Agent vs Custom Agents

The Expertise Agent is a powerful agent capable of responding to open ended questions; however, you may want to create custom agents for repeatable tasks and standardized outputs. This section explains how the two differ and provides guidance on when to use which type of agent.

The Expertise Agent:

• Is suitable for one-off tasks and exploratory questions
• Dynamically assembles the tools it needs for each request, which makes it flexible and able to respond to open ended questions
• Provides strong answers and proactive recommendations

Custom agents:

• Are suitable for repeat tasks that require consistent outputs
• Have predefined instructions and toolset, ensuring their behavior and output on each run
• Provide reliable outputs

How to Create a Custom Agent

1. In the left navigation, hover over Settings, then select Agent Studio.
2. Select New Agent.

3. In the Agent Builder chat, describe the custom agent you would like to create.

   Tip: You can include information like sample user prompts that should invoke this custom agent or how to format responses.

   A new custom agent is created with the Draft status and the Agent Builder responds with details about the custom agent it created.

4. After a few moments, another message appears that indicates whether validation of the custom agent was successful or not.
   • Validation Failed—The custom agent you described wasn’t created successfully due to issues such as inaccessible data or unsupported capabilities. Expand this message to view the details and suggestions for how to fix the custom agent. A custom agent cannot be published until validation issues are fixed.
   • Validation Passed—No issues were found with the custom agent. Expand this message to view the details.
5. The following information appears to the right:
   • Custom Agent Name—A name is automatically generated based on your prompt. You can rename the agent by selecting the More (“   ”) icon > Rename.
   • Custom Agent ID
   • Custom Agent Description—A description is automatically generated based on your prompt.
   • Instructions—The Agent Builder’s interpretation of your request. Use this section to confirm the Agent Builder understood your request correctly.
   • Guardrails—The custom agent’s limitations. Use this section to confirm the custom agent can accomplish the tasks or actions you requested.
   • Capabilities—A list of the tools or actions the custom agent will use to process user prompts and prepare responses. Use this section to confirm the custom agent will be able to process user prompts and respond correctly.
6. Refine the custom agent through conversation. As you make changes, the information to the right updates and the custom agent is re-validated.

7. To publish the custom agent, select Publish to the upper-right.

   Note: The Publish button does not appear if validation failed. You can fix validation issues through the Agent Builder chat.

Custom Agent Examples

Here are some examples of custom agents to provide inspiration.

• Status Report custom agent—Provide a summary of project health, risks, and blockers for executives in a consistent format.
• Quarterly Business Review Prep custom agent—Provide a summary of the quarter’s successes, risks, and decisions in preparation for the client review.
• Billing Reconciliation custom agent—Provide a list of ready to invoice items, pending items, and items stuck in review, in a consistent layout and format.
• Bench Report custom agent—Provide a summary of team members who are rolling off projects and team members who are underutilized, in a consistent layout and format.

Related Articles

• Kantata AI Overview
• Expertise Agent Overview
• Agent Studio and Custom Agents Overview
• Kantata MCP Server Overview
• Access Group Set: Kantata AI