Skip to main content

Overview

The FlowX Agent Designer allows users to create AI Agents, such as one for extracting transaction details from a card statement, using a visual, drag-and-drop interface. Furthermore, FlowX leverages AI to accelerate the agent building and modification process.

Agent Designer interface

The agent designer interface consists of:
  • Left column: Lists all available preconfigured AI actions.
  • Canvas: The main area where AI actions are dropped and connected to design the agent flow.

Steps for Building an AI Agent

1

Manual Design (Drag and Drop)

Manual Design (Drag and Drop)
2

Select AI Actions

Use the left column to find available preconfigured AI actions. These include:
  • Text extraction tools.
  • Document utilities.
  • Custom AI prompt nodes.
3

Design the Flow

Drag and drop the necessary AI actions onto the canvas.
4

Configure Actions

Once placed, configure the AI actions. For an agent extracting card statement details, actions are configured to:
  • Parse the document.
  • Extract specific data (for example, card holder name and address).
  • Send the results to the next action in the agent flow.
5

Connect Actions

Connect the AI actions in the desired sequence.
6

Run a Test

Once the AI actions are connected, run a test to validate the new AI agent.

AI Context Understanding

  • AI Context Understanding: The FlowX core understands the entire context, the process, and the user’s intent.
  • Modifying Agents with Prompts: To make a change to an existing agent:
    • Open the Command Center.
    • Describe the request in a prompt.
    • The AI will handle the update automatically.
  • User Review: The user is always required to review, approve, or adjust the AI-handled update as needed.
  • Final Validation: After any modification, run the test again to ensure the agent is ready.

Potential Applications of AI Agents

AI agents built with FlowX can be used in banking and other industries to:
  • Extract data.
  • Verify consistency between multiple documents.
  • Run risk checks.
  • Enforce policy rules (for example, flagging overdue invoices or mismatch terms).