Note: We are actively improving the platform and adding new features. This information reflects the current state of the platform and may evolve as we continue to develop the product. If you have questions or feedback, please reach out via the Support button in the top right.

In this guide, we’ll cover the essential concepts required to create and manage AI agent flows on our platform. These foundational elements will help you build a flexible and effective flow that can adapt to different customer interactions.

​
Key Concepts

​
Flows

A Flow is a structured sequence of states and transitions that guides the conversation between the AI agent and the customer. Flows define the logic and structure of the interaction, ensuring the AI agent responds appropriately based on customer input. Typically, a flow starts with an initial state and progresses through various states, injecting new instructions or functions as the conversation evolves.

​
States

States represent individual steps within a flow. Each state corresponds to a specific point in the conversation where the AI agent interacts with the customer. States can include instructions (known as Thoughts) and Functions, which together determine the AI agent’s behavior and responses at each stage of the conversation. Transitions between states are defined based on customer input or function responses, creating a dynamic conversational flow.

​
Functions

Functions are predefined operations the AI agent can perform during a conversation. Examples include retrieving customer information, updating account details, or sending notifications. Functions are associated with specific states and are triggered as the conversation progresses, allowing the AI agent to execute actions based on the current context.

Note: Functions can be predefined or custom-created. Custom functions use webhooks to communicate with external systems, where users can define function logic and host it externally.

​
Parameters

Parameters are data variables used within a flow. They allow functions to access and store information, such as customer details or conversation context. There are two types of parameters:

  • Normal Parameters: Passed directly to functions when they are called.
  • Internal Parameters: Set by functions and used internally to maintain data consistency within the flow.

​
Events

Events are triggers that facilitate transitions between states. When a function completes an action, it may return an event, which prompts the flow to move to a new state. For example, if a function retrieves payment plan options, it might return a Payment Plans event, transitioning the conversation to a state where those options are presented to the customer.

​
Thoughts

Thoughts are the instructional prompts associated with each state, guiding the AI agent’s responses. Thoughts define what the AI agent should say or do in that state, creating a smooth and guided experience for the customer. Well-defined thoughts are essential for clarity and ensuring the AI agent behaves predictably at each stage.

Check out our guide on Designing Thoughts for best practices and examples.

​
Additional Resources

After familiarizing yourself with these key concepts, you can proceed to build a flow, set up custom functions, and define parameters.

For detailed steps on creating a flow from scratch, setting up functions, and testing flows, refer to our other guides:

With these foundational concepts, you’re well-equipped to build effective, responsive flows that enhance customer interactions on our platform. Continuous testing and refinement are recommended to ensure each flow operates smoothly and meets your customer service goals.

ReportingFlows