Strategies & Workflows

Written Flows

Written Flows let you automate consumer engagement across email, SMS, and RCS channels. They are designed to help teams build structured, repeatable messaging strategies while still allowing AI to respond dynamically within the flow you define.

Use written flows when you want to guide digital conversations with clear logic, controlled messaging, and built-in testing tools before deployment.

How Written Flows Work

Written flows are built using a node-based editor. Each flow defines how the system should respond, when it should wait, and how consumers should be routed based on their responses or status.

This structure makes it easy to design consistent engagement strategies while still adapting to different conversation paths.

AI Flow Editor

Written flows can be edited using the AI Flow Editor, similar to call flows. You can use AI to:

  • update flow structure

  • adjust logic

  • refine messaging

  • create or revise conversational paths

The AI Flow Editor is currently in beta. Because the editor is still evolving, we recommend testing all updates in the Simulator before moving a written flow into production.

Nodes

Written flows are made up of nodes. Each node controls a specific step in the conversation.

Branch

A Branch node is used to classify or route consumers based on specific conditions.

This is useful when you want to send consumers down different paths depending on factors such as eligibility, account status, or prior responses.

Send Message

A Send Message node sends a templated response that the AI can reference during the conversation.

This helps ensure that important messaging stays structured and aligned with your strategy.

Wait for Reply

A Wait for Reply node pauses the flow until the consumer responds.

Use this when the next step in the conversation depends on consumer input and you do not want the system to continue automatically.

End

An End node is a terminal step that closes the flow.

Use this when the conversation has reached its intended conclusion and no additional actions should be taken.

Strategy Information

Each written strategy includes the following core metadata:

  • Name - the strategy identifier

  • Description - a summary of the strategy's purpose

  • Version - the current version number

This information helps teams organize, identify, and manage different written strategies over time.

Simulator

Use the Simulator to test conversations and preview how your agent will respond before deploying a written flow.

The Simulator allows you to:

  • select a simulated debtor

  • run through a conversation flow

  • test how the agent responds at each step

  • validate logic and messaging before launch

We recommend using the Simulator whenever you create a new written flow or make changes to an existing one.

Evaluations

You can run predefined evaluations against written flows to assess performance and quality.

Available evaluations include:

Performance

Measures overall quality of engagement.

Collections

Designed for first-party collections use cases.

Servicing

Designed for servicing-related use cases.

These evaluations can help you understand how well a written flow is performing and whether it aligns with your intended use case.

Custom Evaluations

Custom evaluations are coming soon.

Until custom evaluations are available, we recommend using the Simulator to test custom conversation scenarios and validate how your written flows behave in edge cases or specialized workflows.

Best Practices

  • Keep flow logic as simple and readable as possible

  • Use branch nodes only when different paths are truly needed

  • Test changes in the Simulator before deployment

  • Review strategy versions carefully when making updates

  • Use predefined evaluations regularly to assess performance

Summary

Written flows give you a structured way to automate digital engagement across email, SMS, and RCS. With node-based logic, AI-assisted editing, simulation tools, and built-in evaluations, written flows help teams design, test, and improve messaging strategies with greater control.

Was this helpful?