Flow Summary Node - User Guide

Overview

The Flow Summary Node enables designers to automatically generate a concise summary of the conversation that has occurred up to that point in a flow. The generated summary is stored in a session variable and can be used later within the flow, displayed to users, or sent to external systems for reporting, analytics, or handoff purposes.

This feature is particularly useful for:

• Capturing conversation context before transfer or escalation

• Generating interaction summaries for CRM or external systems

• Storing conversation insights for analytics

• Reducing the need to process entire conversation transcripts downstream

---

How It Works

Behind the scenes, the Flow Summary Node:

1. Collects the conversation transcript accumulated up to the node.

2. Sends the transcript along with the configured summary prompt to a Large Language Model (LLM).

3. Receives a generated summary from the model.

4. Stores the summary in a session variable for later use within the flow.

Optionally, the conversation transcript can be processed through PII Redaction before being sent for summarization.

---

Configuring the Flow Summary Node

Navigate to flows and choose an existing application and you see a node called "Flow Summary"

1. Summary Variable

Use the Flow Summary Node to generate a concise summary of the conversation based on a configured summary prompt. The generated summary is automatically stored in a session variable, making it available for use throughout the flow, integrations, or downstream processes.

Configuration: Specify the session variable where the generated conversation summary will be saved. This variable can later be referenced in other nodes, API calls, agent handoffs, or reporting workflows.

---

2. Output Format

Choose the format in which the summary will be generated and stored.

The selected format determines how the summary content is structured for downstream usage.

---

3. Language Selection

The summary language is controlled by the languages configured in the Global Flow Configuration.

Options include:

• Same as Conversation

• Specific configured flow language

Note: Only languages configured for the flow are available for selection.

---

4. Redact PII

Redact PII provides an additional privacy and compliance layer.

When enabled:

• The conversation transcript is first sent to the Presidio Analyzer.

• Personally Identifiable Information (PII) is detected.

• Identified PII elements are redacted before summarization.

• The redacted transcript is then sent to the LLM for summary generation.

Supported PII may include:

• Customer names

• Phone numbers

• E-Mail addresses

• Credit card numbers

• Account identifiers

• Other configured sensitive information

Benefits

• Protects customer privacy

• Reduces exposure of sensitive information

• Helps support compliance and governance requirements

---

5. Conversation Insights

When enabled, additional conversational intelligence is generated and included in the output.

Available insights include:

• Intent

• Entities

• Sentiment

• Sentiment Source

• Summary

These insights are stored alongside the generated summary and can be used in reporting, integrations, or workflow automation.

---

Accessing the Output

The generated data is available through the configured output variable.

Available Output Keys

Key	Description
summary	Generated conversation summary
intent	Detected customer intent
entities	Extracted entities from the conversation
sentiment	Overall conversation sentiment
sentimentsource	Source used for sentiment analysis

---

Using the Summary Within a Flow

The stored summary can be consumed in multiple ways:

Use a Say Node to present the generated summary.

---

External API-Integration

The summary can also be passed to an API Call Node and sent to external platforms such as:

• CRM systems

• Analytics platforms

• Ticketing systems

• Knowledge management solutions

• iX Wisdom or other integrated services

The API can consume:

• Summary

• Intent

• Entities

• Sentiment

• Other conversation insights

---

Example Use Case

Customer Support Handoff

1. Customer interacts with a virtual assistant.

2. Flow Summary Node generates a concise conversation summary.

3. Summary is stored in a session variable.

4. The summary is sent to a ticketing or agent-assist system.

5. Human agents receive conversation context without reviewing the full transcript.

Result: Faster handoffs and improved agent efficiency.

---

Bot Testing

Best Practices

• Enable PII Redaction when customer data may contain sensitive information.

• Use Conversation Insights when analytics, reporting, or routing decisions are required.

• Store summaries in a clearly named session variable for easy maintenance.

• Pass summaries to downstream systems instead of entire transcripts whenever possible.

• Validate generated summaries during testing to ensure they meet business requirements.

---

Key Benefits

• Automatically summarizes conversation history

• Stores summaries in reusable session variables

• Supports privacy protection through PII redaction

• Generates conversation insights such as intent and sentiment

• Integrates easily with API calls and external systems

• Reduces dependency on full conversation transcripts

• Improves agent handoff and reporting workflows

---

Conclusion

The Flow Summary Node provides an efficient way to capture and reuse conversation context within a flow. By combining AI-powered summarization, optional PII redaction, and conversation insights, it helps organizations streamline integrations, improve customer support workflows, and enhance operational reporting while maintaining privacy and compliance standards.