Skip to main content

Flow Documentation

This guide will help you provide your AI flow with the information it needs to handle customer conversations effectively. Flow documentation allows you to attach relevant documents, FAQs, and contextual information that your flow can reference during conversations.

Overview

Flow documentation is a key feature that enables your AI agent to access and utilize reference materials during conversations. By configuring documentation for your flow, you ensure that the AI has access to the most current and relevant information to provide accurate, helpful responses to customers.

Accessing Flow Documentation

To access the documentation settings for a flow:
  1. Navigate to Build > Flows
  2. Open the flow you want to configure by clicking on it
  3. Click on the Configure button/link at the top
  4. Navigate to the Documentation tab

Documentation Configuration Options

Amount of Documentation

Controls how much documentation and FAQ content is provided to the AI after each customer message.
  • Setting: Slider control (scale of 1-10)
  • Default: 5
  • Purpose: Balances between providing sufficient context while maintaining conversation efficiency
  • Higher values: More documentation segments are included with each response, providing richer context
  • Lower values: Less documentation is included, keeping responses faster and more focused
Best Practice: Start with a value of 5 and adjust based on:
  • Complexity of your documentation
  • Length of customer interactions
  • Response time requirements

Enable Query Rewriting

A toggle that activates the query rewriting agent to improve document search accuracy.
  • Purpose: Enhances the ability to find relevant documents when customer queries don’t exactly match documentation terms
  • Use Case: Enable this when your documentation uses specialized terminology or when customers use different phrasing.
  • Example: If documentation says “Energy Rebate Program” but customers say “discount on power bill”, query rewriting helps bridge that gap
  • Performance: May add slight latency to document retrieval

Query Rewriter Context

Provides additional context that helps the query rewriting agent improve search results.
  • Format: Text field for free-form context
  • Content Examples:
  • Industry-specific terminology and abbreviations
  • Common customer phrasing variations
  • Synonyms for key concepts
  • Important context about your business domain
  • Example: “Customers refer to ‘solar panels’ as ‘PV systems’ or ‘sun panels’. ‘Rebate programs’ include federal and state incentives.”

Documents

Specifies the reference documents that your AI agent should have access to during conversations.
  • Purpose: Provides comprehensive reference material
  • Document Types Supported: PDF files (and other formats depending on your system)
  • How to Add:
  1. Click “Create New” button
  2. Select a document from your organization’s document library
  3. The document will be indexed and available for the AI to reference
  • Multiple Documents: You can attach multiple documents; the AI will search across all of them
  • Document Management: Documents are managed in the Docs > Manage section of Hub CX

Frequently Asked Questions (FAQs)

Pre-configured questions and answers that your agent should reference during conversations.
  • Purpose: Ensures consistent, accurate answers to common customer questions
  • How to Add:
  1. Click “Create New” button
  2. Select from existing FAQs in your organization
  3. Multiple FAQs can be associated with a single flow
  • How It Works:
  • When relevant to the customer’s query, the FAQ answers are provided to the AI
  • The AI uses FAQ content to inform its responses while maintaining natural conversation flow
  • FAQ answers are formatted for easy AI consumption
  • Example: If “Energy Bill Relief Fund”, “Cheaper Home Batteries Program”, and “Grace AI Factsheet” are configured, the AI can reference these when appropriate

Workflow: Setting Up Flow Documentation

Step 1: Gather Your Materials

  • Collect all relevant documents (PDFs, guides, policies, procedures)
  • Identify key FAQs specific to this flow
  • Note any specialized terminology or customer communication patterns

Step 2: Upload Documents

  1. Go to Docs > Manage in your Hub CX dashboard
  2. Upload your reference documents
  3. Ensure documents are well-organized and clearly named

Step 3: Create or Organize FAQs

  1. Go to Docs > FAQ
  2. Create or verify FAQ entries for common customer questions
  3. Ensure answers are clear and concise

Step 4: Configure the Flow’s Documentation Tab

  1. Open your flow and go to Description > Documentation
  2. Set the “Amount of Documentation” slider to an appropriate level (start with 5)
  3. Enable Query Rewriting if your documentation uses specialized terminology
  4. Add Query Rewriter Context if needed
  5. Attach relevant documents
  6. Attach relevant FAQs
  7. Click Save

Best Practices

Documentation Quality

  • Keep it Current: Regularly review and update documents to reflect current policies and procedures
  • Organize Logically: Use clear titles and structure in documents for easier retrieval
  • Remove Duplicates: Avoid uploading multiple versions of the same document
  • Concise Writing: Shorter, focused documents are easier for the AI to search and reference

Query Rewriting

  • Enable for Complex Domains: Use query rewriting if your industry has specialized terminology
  • Provide Examples: In Query Rewriter Context, include examples of how customers phrase questions differently
  • Test Results: Monitor conversations to see if query rewriting improves document relevance

Amount of Documentation

  • Test Different Levels: Experiment with different values to find the optimal balance
  • Monitor Performance: Higher values provide more context but may slow response times
  • Adjust by Flow Type: Complex flows may need higher values; simple informational flows may use lower values
  • Consider Conversation Length: Longer, more detailed conversations may benefit from higher documentation amounts

FAQ Management

  • Keep Answers Brief: FAQs are most effective when answers are 1-3 sentences
  • Update Regularly: Remove outdated FAQs and add new ones based on customer conversations
  • Link to Documents: Reference specific documents in FAQ answers when more detail is needed
  • Test Relevance: Monitor if the AI is using FAQs appropriately

Configuration Example

Here’s an example configuration for an Energy Services flow:
SettingValueReason
Amount of Documentation6Customers ask complex questions about rates, programs, and services
Enable Query RewritingYesCustomers use varied terminology (e.g., “my bill” vs. “charges”, “solar” vs. “PV”)
Query Rewriter Context”Customers may refer to energy rates as ‘charges’, ‘costs’, or ‘rates’. Solar panels are also called ‘PV systems’ or ‘solar array’.”Improves search accuracy
Documents- Company energy plans guide
- Billing and rates manual
- Rebate programs guide
- Troubleshooting guide
Comprehensive reference material
FAQs- How do I reduce my energy bill?
- What rebates are available?
- How do I report an outage?
- How do I view my usage?
Answers to most common questions

Troubleshooting

AI Isn’t Finding Relevant Documents

  • Check Query Rewriting: Enable it if not already active
  • Improve Context: Add more examples to Query Rewriter Context
  • Verify Documents: Ensure the right documents are attached
  • Increase Documentation Amount: Try increasing the slider value

Response Times Are Slow

  • Reduce Documentation Amount: Lower the slider value
  • Review Document Size: Consider breaking large documents into smaller, more focused files
  • Disable Query Rewriting: This can add latency; test performance without it

AI Seems Inconsistent

  • Update FAQs: Verify FAQ answers are clear and consistent
  • Check Document Currency: Ensure documents reflect current policies
  • Review Query Rewriter Context: Make sure it accurately describes customer terminology

See Also


This feature helps ensure your AI agents have access to accurate, current information for every customer interaction.