Skip to main content

Conversation-Based Knowledge Search in Live Chat

This guide explains how to implement and optimize conversation-based knowledge search, which automatically suggests relevant knowledge articles to agents during live chat interactions based on real-time conversation analysis.

Conversation-based search is an AI-powered feature that:

  • Analyzes live chat conversations in real-time
  • Identifies customer questions, issues, and intent
  • Searches the knowledge base for relevant articles
  • Presents contextual suggestions to agents
  • Enables faster, more accurate responses during chat

Prerequisites​

Before configuring conversation-based search, ensure you have:

  • A well-populated knowledge base
  • Integration with your live chat platform
  • Admin access to Knowledge Management
  • Support agent permissions

Integration Configuration​

Connect your chat platform:

  1. Navigate to Admin > Integrations > Chat Platforms
  2. Select your chat platform (Intercom, Drift, LivePerson, etc.)
  3. Configure the integration:
    • API credentials
    • Webhook endpoints
    • Event subscriptions
    • Authentication method
  4. Test the connection
  5. Enable conversation-based search

Search Configuration​

Customize the search behavior:

  1. Go to Admin > Search > Conversation-Based Search
  2. Configure search settings:
    • Real-time analysis interval (how often to analyze new messages)
    • Context window (how many previous messages to include)
    • Relevance threshold (minimum confidence score)
    • Maximum suggestions per context change
  3. Set up result filtering:
    • By audience (match customer attributes)
    • By product (if specified in conversation)
    • By content type (articles, guides, FAQs)
  4. Save your configuration

Agent Experience​

Viewing Suggested Articles​

Agents will see suggested articles in their chat interface:

  1. During an active chat, the Knowledge panel appears alongside
  2. As the conversation progresses, suggestions update automatically
  3. Each suggestion includes:
    • Article title
    • Relevance score and match reason
    • Brief excerpt
    • Last updated date
  4. Agents can click to view the full article without leaving the chat

Using Knowledge in Responses​

Incorporate knowledge into chat responses:

  1. View a suggested article
  2. Click Use in Response to:
    • Insert the full article
    • Insert a specific section
    • Insert with a link to the article
    • Send as a rich card (if supported by chat platform)
  3. Customize the inserted content
  4. Send the response to the customer

Manual Search During Chat​

Agents can also perform manual searches:

  1. In the Knowledge panel, click Search
  2. The search is pre-populated with conversation context
  3. Agents can modify the search terms
  4. Results are filtered based on conversation context
  5. Agents can sort and filter results

Providing Feedback​

Improve search quality through feedback:

  1. For each suggestion, agents can:
    • Mark as helpful
    • Mark as not relevant
    • Provide specific feedback
  2. This feedback trains the AI to improve future suggestions

Advanced Features​

Proactive Customer Suggestions​

Offer knowledge directly to customers:

  1. Enable Customer Knowledge Suggestions in settings
  2. Configure when to offer suggestions:
    • After specific customer questions
    • When confidence score exceeds threshold
    • During wait times
  3. Customize how suggestions appear to customers
  4. Set up tracking to measure customer engagement

Intent Detection​

Leverage advanced intent recognition:

  1. Go to Admin > Search > Intent Configuration
  2. Create custom intents for your business
  3. Map intents to specific knowledge categories
  4. Train the system with example phrases
  5. Enable intent-based suggestions in conversation search

Sentiment-Aware Suggestions​

Adapt suggestions based on customer sentiment:

  1. Enable Sentiment Analysis in settings
  2. Configure response strategies for different sentiments:
    • Positive: Standard knowledge suggestions
    • Neutral: Detailed technical content
    • Negative: Empathetic content and escalation paths
  3. Set up sentiment thresholds and triggers
  4. Monitor sentiment trends in analytics

Analytics and Optimization​

Performance Metrics​

Track the effectiveness of conversation-based search:

  1. Navigate to Analytics > Chat Search
  2. View key metrics:
    • Suggestion accuracy rate
    • Agent utilization rate (% of suggestions used)
    • Impact on resolution time
    • Customer satisfaction correlation
    • Knowledge gaps identified
  3. Filter by time period, agent team, or chat category

Conversation Analysis​

Gain insights from chat interactions:

  1. Go to Analytics > Conversation Insights
  2. Review the Knowledge Impact report
  3. Identify:
    • Common questions and topics
    • Effective knowledge responses
    • Missing information
    • Resolution patterns
  4. Use these insights to improve knowledge content

A/B Testing​

Test different search configurations:

  1. Navigate to Admin > Search > A/B Testing
  2. Create a new test for chat interactions:
    • Define test variants (different suggestion algorithms)
    • Set test duration and scope
    • Define success metrics (resolution time, CSAT)
  3. Run the test
  4. Review results and implement the winning configuration

Integration with Chat Workflows​

Automated Responses​

Set up knowledge-powered automated responses:

  1. Go to Admin > Automation > Chat Responses
  2. Create rules for automatic responses:
    • For common questions with high-confidence matches
    • During agent handoffs or wait times
    • For specific detected intents
  3. Configure when to use automation vs. agent review
  4. Monitor automation performance and customer satisfaction

Guided Conversations​

Implement guided conversation flows:

  1. Navigate to Admin > Chat > Guided Flows
  2. Create knowledge-based conversation flows:
    • Diagnostic trees for troubleshooting
    • Step-by-step guides for common processes
    • Decision trees for product selection
  3. Link flows to specific detected intents
  4. Allow agents to initiate flows during conversations

Optimizing for Real-Time Interactions​

Content Formatting for Chat​

Adapt knowledge content for chat interactions:

  1. Go to Admin > Content > Chat Formatting
  2. Configure how content is formatted for chat:
    • Break long content into digestible messages
    • Convert bullets to numbered steps
    • Optimize images and attachments
    • Create chat-friendly versions of complex tables
  3. Preview chat renderings of your content
  4. Apply formatting rules to specific content types

Response Time Optimization​

Improve agent response speed:

  1. Create quick-response versions of common articles
  2. Enable keyboard shortcuts for inserting knowledge
  3. Implement type-ahead suggestions based on knowledge
  4. Create pre-written responses for high-volume topics

Best Practices​

Content Strategy​

  • Create chat-optimized content: Shorter, more direct articles work best
  • Use clear headings: Help agents quickly scan for relevant sections
  • Include conversation starters: Suggest follow-up questions
  • Develop troubleshooting flows: Step-by-step guides work well in chat
  • Update based on conversations: Regularly review chat logs for content ideas

Agent Training​

  • Provide platform-specific training: Ensure agents know how to use suggestions
  • Encourage knowledge contribution: Have agents flag missing information
  • Share successful examples: Highlight effective knowledge usage in chats
  • Create chat-specific guidelines: Develop best practices for knowledge in chat

Technical Optimization​

  • Balance speed and accuracy: Tune settings for your specific needs
  • Optimize for mobile chat: Ensure content works well on all devices
  • Consider bandwidth limitations: Optimize images and attachments
  • Test with realistic conversation volume: Ensure performance at scale

Troubleshooting​

Common Issues​

IssueSolution
Delayed suggestionsAdjust analysis interval and optimize indexing
Context misinterpretationTune context window size and provide feedback
Formatting problems in chatReview and update chat formatting rules
Customer confusionImprove clarity of customer-facing suggestions

Next Steps​