Implementing Genesys Cloud Knowledge Base Integration for Agent-Assist Suggestions

How-To-Articles
1

Implementing Genesys Cloud Knowledge Base Integration for Agent-Assist Suggestions

What This Guide Covers

  • You will architect an intelligent knowledge ecosystem that automatically serves relevant articles to agents during live digital and voice interactions.
  • You will configure the Genesys Cloud Knowledge Workbench to process natural language queries and map them to high-fidelity support content.
  • When complete, your agents will spend less time searching for answers and more time resolving customer issues, leading to significant improvements in First Contact Resolution (FCR) and Average Handle Time (AHT).

Prerequisites, Roles & Licensing

  • Licensing: Genesys Cloud CX 2 or CX 3. (Knowledge Management is included in these tiers).
  • Permissions:
    • Knowledge > Knowledge Base > View/Edit/Publish
    • Knowledge > Workbench > View
    • Conversation > Agent Assist > View
  • OAuth Scopes: knowledge, conversations.
  • Content: You should have at least 5-10 pre-written technical support articles in Markdown or HTML format ready for import.

The Implementation Deep-Dive

1. Architecting the Knowledge Base Structure

A Knowledge Base (KB) is only as good as its taxonomy. In Genesys Cloud, you must move beyond simple “folders” and use Categories and Labels to drive the AI search engine.

The Step:

  1. Navigate to Admin > Knowledge > Knowledge Workbench.
  2. Click Create Knowledge Base and select your language (e.g., English - United States).
  3. The Critical Part: Define Categories that match your support departments (e.g., “Hardware”, “Billing”, “Connectivity”).
  4. Add Labels for synonyms (e.g., for “Refund,” add labels like “Money back,” “Reimbursement,” “Chargeback”).
  5. Architectural Reasoning: Labels are used by the NLU (Natural Language Understanding) engine to expand the search intent. If a customer says “I want my cash back,” the label “Money back” helps the AI bridge the gap to the “Refund Policy” article.

2. Enabling Agent Assist for Real-Time Suggestions

Once the KB is populated and published, you must tell the Genesys Cloud UI to surface these articles to agents during interactions.

The Step:

  1. Go to Admin > Message > Messenger Deployments (for digital) or Admin > Telephony > Queues (for voice).
  2. Under the Knowledge tab, select the Knowledge Base you created.
  3. Enable Agent Assist.
  4. Set the Suggestion Confidence Threshold to 0.6 (60%).
  5. The Trap: Setting the threshold too low (e.g., 0.3) will result in “Knowledge Spam,” where agents are bombarded with irrelevant suggestions, causing them to ignore the tool entirely. Setting it too high (0.9) will result in “Knowledge Silence.” Start at 0.6 and tune based on feedback.

3. Implementing Knowledge in Architect Bot Flows

Knowledge isn’t just for agents; it can power your self-service bots.

The Step:

  1. In Architect, create a Bot Flow.
  2. Use the Knowledge Search action.
  3. Pass the Message.Message.Body as the query.
  4. If an article is found with high confidence, use the Show Knowledge Article action to present the summary to the customer.
  5. Architectural Reasoning: This “Knowledge-First” bot approach reduces the need to build complex intent-based dialogs for simple FAQ-style questions.

4. Training the AI (The Knowledge Workbench)

The Knowledge Workbench provides a “Search Optimizer” that shows you what customers are searching for and what articles were (or weren’t) found.

The Step:

  1. In the Workbench, navigate to Search Optimizer.
  2. Review “Queries with No Results.”
  3. Action: Either create a new article for these queries or link the query as a synonym to an existing article.
  4. Use the Test Search tool to verify that specific technical phrases (e.g., “DHCP Lease Timeout”) return the correct engineering guide.

Validation, Edge Cases & Troubleshooting

Edge Case 1: The “Everything is Relevant” Problem

  • The Failure: The agent assist panel shows 5 articles for every single message, none of which are the right ones.
  • The Root Cause: Overlapping keywords. If every article contains the word “Genesys,” the search engine may struggle to find unique intents.
  • The Solution: Use Negative Keywords or refine your article titles to be more specific. Ensure that “Refund Policy” and “Refund Procedure” are consolidated into a single master article with clear sub-headings.

Edge Case 2: Stale Content in the Cache

  • The Failure: You updated an article in the Workbench, but agents are still seeing the old version.
  • The Root Cause: Articles must be Published to be live. Furthermore, there can be a 5-15 minute propagation delay across the Genesys Cloud global CDN.
  • The Solution: Check the “Version History” in the Workbench. Ensure the latest version is marked as “Published.” If the issue persists, have the agent refresh their browser to clear the local UI cache.

Edge Case 3: Image Rendering Issues in the Agent Panel

  • The Failure: The article looks great in the editor, but the images are broken links in the Agent Assist panel.
  • The Root Cause: Images are hosted on an internal corporate drive or a site that requires a separate login which the Genesys UI cannot access.
  • The Solution: Host all knowledge assets on a public, CORS-enabled CDN or upload them directly into the Genesys Cloud Knowledge Workbench asset manager.

Official References