Building Custom Agent Assist Knowledge Surfacing Rules based on Customer Journey Data

How-To-Articles
1

Building Custom Agent Assist Knowledge Surfacing Rules based on Customer Journey Data

What This Guide Covers

  • Architecting proactive knowledge article surfacing within the Genesys Cloud Agent Assist panel.
  • Leveraging Predictive Engagement (Journey Data) to trigger specific knowledge recommendations before the customer even asks a question, based on their recent web browsing history.
  • The end result is an ultra-contextual Agent Assist experience where an agent handling an inbound chat sees exactly which support articles the customer was reading on the website 30 seconds ago, allowing the agent to instantly provide the right solution rather than asking redundant discovery questions.

Prerequisites, Roles & Licensing

  • Licensing: Genesys Cloud CX 2 or 3 with the AI Experience add-on and Predictive Engagement enabled.
  • Permissions: Knowledge > Article > View, Journey > Action Map > Edit, Architect > Flow > Edit.
  • Infrastructure: The Genesys Cloud Journey snippet deployed on your website, and a populated Knowledge Base within Genesys Cloud.

The Implementation Deep-Dive

1. The Limitation of Standard Agent Assist

Standard Agent Assist uses real-time transcription (NLU) to listen to the call or read the chat. If the customer says “I need help resetting my router,” Agent Assist searches the knowledge base for “router reset” and surfaces the article.

The Trap:
This is reactive. It requires the customer to explain their problem to the agent. But in digital channels, the customer has usually spent 10 minutes frantically clicking through your self-service website before starting the chat. If they were just looking at the “Error 404: Router Not Found” page, why make them explain it?

2. Capturing Journey Data via Predictive Engagement

First, you must track the user’s web behavior.

Implementation Steps:

  1. Web Tracking: Ensure the Genesys Cloud Journey snippet (journey.js) is deployed on your website.
  2. Define Segments: In Admin > Predictive Engagement > Segments, create a segment named Router_Troubleshooting.
  3. Set the segment condition: If the user views 2 or more pages where the URL contains /support/router or /error/404.
  4. Action Map: Create an Action Map linked to this segment. The action is to offer a Web Chat.
  5. The Secret Sauce: When the chat is offered via the Action Map, Predictive Engagement attaches the entire Journey Context (the Segment ID, the last viewed URL, the search terms used on your site) to the Conversation object in Genesys Cloud.

3. Extracting Journey Data in Architect

When the chat enters the Inbound Message Flow, you must extract this context so Agent Assist can use it.

Implementation Steps:

  1. In your Architect Inbound Message Flow, use the Get Journey Session action.
  2. Extract the Session.LastViewedUrl or the active Segments array.
  3. Use a Set Participant Data action to map this data to a custom string variable, e.g., Context.CustomerRecentSearch = “Router Troubleshooting”.

4. Configuring Agent Assist Surfacing Rules

Now, you configure Agent Assist to read that Participant Data and proactively surface the article.

Implementation Steps:

  1. Navigate to Admin > Contact Center > Agent Assist.
  2. Open your active Assistant profile.
  3. Under Knowledge Surfacing Rules (or via the Assistant Script integration), you can define custom trigger conditions.
  4. The Trigger: Instead of relying solely on the real-time transcript, configure a rule that says:
    If Participant Data Context.CustomerRecentSearch equals “Router Troubleshooting”, immediately pin the Knowledge Article ID KA-1024 (Router Factory Reset Guide) to the top of the Agent Assist panel.
  5. The Agent Experience: The interaction alerts the agent. The agent accepts. Before the customer even types their first message, the Agent Assist panel flashes: “Based on the customer’s web journey, they were looking at Router Errors. Suggested Article: How to Factory Reset.” The agent can instantly type: “Hi! I see you might be having trouble with your router. Would you like me to walk you through a factory reset?”

Validation, Edge Cases & Troubleshooting

Edge Case 1: Over-Surfacing and “Creepiness”

  • The Failure Condition: The customer is browsing the “Divorce Lawyers” page on a legal services website. They initiate a chat. The agent immediately greets them with: “I see you are getting divorced! Here is our article on splitting assets.” The customer is horrified by the privacy intrusion and abandons the chat.
  • The Root Cause: Exposing raw, sensitive web browsing history directly to the agent’s greeting.
  • The Solution: Use Journey Data to surface internal process documentation to the agent, but train the agent not to reveal the exact tracking mechanism. Alternatively, build Architect rules that categorize sensitive URLs into broad intents (e.g., Intent=Family_Law_General) rather than passing the exact URL string.

Edge Case 2: Stale Journey Data on Mobile

  • The Failure Condition: A user browses the website on their laptop. 3 hours later, they open the brand’s mobile app and start an in-app chat. Agent Assist surfaces articles related to the laptop browsing session, which is completely irrelevant to their current mobile issue.
  • The Root Cause: Journey Sessions persist. If the user identity is stitched across devices, the LastViewedUrl might reflect an old session.
  • The Solution: In your Architect flow, always evaluate the Session.Duration or check the timestamp of the last web event. If the last web event occurred more than 15 minutes ago, do not pass the Journey Data to Agent Assist. Rely exclusively on the real-time NLP transcript for knowledge surfacing.

Official References