Building Proactive Chat Invitations using Journey SDK and Predictive Engagement

How-To-Articles
1

Building Proactive Chat Invitations using Journey SDK and Predictive Engagement

What This Guide Covers

  • Implementing the Genesys Cloud Journey SDK to track real-time customer behavior.
  • Configuring Predictive Engagement (GPE) Action Maps to trigger proactive chat based on intent and probability of abandonment.
  • Customizing the “Offer” UI to match brand aesthetics while maintaining high conversion rates.

Prerequisites, Roles & Licensing

  • Licensing: Genesys Cloud CX 2 or 3 (CX 3 includes GPE by default; CX 2 requires the GPE add-on).
  • Permissions:
    • Journey > Action Map > View/Edit
    • Journey > Outcome > View/Edit
  • External Dependencies: The Genesys Cloud Messenger snippet must be deployed on the target website.

The Implementation Deep-Dive

1. Instrumentation: Deploying the Journey SDK

To trigger an invitation, Genesys must first “see” the customer. This requires more than just the base Messenger snippet.

  • Enable Tracking: In your Messenger Deployment, ensure “Journey Tracking” is enabled.
  • Identify Outcomes: Define what a “Success” looks like (e.g., reaching a checkout page or downloading a whitepaper). In Admin > Journey > Outcomes, create a new outcome (e.g., “Order Completed”).
  • The Trap: The “PII Leak”: By default, the Journey SDK tracks page titles and URLs. If your URLs contain sensitive data (e.g., checkout?email=user@example.com), this data is sent to Genesys Cloud. The solution: Use the journey:data command to scrub or mask sensitive URL parameters before the SDK processes the page view.

2. Predictive Engagement: The Action Map

The Action Map is the “Brain” that decides when to pop the chat.

  • Trigger Condition: Instead of a simple “Time on Page” trigger, use Segment Membership or Outcome Probability. For example, trigger the chat if the “Cart Abandonment” probability exceeds 70%.
  • Action Type: Select “Web Messaging Offer.”
  • Architectural Reasoning: Predictive Engagement uses machine learning to identify the “Sweet Spot”-the moment a customer is frustrated enough to need help but hasn’t yet closed the tab. Triggering too early is annoying; triggering too late is useless. By using Outcome Probability, you are letting the AI optimize the conversion rate rather than relying on arbitrary timers.

3. Customizing the Invitation (The Offer)

The default invitation is often too generic. To increase engagement, you must customize the payload.

  • Attributes: Pass custom data to the Architect flow that handles the invitation. For example, if the user is looking at a specific product, pass the productID so the agent can see it immediately.
  • The Trap: The “Over-Triggering” Burnout: If a customer declines an invitation, you should not show it again for at least 24 hours. The solution: Configure the “Frequency Caps” in the Action Map settings to prevent harassing the customer across multiple page loads.

Validation, Edge Cases & Troubleshooting

Edge Case 1: Silent Failures in Single Page Apps (SPA)

  • The failure condition: Action Maps only fire on the first page load; subsequent navigation (via React/Vue/Angular router) does not trigger invitations.
  • The root cause: SPAs do not trigger a full browser “page load” event. The Journey SDK needs to be manually notified of route changes.
  • The solution: Call Genesys("command", "Journey.record", { pageview: { ... } }) on every route change in your SPA logic.

Edge Case 2: Invitation Appearance Latency

  • The failure condition: The AI decides to trigger the chat, but the invitation takes 3-5 seconds to appear, by which time the user has scrolled past the relevant section.
  • The root cause: High network latency or a heavy “Condition” stack in the Action Map (evaluating too many segments).
  • The solution: Simplify Segment definitions and ensure your ac.js (Journey SDK) is loaded with async but high priority in the HTML header.

Official References