You added an AI agent to your product. The widget is live, sessions are being created, but your deflection numbers are lower than expected. The question is: where exactly are users dropping off?

The Widget Usage Funnel gives you an answer.

The four stages

Every widget interaction passes through the same sequence:

Widget opened — the user clicked to open the widget
First message sent — the user typed something and submitted it
Conversation completed — the user exchanged at least two messages and received at least one confident response
Positive outcome — the conversation ended without escalation and with neutral or positive feedback

Most teams see meaningful drop-off at every stage. Understanding where it happens — and why — is what the funnel is for.

How to use the funnel

The funnel isn't a metric to optimize in isolation. It's a diagnostic tool.

Low open-to-message rate: Check placement, timing, and the placeholder text. Consider whether the widget is auto-opening when it shouldn't be.

High first-message drop-off: Review the single-exchange conversations with negative feedback. Are they hitting knowledge gaps?

High conversation-completion but low positive outcomes: Pull the specific sessions in the Conversation History Viewer. Read what happened at the end of those conversations.

Each drop-off point has a different cause and a different fix. The funnel tells you which one to work on.