Analyzing conversation flows for advanced AI agents with the conversation journey dashboard

What's my plan?

Add-on

AI agents - Advanced

The Conversation journey dashboard provides an aggregated view of conversations held by zero-training and expression-based AI agents. The tool helps you discover opportunities for optimization by visualizing the conversation journeys. It helps you navigate your dialogue flows to understand the percentage of customers that go through each reply, find patterns, surface problematic areas, and easily track drop-off rates.

Note: The Conversation journey dashboard isn't available for agentic AI agents.

This article contains the following topics:

• Layout
• Use Cases

Layout

The Conversation journey dashboard works best when resolutions states are configured. If you haven't set up resolutions states yet, follow the instructions here.

Global filters

Filter	Description
Language	Filters out the replies in the selected language
Label	Show results related to a specific label(s)
Resolution	Filters reply in the conversation journey explorer with the selected resolution. Note that the resolution filter may mess up the conversation flow.
Advanced	Include/exclude test conversations generated from the "Test AI agent" widget
Date	Select a timeframe from 1st March 2023 till the present date

Main Top KPIs

KPI	Description
Total Conversations	Conversations held by the AI agent in a selected timeframe
Conversations with interrupted Dialogs	Conversations with at least one dialogue where the end-user has interrupted or broken the designed flow by triggering another dialogue
Drop-off rates	Conversations that are abandoned by the end user. A conversation is considered a drop-off if: The final resolution is undefined The last triggered dialogue was not completed by the end-user

Conversation Journey Explorer Tool

Building Blocks

The Conversation Journey Explorer Tool is an aggregated view of all the conversation journeys in a selected time frame.

Block	Description	Visual
Reply Column	Each column represents a step in the conversation journey. The default is to start from the first triggered reply in the conversation journey. The following steps will be built to the right as you select replies and specify the path. The header of the column states: the reply’s position in the journey ex: the Missing Item reply is amongst the replies that were triggered 2nd after the Order Statusreply. continuing conversation percentage on the top right corner of the column visualize the portion of conversations that continued after the welcome reply. ex: 98% of the 800 conversations that started with the Order Status Reply continued to have a 2nd reply in the journey. This indicates that the remaining 2% were either dropped off or the session simply ended
Reply	Each block in the column represents a reply. These replies are the aggregated amount of times the reply was found in a conversation journey in a specific order/stage. ex: The German reply Welcome Reply was triggered as a 1st reply in 5% of the conversation journeys Replies that have a Generative replies block will be marked with a Generative replies label
Selected Reply	This is the state of a selected reply when a conversation journey is built. The reply is pinned to the top of the column. In this state, you can view some insights: Conversation Volumes Conversations Ended: this is the portion of conversations that ended with this reply in the journey. The ended conversations are categorized by the final resolution. Upon hovering over the bar, you can focus on specific resolution details and drill further into these conversations in the conversation logs through the “Open detailed view” button Drop-off conversations: this is the portion of conversations where the messaging conversations are abandoned when the journey reaches this reply.
Detailed View	Clicking on "Open detailed view" will open a drawer to the right hand side of the dashboard. This view will surface sample conversations for the Resolution State,  of the specific reply that you've selected. Simply click the links to be taken directly to the conversation log.
Selected Reply - Summary	We know that the conversation journey explorer can be a bit overwhelming at first. The summary option was introduced to help you understand where are you in the journey and what is going on with the conversations in this journey step.

Use Cases

Escalation pathways - identify replies that have a non-intended high escalation rate. This is often an indication of a poorly designed dialogue flow where your end users don't get their inquiries fulfilled and human intervention is needed.

Reply loops - identify areas where your end users get stuck in a conversational loop. You can rethink and edit the dialogue design to avoid a deadlock. Another potential solution would be providing a fallback answer if your end-user has been stuck in one of your designed dialogues.

Interrupted flows - identify areas where your end users constantly break your designed flows. This could be an indication of poor knowledge training where your end users' inquiries are misunderstood by your AI agent. You can overcome the problem by launching more intents or training existing intents with more expressions.

FAQs

• Why are there discrepancies between the Conversation journey tab of the AI agent analytics page and the Performance summary tab? (For example, the conversation volumes are different.)
  The Conversation journey tab is updated once a day, at midnight, while the Performance summary tab is updated every four hours. Because of this, metrics will be different between the two tabs as each is a snapshot of a different point in time.

• Why do I have a high drop-off rate?
  A conversation is considered dropped off if the final resolution is undefined. The metrics presented in the Conversation journey dashboard are based on the resolutions states. To unlock the full potential of this tool, we recommend configuring the resolution states.