Introducing Conversation Insights: Connect chat conversations to the customer journey.
10-29-2026
Customer satisfaction with a brand’s AI-powered chat is often lacking. In fact, this is the scenario that commonly plays out.
A customer explains their issue, the agent misses the point, and the customer requests a human representative. By that point, the customer is likely frustrated, and their next move is probably one the brand wants to avoid: cart abandonment, calling customer support, or buying from a competitor.
Gartner research backs this up — only 27% of customers say they would be willing to try a chatbot again after a negative experience. This creates what’s known as the "leaky bucket" problem, where one unsuccessful interaction discourages future use, even as the chatbot's capabilities improve. It not only makes a one-time conversion difficult but can affect a customer’s long-term opinion of your brand.
But how do you know how well your chat agent is working? The answer is in the conversations themselves.
Chat transcripts are a gold mine of insights.
Chat experiences generate massive amounts of rich data, but that data is usually siloed from other customer behaviors in disconnected systems and is not tied to customer identity. Did customers who interacted with your agent go on to make a purchase or did they initiate a refund? These questions are hard to answer and connecting the dots takes significant data engineering. Improving AI-powered chat experiences requires feedback tied to real outcomes. Without that link, you’re blindly tuning the model.
Adobe Customer Journey Analytics turns chat conversations into actionable insights.
Conversation Insights — now available to all Customer Journey Analytics and Customer Journey Analytics B2B customers — uses AI to extract signals such as topics, tone, keywords, intent, and sentiment from unstructured chat and analyzes them in the context of the entire customer journey. It’s now possible to have a holistic view of each customer’s experience instead of an isolated transcript.
These new capabilities let your team measure the ROI of your chat agents and see where the models need tuning. They answer questions like “Are the agent's responses helpful?” and “Did those customers go on to convert?” They also help you spot friction points and gather customer feedback to inform roadmap priorities.
How it works.
Conversation Insights ingests unstructured chat data and extracts topics, tone, and sentiment. These signals blend with your existing Customer Journey Analytics data, allowing you to analyze them alongside customer behavior on your website and across other channels. For example, you can explore trending topics to see what customers are talking about and how those conversations change over time.
You can also create custom topics and subtopics (in development) that reflect how your business talks about customers, instead of relying on generic categories. Then build segments from keywords in prompts and responses using semantic matching that catches variations (like “cancel” and “end subscription”), helping you identify the conversations that matter most.
With the conversation summary tool, you can focus on a group of conversations, such as negatively rated chats, and get a summary of what they have in common. You can further check subtopics with their share of keywords along with representative customer quotes.
Additionally, you can go beyond the numbers by opening individual conversations and reading them as the customer experienced them. See exactly where the experience broke down and what the customer said right before they left. Sentiment and customer feedback are visualized for whichever group you select.
Ready to get started? Follow the step-by-step guide to start analyzing your chat conversations today.
Recommended for you
