Agent Conversation Transcripts

​

Overview

The Conversation Transcript View gives you a full-screen read of a single agent conversation, laid out alongside the user’s context (page, company, platform) and the specific signals detected within it.

​

Use Cases

• Investigate a support case or complaint by reading the exact exchange between a user and the agent.
• Audit agent quality by reviewing conversations flagged with a failure or frustration signal.
• Verify a session replay or export a conversation for further review outside Userpilot.

​

Getting Started

You can open a full conversation transcript from three places:

• A conversation row inside a Topics slideout.
• A conversation row inside a Failure signals slideout.
• A conversation row in a user’s Conversations tab on their profile.

​

What you can do from here

1. Follow the exchange exactly as it happened — every prompt and reply in order, including if the user changed page mid-conversation, and how long the agent took to respond to each message.
2. Understand who you’re looking at, without leaving the page — the user’s identity, company, platform, and location sit alongside the conversation, and you can jump straight to their profile.
3. Tell whether this is a one-off or a pattern — see how many other conversations match the same topic or failure signal, or pull up everything else this user has talked to the agent about.
4. Jump straight to the moments that matter — filter down to just the positive or negative feedback, frustration, failure, or slow-response moments in the conversation, instead of reading the whole thing to find them.
5. Search within a long conversation to find a specific word or phrase without scrolling.

• Signals identified throughout a conversation include:

​

Data capture and privacy settings

From the Data Capture & Privacy Settings page, you can turn capture of conversation content on or off separately for production and staging, and set rules to keep sensitive data out of what’s stored — useful if your agent might encounter things like card numbers or other PII in the course of a conversation.

• Exclude keywords – define a keyword or phrase to watch for. Any message that matches is masked within the conversation — since a keyword alone can’t reliably bound where a sensitive value starts and ends — and you can scope this to user messages, agent messages, or both.

​

FAQs