Conversations

The Conversations screen is where you examine the raw evidence behind your bot's scores and issues. It allows you to read actual user transcripts, filter for specific scenarios, and see how conversation volume and outcomes develop over time.

Use this screen when you want to read exactly what users are saying, investigate a specific issue, or understand the daily traffic flow of your assistant

Search, filters, and filter chips

To find exactly what you are looking for, you can use the search bar and the filter menu at the top of the screen.

You can search for specific words or phrases used by the user or the bot. You can also apply filters to narrow the list down by:
  • Topic
  • Resolution status (Resolved, Escalated, Unresolved)
  • Sentiment (Positive, Neutral, Negative)
  • Flags and issue types

When you apply a filter, it appears as a filter chip below the search bar. These chips make it easy to see exactly which conditions are currently active. You can clear individual filters by clicking the "X" on the chip, or clear all of them at once to return to the full

Copy URL with filters

When you find an interesting pattern or a group of conversations that requires attention, you do not need to take screenshots to show your team.

As you add filters, the URL in your browser automatically updates. By clicking Copy URL (or simply copying the link from your browser's address bar), you capture the exact view you are looking at.

When you share this link with a coworker, they will open Inquio with the exact same search terms, date range, and filter chips applied, allowing them to read the same list of transcripts.

Conversations-over-time chart and legend

Below the filters, you will find the conversations-over-time chart. This visualizes your bot's traffic and outcomes across the selected period.

The colors in this chart are not just decorative; they carry all the meaning regarding how conversations ended. Reading the legend line by line explains your bot's daily performance:
  • Gray (Unclassified): Conversations that do not carry a specific issue, kudos, or bounce flag.
  • Green (Kudos): Conversations where the bot performed well and received a positive signal, satisfaction, or gratitude from the user.
  • Yellow (Neutral): Standard interactions that progressed normally without a strong positive or negative flag.
  • Shades of red (Issues and Incidents): Red indicates problems in the conversation. The specific shade tells you what went wrong:
  • Darkest red (Incident): Safety and Compliance risks, attacks, or severe violations that require immediate review.
  • Other red shades: Represent specific dimension failures, including Resolution Accuracy Issue, Containment Quality Issue, and Dialog Fluency Issue.
  • Purple (Bounced): Conversations where the user dropped off, left without engaging further, or abandoned the chat immediately.

Note: If you apply filters, this chart updates to show only the volume and outcomes for the filtered subset of conversations.

Frequent topics

Next to the chart, you will see the Frequent topics block.

This section dynamically updates based on your active filters. If you are looking at the full dataset, it shows the most common subjects overall. If you filter for a specific issue, the Frequent topics block instantly changes to show you exactly which subjects are causing that problem the most.

Clicking on any topic in this list automatically adds it as a filter chip, narrowing the view further.

The conversation list

The main body of the screen is the conversation list. It displays the transcripts chronologically, based on your current filters and date range.

For each row in the list, you will see:
  • the date and time of the conversation
  • the identified Topic
  • the resolution status (Resolved, Escalated, Unresolved)
  • any identified issues or safety incidents (Flags)

Clicking on a row opens the full conversation transcript. Here, you can read the exact back-and-forth between the user and the assistant, see where issues occurred, and understand the context behind the metrics before making changes to your bot.

Related