We've made a series of improvements to how filter chips are displayed across Signals and Conversation Explorer. These updates make it much clearer what filters are applied to your data at a glance:

When a logged-out user clicked a link (e.g., from a subscription email) and was redirected to login, the destination URL was being lost. Users are now correctly routed to their intended destination after logging in. (This was previously fixed for deep links to conversations — this fix extends that behavior more broadly.)

Users can now move insight cards to any location on their custom dashboards! To do so, simply navigate to the overflow menu on the insight card and update the order.

Previously, many reports across Signals would change the sorting (order of data shown), based on volume. This made it hard to do apples to apples comparisons when changing filters or looking at side-by-side reports. We’ve updated this, so now:

Note: This work is a precursor to adding expansive “group by” capabilities across Signals!

When users click “Go Deeper” on model results reports, the KBo analysis will now include specific call quotes in its analysis.

This is a powerful feature when analyzing model results, and it just got significantly more useful! Remember that Go Deeper looks directly at model result reasoning, meaning that when you use this feature you’re analyzing a sample of the conversations with that model result and digging into the why.

Users can now coach conversations even when no evaluation is completed. One example of when this is valuable: when an AI-triggered Safety Event Workflow occurs, and an evaluation is not yet completed on the conversation, the reviewer can now immediately coach.

What’s included:

We’ve improved clarity and consistency in pinning behavior. Now:

We’ve changed how percentages are calculated in stacked bar charts to better reflect individual group performance.

Before:
Percentages were based on the total across all groups, making individual breakdowns harder to interpret.

Now:
Percentages are calculated using the total within each individual group/bar.

Example — Group by Agent (Jane Doe):

Previous calculation (across all agents, total = 200):

New calculation (within agent total):

This change makes it easier to understand performance at the individual level.

Before:

After:

We’ve updated the date selector to include more, and higher value, dynamic date ranges. The available options are now:

These dynamic date options are especially powerful when used in saved filters (e.g., pinning a saved filter with “this month” in it will mean the dashboard always shows the current month, vs. a static date range).

After our exciting launch of KBo Research Mode, we’ve made a few additional updates to further refine the experience. When a user clicks the “Open in KBo to explore deeper insights” link from the email, the chat in Platform will have a distinct look and feel, with a consistent chat name, an analysis goal banner that displays the user’s (editable) prompt, and a “View Conversations” action to see the analyzed set of conversations in Convo Explorer.

Now, users can apply saved filters on charts, making it quicker and easier to configure reports. To do so, go to the Configure menu on a chart, click the Filters dropdown, and select “Apply Saved Filter”.

How it works: