Agent Conversation Search and Metric Creation

This update adds agent conversation search, and lets the agent create a metric mid-conversation to continue answering.

You can find a past analysis by what was said rather than by title, and when a question needs a metric the ontology does not have, it gets created there and then so the answer continues.

Why we made this

As conversations with the agent pile up, revisiting a figure you checked last week means opening conversations one by one from their titles alone. And if a question needs a metric that has not been defined yet, you stop the conversation, go define it, and come back to ask from the beginning.

This update closes those gaps so you can find past conversations again and keep a conversation going.

Conversation content search

Find conversations by what was exchanged in them.

• Full-text search: Search the contents of a conversation, not just its title.
• Search in the sidebar: Enter a term directly in the conversation list to narrow it down.
• Direct navigation: Open what you find and continue asking from that context.

Now you can pick up the conversation where you checked a figure last month or discussed a root cause, instead of asking all over again.

Metric creation mid-conversation

When a question needs a metric that does not exist, it gets created in the conversation.

• Recognizing the gap: The agent first confirms that the analysis needs a metric that is missing.
• Defining it together: It checks what criteria to build it on, then creates the metric.
• Continuing the answer: The new metric is aggregated immediately and the answer continues.

Now, when you need a perspective nobody has defined yet, you can set up the new criteria without stopping the analysis.

You can try it right away from the Deskroom sidebar > Agent. Spend the time you used to spend finding past conversations and preparing metrics on making decisions from your data.