Insights answers one question: is this working, and where is it not? Everything on the page is scoped to the project and the date range at the top.
The numbers worth acting on
Escalation rate - how often an agent hands a conversation to a person. Climbing means the agent is out of its depth: usually a missing document, occasionally a job description that promises more than it knows.
Retrieval hit rate - how often a search of your documents found something useful. A low rate with plenty of searches means people are asking about things you have not uploaded. The content gaps list on the Detailed Insights page names those questions directly; it is the most useful diagnostic list available.
The rest, in plain terms
- Tokens used - exactly how many input and output model tokens your agents consumed over the period.
- Tasks done on their own - autonomous tasks, split by finished, failed, and waiting on your approval.
- Collaboration between agents - how many tasks and conversations your agents handed off or delegated to specialized colleagues.
- What each agent did on its own - independent performance breakdown per employee, including tokens and cost.
- Approval turnaround - how long things sit waiting for a person. That one is about you, not the agent.
Detailed Insights & Disliked replies
For deeper analysis, click View Detailed Insights at any time. Clients can rate replies with a thumbs-down; each is listed with the question that produced it so you can inspect the transcript and close the gap.
A sensible weekly loop
- Check the executive Insights dashboard to review tokens used, autonomous tasks finished, and agent collaborations.
- Open Detailed Insights to check content gaps and upload one document closing the biggest gap.
- Review any disliked replies and glance at the escalation rate.
Ten minutes, and it compounds.