What it looks for
The insight engine watches your workspace and raises findings when something is worth your attention:- A metric moved more than its normal variation
- A pattern in failures, for example a cluster of conversations failing on the same topic
- Something is stuck, like conversations sitting unanswered or a channel that has gone quiet
- A configuration problem, like documents that never finished indexing so the agent cannot reach them
Investigating a conversation
From a finding, or from any conversation, you can trace what happened: which specialist handled each turn, what was retrieved, where time went, and where a turn failed.This view describes the conversation in your terms. It does not expose the internals of our
infrastructure. If you need engineering-level detail on a specific incident, ask your operator.
Insights and knowledge gaps
The two are related and worth distinguishing.
Knowledge gaps run weekly and are a backlog you work through. Insights are closer to real time and
are things you look at now.
Acting on a finding
Most findings resolve into one of:- Write a knowledge entry, if customers keep asking something you cannot answer
- Adjust a rule, if the agent escalates too often or not often enough
- Fix a configuration, if a channel, a calendar or an indexing job is not doing its job
- Nothing, if the change is explained by something you already know about, like a campaign