Auron
Web consoleConfigure

Signals and signal categories

Defining what Auron looks for in conversations, grouping it, attaching it to agents, and reading what was found.

A signal is something worth noticing in a conversation: a pricing enquiry, a competitor mentioned, a commitment made, a risk raised. Auron only finds what you have defined, which means this screen decides how much a transcript is worth to you.

Defining a signal

A signal definition has:

  • A name, in the form your organization will read in reports.
  • A description of what it is for.
  • A category.
  • Example phrases, added one at a time, which are what teach detection what this looks like in real speech.
  • Labels.

Example phrases carry most of the weight. Two or three phrases in the words your customers actually use beat a long abstract description.

Signals can also be imported in bulk from a CSV file, with the import reporting its progress. That is the practical route when you are bringing an existing taxonomy into Auron.

Categories

Signal categories group definitions so reports and dashboards can roll them up. Categories are managed on their own screen, and each one can be opened to work on the signals inside it.

Attaching signals to agents

A definition does nothing until it is attached to an agent, on the agent's Signals tab. An agent with no signals attached still runs the conversation and produces a transcript; it just detects nothing. This is the most common reason a session comes back with an empty Signals tab.

Reading what was found

Detected signals appear in three places:

  • Inside the session they came from, as its Signals tab.
  • On the organization-wide signals screen, as a feed across every session.
  • In Analytics, as trends, confidence distributions and per-agent contribution.

Each detection carries a confidence percentage, the insight in a sentence, and the quote it was drawn from. The quote is the evidence, and it is what lets you judge the interpretation.

Enrichment

The signals screen can trigger an enrichment pass, which works over what has been detected to produce the composite views the dashboards and the knowledge graph read from. It reports its status while it runs.

How agent behaviour is governed

The rules that shape how signals are extracted and reviewed sit in the AI constitution, separately from the definitions here.

On this page