Custom conversation signals

Define a signal

Active-signal plan limits

The limit counts active signals, not every saved row. Deactivate a signal before activating another when the Stand-in is at its plan limit.

What happens on a match

• The AI conversation pipeline evaluates supported conversation evidence against the Stand-in’s active definitions.
• A first match adds the configured label and evidence summary to the session without posting a visitor-visible message.
• The label becomes visible in the live monitoring view and remains with the completed conversation in History.
• If Notify is enabled, the Stand-in owner receives the configured first-match alert. Later matching messages do not create repeated first-match alerts for that label.
• Digest analysis can group and cite matching conversations when the organization has the corresponding AI insight capability.

Test a signal definition

Use Try it out with one message that clearly should match, one near miss, and one message that uses different wording for the same evidence. Review the resulting label behavior rather than judging only whether the Stand-in’s prose sounds relevant.

A useful definition describes observable conversation evidence and exclusions. Broad topics such as “pricing” tend to create noisy labels; a concrete signal such as “visitor is blocked by a procurement requirement” gives the classifier a testable boundary.

Difference from Product Discovery

• A custom signal classifies evidence; it does not initiate a consent-based research interview.
• Product Discovery has its own built-in label, research fields, consent rules, and adaptive interview behavior.
• The built-in Product Discovery label does not consume the custom-signal active limit.
• A conversation can contain both a Product Discovery finding and one or more custom signal labels when their independent conditions are met.

Limits and boundaries

• Signal labels are AI classifications and should be reviewed against their cited conversation evidence.
• Disabling the skill stops future matching; it does not erase labels already retained with completed conversations.
• Changing a definition affects future evaluation and does not retroactively relabel all retained History.
• Signal notifications require a working browser or push notification channel and do not make the owner eligible for live routing.

Example

The screen below shows the feature in its normal Stand context. Labels and surrounding controls may vary with account state and plan.

Related features

• Product Discovery — Product Discovery lets an AI Stand-in recognize a relevant research opportunity, ask the visitor for explicit consent, and conduct an adaptive interview without replacing the visitor’s original request. Grounded findings are saved with transcript evidence and a built-in Product Discovery label.
• History search and filters — History filters narrow the completed-conversation record by responder, member scope, site, visitor contact, label, date, and transcript text.
• AI digest insights — AI digest insights add evidence-backed themes, Product Discovery findings, and custom-signal matches to the standard chat recap.