Use cases

Nyeti turns message text into routing signals for support, customer-success, and safety workflows. It can run before an LLM, before a queue decision, or after a completed call. It analyzes language; it does not make a final decision about a person or case.

1. Prioritize a support queue

Use urgency and emotional direction to move messages that need prompt attention ahead of routine requests. Keep a person in control of the final escalation policy.

JavaScript

const { annotation, routing } = await analyze(message);

if (annotation.urgency_level === "critical") {
  queue.routeToPriorityReview({ ticketId, posture: routing.response_posture });
}

2. Catch cancellation language before replying

A cancellation request that combines negative language with urgency can move into a retention or human-review flow before an automated response is generated.

JSON

// "I need this fixed today. Your bot is useless and I want to cancel."
{
  "annotation": {
    "emotional_valence": "negative",
    "urgency_level": "high",
    "engagement_state": "disengaged"
  },
  "routing": {
    "response_posture": "EMPATHIC_SOLVE",
    "suggested_action": "acknowledge_then_fix"
  }
}

3. Simplify when cognitive load is high

When a message signals confusion or overload, shorten the next response, offer one step at a time, or route the case to someone who can walk through it with the customer.

4. Review completed voice-agent calls

The ElevenLabs post-call receiver checks the transcript for false resolution, repeated needs, deterioration across turns, and crisis language. Use review_status andsuggested_action to build a focused quality-review queue rather than listening to every call.

See conversation integrations

5. Add a safety handoff for crisis language

CRISIS_SUPPORT is a routing signal produced when configured crisis-language markers match. Use it to hand the case to a trained responder or an established safety workflow instead of returning a generic automated reply. Nyeti is not an emergency service or a diagnosis tool.

6. Respect regional language cues

cultural_context can surface configured East and Southern African language markers. Use those markers to avoid a one-size-fits-all tone, not to infer identity or make eligibility decisions.

Try a message in the Analyze reference or review quota behavior before planning volume.