Call intelligence.
Verified voice workflows.

Noesis identifies people, IVRs, voicemail and other agents. Route conversations in your existing stack. Lexis runs defined workflows with caller verification, access controls and human hand-offs.

NoesisIntelligenceCall intelligence
Noesis · call routing preview
On the other end
IVR

“For account services, press one.”

NoesisListen · classify · adapt
For human conversationsYour high-fidelity model

Preserve the customer experience.

Standby
For automated endpointsYour lower-cost model

Match inference cost to the interaction.

Active route
Continuous classification

An automated menu is on the line. Use a lower-cost route while you navigate it.

Call classification and routing.

Keep your high-fidelity model for human conversations. Choose a lower-cost route for automated endpoints. Set the route in your orchestration layer.

At call start

Initial recipient classification.

Noesis classifies the first few seconds, selects a route and lets your existing stack take over. Classification runs at the beginning of the call.

Throughout the call

Continuous recipient classification.

Keep listening for an IVR handing off to a person, or voicemail picking up mid-dial. Re-route when the conversation changes.

In your stack

Existing orchestration platforms.

Integrate with LiveKit, Pipecat and their derivatives, or run inside your own infrastructure. Voice AI companies already use Detectif.ai in production.

Detection performance, independently scored.

24 ms per clip#1 commercial latencySee the Podonos benchmark

The 24 ms figure is Aletheia’s vendor-reported latency on the independent Podonos audio deepfake benchmark. It is not a measured latency or accuracy result for call-recipient classification.

Lexis: verified voice workflows.

Check the caller, apply access controls and pass exceptions to your team. The example below follows a payment request through those checks.

LexisSpeechVoice agents
  1. Noesis

    Classify the recipient

    Identify a person, IVR, voicemail or another agent and select the appropriate route.

  2. Aletheia + Eidos

    Check the caller

    Inspect authenticity and compare the enrolled voice before a sensitive request.

  3. Workflow rules

    Apply access controls

    Check permitted actions, tool access and additional verification requirements.

  4. Lexis

    Act or hand off

    Continue in scope, or pass the conversation and the evidence to your team.

Lexis · payment confirmation
ConversationExample
  1. NoesisHuman recipient · conversation route selected
  2. Lexis

    Thanks for calling. How can I help today?

  3. Caller

    I need to move $5,000 to a new payee.

  4. Access rulesNew payee · additional verification required
  5. AletheiaSynthetic score 0.93 · voice flagged
  6. EidosSpeaker match 0.86 · authenticity check failed
  7. Lexis

    I’ll connect you with a specialist to review this request.

  8. OutcomePayment held · call and evidence sent to a specialist

Built for the operating decision.

Defined conversations

Build around bounded intents, not an unrestricted agent.

Human escalation

Set triggers, ownership, and operator context.

Decision evidence

Retain the workflow state needed for review.

Plan the deployment.

Bring your traffic, orchestration stack and routing policy. Start with a scoped evaluation of classification quality, hand-offs and inference cost.

Talk to the team
  1. Which destinations dominate your calls: humans, IVRs, voicemail or other agents?
  2. Should classification run once, or continue throughout the call?
  3. Which models should receive human and automated conversations?
  4. Where must audio run, and what evidence should the routing decision retain?

Evaluate your call traffic.

Free evaluation on your live traffic. No commitment.