Keep people in the conversation.
Human recognition belongs beside voicemail detection. Our guarded research candidate correctly recognized 119 of 124 human recordings in the historical cohort.
Explore human recognition →REAL-TIME VOICEMAIL DETECTION
The next call.
A clearer answer.
Human or voicemail? Give your calling workflow a clear signal to make its next move.
Meet Kooya Answer. Audio goes in. A human, machine or unknown result comes back—with measured audio usage and a path into your application.
96.68%
overall accuracy in a historical
1,445-recording research replay ↗
THE ANSWER LAYER
ILLUSTRATIVE RESULTVoicemail detected
machineOne result. Your next move.01 → 02 → 03
BUILT AROUND THE ANSWERHuman recognitionVoicemail detectionClear result contractsAudio usage visibility
WHY AN ANSWER LAYER?
Make the connection
count.
For dialers, voice agents and calling platforms, the moment a call is answered is a decision point. Put voicemail detection into that workflow.
Give voicemail its own path.
Use a machine result to choose a voicemail workflow, log an outcome or hand control back to your dialer. Your application decides what happens to the call.
Understand the result →Measure the work in audio.
The service plan tracks analyzed audio milliseconds and evaluation counts. See the actual detection work, independently of whole-call duration.
See the usage model →ONE SHARED RESULT CONTRACT
An answer your
application can use.
The same outcome vocabulary across the supported audio connections. Select an example to see how your application can use the judgment.
Illustrative payloads, not live classification. Explore the service methods and result paths.
YOUR APPLICATION ← KOOYA ANSWEREXAMPLE
TERMINAL JUDGMENT
Voicemail detected
{
"evaluation_id": "eval_example_01",
"correlation_id": "your-call-reference",
"result": {
"label": "machine",
"audio_analyzed_ms": 6000,
"decision_audio_ms": 6000
}
}Your next move: select your voicemail workflow, then acknowledge the result.
HTTP response · WSS judgment · RTC data · webhookWHERE IT FITS
The missing signal
in your calling stack.
Outbound dialers
Bring the answer classification into campaign routing, agent handoff and disposition reporting.
CALL OPERATIONSVoice AI workflows
Let your voice application choose its next step from the answer signal. Voice generation stays in your existing stack.
VOICE APPLICATIONSTelephony platforms
Evaluate the opening audio of a call with a defined input and result contract. Start with a scoped integration pilot.
PLATFORM INTEGRATIONSWhat is voicemail detection?
Voicemail detection, often called answering machine detection or AMD, analyzes opening call audio to classify an answer as a human, a machine or unknown. Your calling application uses that classification to choose its next step.
How accurate is Kooya Answer?
The guarded candidate recorded 96.68% overall accuracy on a historical 1,445-recording research replay, with 95.97% human recall and 96.74% machine recall. The cohort was machine-heavy. These results describe that research set; your calls need a scoped evaluation. See the full evidence and timing.
Does detection happen instantly?
Detection consumes audio as it arrives, and a result is delivered when the evaluator commits. Timing varies with the audio and configuration. The guarded research replay had a 10.32-second median audio decision point; no subsecond latency guarantee is offered.
How do I connect and receive the answer?
Send a PCM16 WAV evaluation over authenticated HTTPS, stream through authenticated WSS, connect a far-end track using LiveKit WebRTC, bridge inbound SIP through a provisioned trunk, or send authenticated SDES-negotiated SRTP over UDP after creating an HTTPS session. Results return over HTTP, WSS, an RTC data message, a configured HTTPS webhook or session polling. Read the integration details.
How can I get access?
Email Contact the Kooya Answer team to discuss a pilot. There is no self-service customer dashboard or instant signup. Integration, evaluation, usage and service terms are agreed during onboarding.