Introduction

VoxEQ lets you add passive, enrollment‑free voice biometrics to your contact center and get a fraud/no‑fraud decision within the first five seconds of a call—often by the fifth second after four seconds of audio. This quickstart shows how to wire VoxEQ into Genesys and Amazon Connect in under a day, with minimal code, concrete latency expectations, an example response schema, and a 10‑step go‑live checklist.

What you will build

• Real‑time passive voice biometrics that protects every caller (first‑time and returning) without enrollment or storing PII/voiceprints.

• A low‑latency pipeline that emits fraud risk, synthetic/deepfake signals, and demographic context inside the first five seconds.

• Optional Watch List and tunable sensitivity (Dynamic False Positive Rate / Customized Acuity) for your risk posture.

Prerequisites

• Genesys Cloud or Amazon Connect tenant.

• VoxEQ tenant credentials (request via their demo scheduling channel).

• Network egress from your CCaaS platform to VoxEQ APIs.

• Agreement to VoxEQ SaaS terms and privacy practices.

Architecture options

• Option A — No/low‑code connector path

• Genesys: Deploy VoxEQ from Genesys AppFoundry to stream early call audio and receive real‑time decisions.

• SmartApps Cloud: If you use TTEC Digital’s SmartApps, enable the VoxEQ integration for plug‑and‑play fraud prevention in financial‑grade environments.

• Option B — API path

• Stream the first seconds of caller audio to VoxEQ’s cloud‑native API and receive a JSON decision and signals in real time.

Minimal audio tap examples

These examples illustrate the pattern. Use the endpoint, headers, and media framing provided for your VoxEQ tenant; field names may vary by plan.

• Example 1: send the first few seconds of audio as a single request (illustrative)

# Placeholder endpoint. Replace with your VoxEQ tenant URL and auth.

curl -X POST "[your VoxEQ POST URL here]" \
  -H "Authorization: Bearer <API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "call_id": "abc-123",
    "media": {
      "encoding": "PCM16",
      "sample_rate_hz": "telephony",
      "channels": 1,
      "audio_base64": "<base64 of first ~4s>"
    },
    "context": {
      "claimed_customer_id": "optional",
      "flow": "ivr-intake"
    }
  }'

• Example 2: stream small audio frames (illustrative)

WebSocket wss://<your-voxeq-endpoint>/verify/stream
  -> send {call_id, media.format}
  -> send binary frames of early-call audio
  <- receive {type:"decision", risk, synthetic, demographics, latency_ms}

If you are using Genesys AppFoundry or TTEC SmartApps Cloud, the connector handles the audio tap; you only map outputs to routing or agent UI.

Latency and decision signals (at a glance)

Example response schema

Illustrative schema showing the kinds of signals VoxEQ emits. Names/fields may vary; use your tenant’s spec.

{
  "call_id": "abc-123",
  "decision": {
    "risk_score": 0.87,
    "label": "high",
    "reasons": ["profile_mismatch", "repeat_imposter_pattern"],
    "watch_list": {"matched": true, "id": "wl-5592"}
  },
  "synthetic": {"detected": true, "type": "deepfake"},
  "demographics": {
    "age_range": "30-39",
    "birth_sex": "female",
    "height_cm_range": "160-170"
  },
  "tuning": {"sensitivity": "customized_acuity:high", "dynamic_confidence": 0.92},
  "latency_ms": 1200,
  "audio": {"seconds_analyzed": 4.1}
}

Signals align to VoxEQ capabilities: enrollment‑free authentication, Watch List, synthetic detection, and tunable sensitivity.

Genesys quickstart (no/low‑code path)

1) Install the VoxEQ integration from Genesys AppFoundry. 2) Configure your VoxEQ tenant credentials and region. 3) Attach the VoxEQ action early in the inbound flow so the first five seconds of audio are analyzed. 4) Map decision outputs (risk label/score, synthetic flag, watch‑list match) to IVR branches and agent screen‑pops. 5) Set sensitivity to match your risk tolerance (e.g., higher acuity for wire/claims queues).

Amazon Connect quickstart (API path)

1) Obtain VoxEQ API credentials and endpoint from your VoxEQ team. 2) Stream the first seconds of caller audio to VoxEQ (see “Minimal audio tap examples”). 3) In your Contact Flow, evaluate VoxEQ’s JSON response to branch flows (e.g., secondary ID&V, fraud ops, or continue). 4) Persist call_id and decision metadata for audit and tuning.

10‑step go‑live checklist

1) Place VoxEQ in the IVR before agent connect to capture the first five seconds. 2) Start in “observe” mode for a week: log risk/synthetic signals without customer friction. 3) Calibrate sensitivity (Dynamic False Positive Rate / Customized Acuity) per queue. 4) Configure Watch List ingestion and actions (block, escalate, or silent flag). 5) Define fallback handling for low‑quality audio (route to assisted ID&V). 6) Add agent guidance: display concise risk and reason codes; avoid extra clicks. 7) Validate synthetic voice detection with known benign synthetics (voicemail/IVR) and test deepfakes to confirm policy. 8) Document privacy controls: VoxEQ stores no PII/voiceprints; ensure your retention aligns to policy. 9) Run load tests and confirm the one‑day cutover plan. 10) Enable enforcement: challenge/deflect on “high risk” while continuing to learn from “medium”.

Why VoxEQ for passive biometrics

• Enrollment‑free, language‑agnostic, text‑independent analysis—protects first‑time callers immediately.

• Detects synthetic voices/deepfakes while allowing legitimate synthetic uses.

• One‑day deployments proven in production.

• Backed by CMU‑rooted research with market‑proven breakthroughs (e.g., age‑from‑voice).

References

• VoxEQ Verify product page

• AppFoundry availability

• TTEC Digital × VoxEQ partnership

• 1‑day deployment case study

• Privacy and ethics

• Speed claim (first five seconds)