Frequently asked questions

SentraKaya.AI is an enterprise-grade, real-time filter for voice, chat and email, built for BPOs and enterprise contact centres. It tells synthetic AI interactions from authentic human ones and routes each one where it belongs.

The name is pronounced Sen-truh-KAH-yuh.

BPO firms and enterprise contact centres that run AI agents alongside human agents, and need to prove who handled each interaction for compliance, billing or security reasons.

• Compliance and SLA proof. Audit trails show that human agents handled the interactions your clients require them for.
• Billing accuracy. Verify whether an interaction was resolved by AI or handled by a human, so tiered pricing and margins are reported correctly.
• Fraud protection. Stay ahead of deepfake voice manipulation and automated text scripts before a breach happens.
• No friction for genuine customers. No registration, biometric enrolment or extra verification step is needed to assess an interaction.
• Quality visibility. Track escalations, bot failures and copilot usage across human and automated agent pools.

It covers voice calls, live chat and email reaching your contact centre or service centre. It is built to catch AI agents and scripts posing as customers, including deepfake voice manipulation and automated text, before they cause financial or reputational harm.

No single signal decides. Evidence from several layers is combined into one score.

• Voice: acoustic artifacts in the waveform, conversational signals such as how quickly a caller stops talking when interrupted, and telephony and SIP metadata such as carrier and packet timing.
• Chat: response timing, browser keystroke and mouse activity, clipboard pastes and network origin, plus language patterns.
• Email: mail-server headers, sending server origin, message timing and linguistic markers.

See the detection framework for the full list.

For voice, SentraKaya.AI scores the caller within the first 2–3 seconds of the greeting and returns one of three outcomes: human, likely AI or inconclusive. Chat and email are assessed in real time.

A policy engine routes each interaction using the confidence score. In the reference design:

• Under 20% (human): straight to a live human agent queue.
• Over 90% (likely AI): to a low-cost automated queue, an AI workflow or API, or a honeypot bot.
• In between, or high risk: active probing, step-up authentication, fraud review or a block.

Event logs and audit tags flow into your CRM, fraud platform, SIEM and agent desktop dashboards.

SentraKaya.AI is designed to assess whether an interaction is automated, not to identify who a person is. It does not require customers to register or enrol, and it is not designed to build voiceprints or other biometric templates for identification.

Read our Privacy Policy for how personal information is handled, including under POPIA and GDPR.

SentraKaya.AI deploys as lightweight real-time middleware, either through an API or a central analytics dashboard, fed by webhooks and SIPREC. It connects to existing contact-centre workflows wherever audio or text can be captured, for example Twilio Media Streams, AWS Connect Kinesis or Genesys AudioHook, with no database overhaul or extra infrastructure.

Typical time to go live: 1–2 weeks.

SentraKaya Labs is based in Cape Town, South Africa.