Accordia Blog
Field notes on AI voice agents, contact centre automation and multilingual customer operations across Southeast Asia.
Malaysians have been trained by two years of deepfake fraud to distrust an unfamiliar voice on an unknown number. That is now the first obstacle every legitimate outbound voice agent has to clear — and most are designed as if it does not exist.
Callers in Malaysia, Singapore and much of Southeast Asia switch languages mid-sentence — and most voice agents are built to pick one language per call. Here is why that breaks, and how to design a voicebot that mirrors the caller instead.
As of 2 August 2026 the EU requires AI systems to identify themselves to the people they speak to, and Malaysia's first horizontal AI statute is in drafting. Here is what a compliant voice agent actually says in its first ten seconds — in English and in Bahasa Malaysia.
Containment rate tells you how many callers never reached a human — not how many got what they called for. Agentic voice agents that complete real tasks need a resolution-first scorecard, and most teams are still reporting the wrong number.
The EU AI Act's new transparency requirements took effect on 2 August 2026, changing how businesses deploy customer-facing AI. Here's what the new rules mean for voice AI agents, automated customer interactions and enterprise AI deployments.
If a voice AI vendor tells you to "just bring your own SIP trunk," here's what that actually means, how call routing works once it's connected, and what to check before you commit.
A voice agent that can only talk isn't that useful on its own. The real value shows up once it can look something up or write something back mid-call, which means giving it scoped, auditable access to your CRM.
Everyone assumes the model is the hard part of building a voice AI agent. In practice, it's almost always something smaller: barge-in, background noise, code-switching, and silence that times out too early.
Malaysian call centers juggle four languages in a single queue, and staffing that around the clock is expensive. Here's why AI voice agents caught on here faster than in most other markets, and where they still fall short.
