Voice AI pilots routinely deliver impressive demonstrations, yet the transition to production reveals a persistent gap between promise and performance. AudioCodes' Ilan Avner articulates the core problem: enterprises systematically underestimate the infrastructure demands of voice AI at scale, treating the underlying technology stack as secondary to the AI model itself. The reality is that connectivity, latency, noise handling, scalability, redundancy, and human-agent handoff are not peripheral concerns—they are foundational requirements that determine whether a voice AI deployment succeeds or stalls. This distinction matters acutely for teams already managing complex omnichannel environments, where voice volumes are climbing and the pressure to consolidate silos is intensifying. The question becomes whether your current infrastructure—whether built on Zendesk, Freshdesk, or proprietary systems—can actually support the operational demands of voice AI, or whether you are optimising for a use case your platform was not designed to handle.
The stalling of voice AI projects after pilot phases suggests a systematic failure in procurement and implementation planning. Teams often evaluate voice AI solutions in isolation, focusing on conversation quality and accuracy metrics, only to discover that production environments demand resilience patterns, failover mechanisms, and integration complexity that were invisible during controlled testing. Platforms like AudioCodes Live Hub attempt to address this by bundling infrastructure, agent assistance, and real-time translation into a cohesive production environment, but the broader implication is that voice AI success now requires architectural thinking, not just model selection. For CX leaders, this means the vendor evaluation process must shift: technical due diligence on infrastructure capability should precede any pilot agreement, and success metrics should include uptime, handoff quality, and latency under load—not merely accuracy on clean audio samples. The contact centre leaders racing to deploy voice AI should be asking whether their chosen platform has genuinely solved the production problem, or whether they are simply repeating the pilot-to-stall cycle with a different vendor.
In this CX Today interview, Marcus Law speaks with Ilan Avner, Director of Product Management at AudioCodes, about the practical reality of moving Voice AI from proof of concept into production. The conversation explores why so many Voice AI projects stall after promising pilots, what enterprises of
The Reality of Voice AI in the Contact Centre: From Pilot to Production CX Today