The App of the Future Is Voice — Not a Screen. Mitel's CTO Luiz Domingos Explains Why.

What It Covers

Mitel CTO Luiz Domingos outlines how enterprise communications is being transformed by AI across contact centers, unified communications, and agentic workflows, arguing that voice will replace traditional screen-based interfaces and that hybrid edge architectures are becoming essential for regulated industries managing latency and data sovereignty.

Key Questions Answered

• •Edge AI for Latency: Deploy AI inference at local data centers or on-premises rather than public cloud for real-time voice applications. In contact center environments, a two-second delay from cloud round-trips makes AI assistance unusable during live calls. Edge deployment eliminates network dependency and keeps sensitive conversation data within enterprise boundaries, addressing both performance and sovereignty requirements simultaneously.
• •RAG Over Generic AI: Enterprises should implement retrieval-augmented generation rather than plain generative AI to extract real business value. Generic models lack enterprise-specific context, making outputs too broad to be actionable. Connecting language models to internal knowledge bases, product documentation, and workflow data produces measurable ROI — Mitel applies this to support routing, agent assistance, and intelligent call handling across 70 million users.
• •Agentic Governance Before Deployment: Before deploying agentic AI workflows, enterprises must define accountability frameworks specifying where human-in-the-loop approval is required. CIOs at industry events are prioritizing governance and auditability over capability features. Agentic systems that autonomously create tickets, update CRMs, or trigger workflows carry reputational and legal liability if unchecked — enterprises need explicit decision-point documentation and audit trails built into architecture.
• •Modernization Prerequisite: Organizations cannot layer AI onto legacy fragmented architectures and expect transformation. Effective AI adoption requires becoming API-first, decoupling the communications layer from the workflow layer, building modular AI services for data ingestion and orchestration, and establishing clean data pipelines with clear governance — gradual hybrid integration outperforms full system replacement while avoiding structural bottlenecks that cap AI value extraction.
• •Voice as the Default Interface: Enterprise applications will shift from screen-based GUIs to voice-first interactions within the near-term horizon. Frontline workers — nurses, field technicians — represent the most underserved segment of digital transformation and benefit most from voice AI that eliminates screen dependency. Transcription and summarization, currently premium features, will commoditize into standard subscriptions the same way call recording transitioned from paid add-on to baseline capability.