Enterprise DNA

Omni by Enterprise DNA

Enterprise DNA Resources

Latest AI and industry news. Practical AI operating-system thinking for owners, operators, and teams doing real work.

220k+

Data professionals

Omni

AI agents and apps

Audit

Map the manual work

News Trending Product

Microsoft MAI-Realtime: A Full-Duplex Voice Model in Testing

Microsoft is testing MAI-Realtime, a full-duplex voice model that listens and speaks simultaneously with no turn-taking required.

Enterprise DNA | | via TestingCatalog
Microsoft MAI-Realtime: A Full-Duplex Voice Model in Testing

Microsoft has been quietly testing a new voice model — and if early reports are accurate, it could be one of the more significant moves in enterprise voice AI this year.

TestingCatalog reported on August 2 that an upcoming model called MAI-Realtime has surfaced as a hidden early-access entry in the Microsoft MAI Playground. A small group of partners are testing it now, with no public launch date confirmed.

What Makes It Different

Most voice AI systems work in turns: the AI speaks, then waits, then listens. MAI-Realtime is bidirectional — it handles both directions simultaneously. That means it can process what you are saying while it is still speaking, interrupt naturally, and respond to cues mid-sentence. The technical term is full-duplex audio.

This is not a minor upgrade. Turn-based voice systems feel robotic because they are. Full-duplex voice is what separates a natural conversation from a customer support phone tree.

According to the report, MAI-Realtime also:

  • Runs web search and other tools during a live conversation — not just answering from a knowledge base
  • Supports 18 languages including English, German, Spanish, French, Japanese, Korean, Chinese, Hindi, Arabic, and more
  • Features two initial voices — Victoria and Grant — described as noticeably more natural than Microsoft’s current MAI Voice 2
  • Includes configurable interruption handling so businesses can tune how the model responds to mid-sentence corrections

No pricing or general availability date has been disclosed.

Why Microsoft Is Doing This Now

Microsoft has been building its MAI model family throughout 2026: MAI-Code-1-Flash for coding at Build 2026, MAI-Cyber-1-Flash for cybersecurity in late July, and now MAI-Realtime for voice. Each model targets a specific enterprise use case where Microsoft wants to offer a native alternative to OpenAI’s products.

The voice AI market is moving fast. OpenAI has GPT-Realtime 2.1. xAI has Grok voice. Competitors like ElevenLabs, Deepgram, and a cluster of enterprise voice startups are all building toward the same goal: natural, tool-enabled, enterprise-grade voice agents.

Microsoft’s advantage is distribution. Azure, Copilot, Teams, and Dynamics are already embedded in enterprise workflows. Adding a native voice model to that stack — one that can run web searches and call external tools — is not a product launch. It is infrastructure.

What This Means for Business

If you are evaluating voice AI for your business right now, three things are worth noting.

The floor is rising fast. What counted as good voice AI 12 months ago sounds robotic by current standards. Any deployment you are planning should be tested against full-duplex systems, not turn-based ones, because that gap will become obvious to your customers.

Microsoft’s entry raises enterprise legitimacy. When Microsoft ships a voice model, procurement teams stop asking “is this real?” and start asking “how do we deploy it?” For businesses that have been waiting for voice AI to mature, MAI-Realtime entering even a private preview is a signal that maturity is arriving.

Tool integration is the differentiator. The detail that MAI-Realtime can run web search and other tools mid-conversation is more significant than the voice quality improvement. A voice agent that can look up a current price, check a calendar, or verify an account in real time is a different product category from one that answers from static knowledge. That is what makes it useful for sales, support, and internal operations — not just demos.

For businesses thinking about deploying voice AI employees — handling inbound inquiries, managing internal helpdesks, or automating routine customer interactions — the direction is clear. The technology is converging on natural, capable, tool-enabled conversation. The question is not whether to adopt it but how to position the rollout when it arrives.


Enterprise DNA builds Omni Voice — AI voice employees for enterprise knowledge discovery, internal reporting, admin automation, and team communication. If you are ready to explore what voice AI can do for your business, book a discovery call with our team.