Two hours into a product training, half the room stopped following around slide five. Not because the content was bad. They were listening in their second language and holding real focus in a language that isn’t yours for two hours straight is exhausting.

Interpreter Agent solves exactly that problem and almost nobody at international companies is using it yet.

❓ What it actually is

Interpreter Agent translates a Teams meeting live into each participant’s own language.

Turning it on takes three clicks: More, then Language and speech, then Turn on Interpreter, followed by picking your language from a dropdown. Teams remembers that choice for next time, so it’s a one-time setup per person, not something you repeat every meeting.

The more interesting decision is the voice. You can let Interpreter simulate your own voice when it translates you, matching your tone and speech pattern or fall back on one of three automated voices. Voice simulation needs your explicit consent first. Your voice data will not be stored unless you explicitly say yes. Skip the consent and Teams defaults to Ava.

Beyond the voice, it also runs in two distinct modes and they solve different problems.

Simultaneous translation is the default mode, live since last year. The agent translates continuously while the speaker keeps talking, no pausing required. Built for broadcast meetings: trainings, webinars, product launches, anything with one speaker and many listeners.

Consecutive interpretation is new. The speaker finishes a thought, pauses and the agent delivers the translation to the whole group before the next person responds. It’s turn-based by design, built for an actual back-and-forth conversation across one language pair (English and German, for example) not a one-to-many broadcast. It’s in public preview now for meeting organizers with a Microsoft 365 Copilot license, with general availability scheduled summer 2026.

Where to actually start

Enabling Interpreter is an admin toggle, the same as the other Phase 1 features in Microsoft’s AI-Powered Workplace framework.

Start with simultaneous translation in a broadcast meeting: a training or an internal webinar where nobody’s negotiating anything. Get comfortable there first. Once you are, move up to consecutive interpretation in a more formal meeting that actually involves discussion, not just one person talking.

Test it before you rely on it. Run a dry pass in an internal meeting first, just to get a feel for the pacing, especially consecutive interpretation’s turn-based rhythm. Don’t let a live business meeting be the first time you’re using either mode.

🎙️ My take

I use this most for long training sessions. Attention span runs noticeably longer when people hear content in their own language instead of translating it in their head the whole time. That’s the actual value, not the novelty of live translation. It’s sustained focus over an hour or two, exactly when most training content loses people anyway.

Two things are holding adoption back. Most people don’t know Interpreter Agent exists. The ones who do are hesitant to actually turn it on. That’s the same story as Facilitator right now: a genuinely useful feature nobody switches on because it defaults to off and turning it on feels like a decision.

Turn it on for one training session. That’s the whole ask. The barrier doesn’t come down because Microsoft shipped a feature, it comes down the first time you actually use it.

⏩ TL;DR

• Interpreter Agent runs two modes: simultaneous translation (default, for broadcast meetings) and consecutive interpretation (new, turn-based, for back-and-forth conversations).
• Consecutive interpretation is in public preview now, general availability scheduled for summer 2026.
• Start with simultaneous in a broadcast meeting, then move to consecutive once you’re comfortable.
• Always test both modes internally before a real meeting, especially consecutive’s turn-based rhythm.
• Awareness and hesitation block adoption more than quality does.