AI Assistant for Microsoft Teams Meetings

There are two ways to add an AI assistant to a Microsoft Teams meeting. You can use the built-in Copilot, tied to your Microsoft 365 subscription, or you can run a desktop assistant that captures system audio, works independently of your license, and never joins the call as a participant. For teams standardized on Microsoft, Copilot is the obvious start. For anyone who wants live help, license-free access, or their audio kept off Microsoft's cloud, the desktop route fits better.

I am the founder of Natively, an open-source (AGPL-3.0) desktop AI assistant for meetings and interviews. Teams is the platform where the built-in option is strongest, so this guide is honest about where Copilot wins and where a separate assistant still earns its place.

Where Microsoft Copilot is genuinely good

I am not going to pretend Copilot is weak. Inside a Microsoft 365 shop it is the most frictionless option there is. It summarizes the meeting, surfaces action items, and connects to the documents, chats, and calendar that already live in your tenant. If your admin has enabled it and your whole team is on Microsoft, turning it on is the sensible default and I would not argue you out of it.

Its real strength is governance rather than magic. Enterprise IT cares about admin controls, identity, compliance boundaries, and data residency, and Copilot answers those questions inside a system the organization already manages. That is worth a lot to a large company, and it is something a small independent tool cannot replicate.

Where a separate assistant still wins

Three gaps make me keep a desktop assistant running even where Copilot exists.

The first is the license wall. Copilot is gated behind a paid add-on that not every seat has. If you are a contractor, a job candidate, or on a plan that does not include it, Copilot simply is not available on that call. A desktop assistant does not care what Microsoft plan anyone is on, because it captures the audio your own machine is playing.

The second is the live moment. Like most built-in tools, Copilot is built around the recap. It tells you what happened after the call. It is not sitting on your screen feeding you an answer while someone is waiting for one. If the value you want is help during the conversation, that is a different product, and the two can run side by side.

The third is where the data goes. Copilot processes in Microsoft's cloud under your tenant's controls, which is fine for internal calls and a harder sell for a confidential interview or an external negotiation. A desktop assistant that can run locally or on your own key keeps that conversation off a shared cloud entirely.

Setting up Natively on Teams

There is nothing to install inside Teams and no admin approval to request. You run the desktop app alongside Teams and it listens to the audio your machine is already producing, whether you use the Teams desktop client or Teams in a browser.

Open Natively before the meeting. Approve the one-time system-audio permission your operating system asks for. Join the Teams call the way you always do. Transcription starts when people talk, and the assistant window stays visible to you while hidden from any screen you share. Because capture happens at the system level with a Rust-based audio pipeline, the same setup works across Teams, Zoom, Meet, Slack, and Webex without changing anything.

One Teams-specific note. In organizations with strict policies, meeting recording and transcription may be governed by tenant rules and regional law. An assistant being invisible to other participants does not change your obligations, so follow company policy and get consent where required, especially on external calls.

Running both without stepping on each other

The framing of Copilot versus a desktop assistant is a little false, because the most practical setup is both. They solve different halves of the problem and do not conflict.

On an internal Teams call where everyone expects a record, let Copilot produce the shared recap that lands in the meeting notes and the calendar event. That is the artifact the team references later. Meanwhile the desktop assistant sits on your screen for the live layer only you see: the answer to a question you did not expect, the follow-up you want to ask, the number you half-caught. One is the team's memory, the other is your in-the-moment support.

Because the desktop assistant captures system audio rather than joining the call, Copilot does not know it is there and there is nothing to reconcile. You are not running two bots that both announce themselves. You are running one official recap tool and one private assistant, and the other participants only ever see the official one.

When capture does not work, and the fix

Two setups trip people up on Teams, and both are quick to fix if you know what you are looking at.

The first is a transcript that only shows your own voice. That means the assistant has your microphone but not the system audio carrying the other participants. On macOS this is almost always the one-time screen-and-audio permission that has not been granted; approve it in system settings, restart the assistant, and both sides appear. Do it well before an important call, not during one.

The second is Teams grabbing exclusive control of your audio device in some configurations. If capture drops the moment a Teams call connects, switching your system output to a standard device, your laptop speakers or plain wired headphones, rather than a virtual or conferencing-specific device, usually clears it. It is a routing quirk, not a limitation of the assistant, and it affects any tool that captures system audio.

Frequently asked questions

Can I use an AI assistant on Teams without Copilot?

Yes. A desktop assistant like Natively captures system audio independently of Microsoft 365, so you get transcription and live help on Teams without a Copilot license and without a bot joining the call.

Does it work with the Teams desktop app and the web version?

Both. Because capture happens at the operating-system level rather than inside Teams, it behaves the same whether you run the Teams desktop client or Teams in a browser.

Will other people in the Teams call see the assistant?

No. There is no bot in the participant list, and Natively's overlay stays hidden from screen share, so what you see stays on your side of the call.

How is it different from Microsoft Copilot?

Copilot is tied to your Microsoft 365 license, processes in Microsoft's cloud, and mainly summarizes after the call. Natively needs no license, can keep processing local or on your own key, and helps live during the call. The two can run side by side.

Try it on your next Teams call

If Copilot already covers your internal recaps, run a desktop assistant on the calls Copilot does not reach: an external interview, a candidate screen, a client negotiation where the audio should stay yours. That is where the gap is most obvious.

Natively is free to try with your own key or a local model, and it works on Teams, Zoom, and Meet the same way. For the full picture across platforms, start with the complete guide to AI meeting assistants.