AI meeting notes:minutes you can rely on,without legal trouble
AI tools transcribe meetings, summarise them and list the agreed tasks. This guide covers the main types of tool, a step by step approach, what needs to be settled before anyone records, and where a person should check the result.
Get in touch- How does AI turn a meeting into minutes?
- Which tools can take meeting notes for you?
- What is a sensible workflow for AI meeting minutes?
- Which prompts produce minutes you can use?
- When may AI take notes in a meeting?
- Where do AI meeting notes most often go wrong?
- Who should check the minutes before they go out?
- Does your team need training to use AI note takers?
- Frequently asked questions
- How to get started with AI meeting notes
- Where the information on this page comes from

Let's talk about your project.
First we check whether the project fits your business model. Then you get a proposal with phases and effort.
Discuss AI meeting notes or call: +49 151 1576 5566AI meeting notes are produced in two steps: a tool turns speech into text, and a language model condenses that text into outcomes, decisions and action items. You can do this inside Teams, Google Meet or Zoom, or afterwards by giving a transcript to an AI assistant such as ChatGPT or Claude. Before you switch on the note taker, settle two things: whether everyone in the meeting knows and agrees, and who checks the minutes before they are sent.
- Teams, Google Meet and Zoom each offer their own AI summaries, built on a transcription of the meeting.
- In Germany, recording someone’s private spoken words without authorisation is a criminal offence under Section 201 of the Criminal Code. Get consent first.
- Under the GDPR you need a legal basis, prior information for participants and, for cloud services, a data processing agreement.
- Treat the minutes as a draft: names, figures, owners and decisions are checked by someone who attended before anything goes out.
On this page
- How does AI turn a meeting into minutes?
- Which tools can take meeting notes for you?
- What is a sensible workflow for AI meeting minutes?
- Which prompts produce minutes you can use?
- When may AI take notes in a meeting?
- Where do AI meeting notes most often go wrong?
- Who should check the minutes before they go out?
- Does your team need training to use AI note takers?
- Frequently asked questions
- How to get started with AI meeting notes
- Where the information on this page comes from
How does AI turn a meeting into minutes?
AI minutes come from two separate jobs, and it helps to keep them apart. First, speech is converted into text, which gives you a transcript. Then a language model reads the transcript and condenses it: what was discussed, what was decided, and who takes on which task by when.
The distinction matters because errors creep in at both stages. If speech recognition mishears a name or a number, the summary carries the mistake forward in clean, confident prose, which makes it hard to spot. Guides to Teams say so plainly: accuracy depends on audio quality and on how well speakers are separated, while crosstalk and jargon make it worse.
In practice, good minutes start with good audio. A decent microphone, one person speaking at a time and clear introductions at the start do more for the result than any instruction you give the AI afterwards.
Which tools can take meeting notes for you?
There are two routes. Either your video conferencing tool takes the notes itself, or you hand an existing transcript or your own notes to an AI assistant afterwards.
Microsoft Teams: the AI summary is called a recap. According to Microsoft, it requires the meeting to be transcribed. It covers key points, decisions, action items and highlights in the transcript. When someone starts transcription, every participant sees a notice.
Google Meet: the “Take notes for me” feature powered by Gemini saves notes to a Google Doc, with a summary and next steps. Google tells all participants that notes are being taken. It works in German and English, but only one language per meeting.
Zoom: AI Companion creates a summary from what was said and sends it by email after the meeting. The host decides who receives it.
AI assistant with a transcript: if you already have a transcript or rough notes, ChatGPT, Claude or Microsoft Copilot can shape them into minutes that follow your own template. This is the most flexible route, but company data belongs in a business account, as explained in our guide to AI assistants for business.
| Route | Where the minutes appear | Strength | What to watch |
|---|---|---|---|
| Teams with Copilot | In the meeting recap | Action items and decisions straight from the transcript | Transcription must be on, everyone sees a notice |
| Google Meet with Gemini | In a Google Doc | Notes can be edited and shared at once | One language per meeting |
| Zoom AI Companion | In an email after the meeting | Summary with no extra effort | Set recipients in advance, share only after review |
| AI assistant with a transcript | In the chat, using your own template | Full control over structure and tone | Business account only, no training on your data |
What is a sensible workflow for AI meeting minutes?
The workflow is much the same whichever tool you use. Set it up properly once and every later meeting takes less effort.
- 01
01
Announce itSay in the invitation that AI will take notes, what the minutes are for and who will receive them. Ask again at the start whether anyone objects.
- 02
02
Fix a templateDecide which sections every set of minutes contains: attendees, topics, decisions, action items with owner and deadline, open questions.
- 03
03
Start the note takerSwitch on transcription or note taking only once everyone has agreed. Ask people to speak one at a time and to say names clearly.
- 04
04
Review the draftSomeone who attended checks names, figures, dates and owners, and above all whether something was actually decided or merely discussed.
- 05
05
Approve and tidy upOnly reviewed minutes go to the distribution list. Delete audio and transcripts according to the retention periods you set in advance.
Which prompts produce minutes you can use?
If you work from a transcript with an AI assistant, the prompt decides the quality. A vague request such as “Summarise this” gets you a vague text. A prompt that sets out structure, audience and limits works far better.
For action minutes: “Turn the transcript below into action minutes. Structure: attendees, topics discussed, decisions made, a table of action items with task, owner and due date, and open questions. Only include what is in the transcript. If no owner or date was given, write ‘not agreed’ rather than filling one in.”
For readers who were not there: “Write a summary of no more than ten sentences for the management team. Start with the decisions, then the risks, then the next steps. No opinions, no guesses.”
For the review before sending: “List every name, number, amount and date in these minutes and show the passage in the transcript each one comes from.” This list speeds up the check, because you go straight to the places where a mistake would be costly.
The key rule is in the first example: tell the AI explicitly not to fill gaps. Language models like to complete missing details with plausible content, and in minutes that quickly becomes a commitment nobody actually made.
When may AI take notes in a meeting?
Two areas of law apply here for businesses in Germany: criminal law and data protection. The points below are not legal advice. They show which questions to settle beforehand.
Criminal law: under Section 201(1) of the German Criminal Code, anyone who without authorisation records the privately spoken words of another person faces up to three years in prison or a fine. An internal meeting is normally not public. If a tool stores the audio, even temporarily, consent from everyone involved is the safe route. Whether a pure live transcription that only processes sound in passing already counts as a recording under Section 201 is disputed in legal commentary, so do not rely on it without legal review.
Data protection: minutes containing names and contributions are personal data. You need a legal basis under Article 6 GDPR. Consent from employees is problematic, because it is rarely truly voluntary towards an employer. If you rely on legitimate interests, document the balancing test. Tell participants in advance who will receive the minutes and how long audio and text will be kept. With cloud services, a data processing agreement under Article 28 GDPR is part of the package.
Works council: if your company in Germany has a works council, it has a say under Section 87(1) no. 6 of the Works Constitution Act in technical systems intended to monitor employees’ behaviour or performance. Legal experts recommend covering transcription tools in a works agreement.
The tools help with transparency. Teams shows everyone a notice as soon as recording or transcription starts, and administrators can require explicit consent from participants, a setting that is off by default. Google Meet shows everyone a pencil icon while Gemini is taking notes.
Where do AI meeting notes most often go wrong?
The typical errors are predictable, which means you can check for them on purpose.
Names and jargon: speech recognition does not know your people, product names or internal abbreviations. It writes a similar sounding word instead, which sometimes even makes sense in context.
Who said what: attributing a comment to a person depends on clean speaker separation. When several people share one microphone in a meeting room, attribution gets blurred.
Decided or just discussed: an idea floated in the room easily turns into a decision in the summary. This is where the most damage is done, because minutes are later treated as a record.
Tone: irony, hesitation and a “let’s see” arrive in the text as a firm statement.
Several languages: Google states that Meet supports only one language per meeting. If your team switches between German and English, expect gaps.
Who should check the minutes before they go out?
Decide in advance who approves them. A practical rule is that the person who chaired the meeting reviews the draft within a day, while memories are fresh. Set automatic summaries to go to that person first rather than to everyone.
Do not check every sentence with the same care. Focus on the places where errors have consequences: action items with owners and deadlines, amounts and quantities, dates, commitments to customers or suppliers, and anything phrased as a decision. An extra sentence about how a discussion went rarely hurts. A task assigned to the wrong person does.
Add a line at the end of each set of minutes saying that they were prepared with AI support and who reviewed them. That makes things clear for recipients and shows who is responsible. For recurring meetings such as project reviews, it pays to move the action list straight into the tool your team works in. Our guide to AI for project management shows how.
Take particular care with performance reviews, job interviews and confidential negotiations. In these conversations, the interest of the people involved in not being recorded often outweighs the convenience. For conversations with staff and applicants, see AI for HR and recruiting.
Does your team need training to use AI note takers?
Yes, to an appropriate extent. Article 4 of the EU AI Act has applied since 2 February 2025 to organisations that use AI systems. The European Commission makes clear that this includes companies whose staff use an AI chat to write text, and gives hallucination as an example of a risk people should know about. Since 27 July 2026 it applies as amended by the Digital Omnibus, Regulation (EU) 2026/1744: you must take measures that support the development of AI literacy, but you do not have to guarantee a particular level for each person. No certificate is required. More in our article on the AI literacy obligation under Article 4.
For meeting minutes this means that anyone who uses a note taker should know when to ask first, which parts to check and where audio and text are stored. If you hold especially confidential meetings, consider whether transcription and summaries can run without the cloud. Our guide to local AI for business covers the options.
Frequently asked questions
Can AI write meeting minutes from a recording?
Yes. Teams, Google Meet and Zoom create summaries directly in the meeting. Alternatively, give the transcript and a template to an AI assistant such as ChatGPT or Claude. The recording must have been made lawfully, and company data belongs in a business account.
Do I have to ask participants before AI takes notes?
If a tool stores the audio, German law (Section 201 of the Criminal Code) requires consent from everyone involved. Whether a purely transient live transcription is covered is disputed. Separately, the GDPR requires you to inform participants in advance. So ask first in every case.
Will other participants notice that AI is taking notes?
In Teams, everyone sees a notice as soon as recording or transcription starts. Google Meet shows everyone a pencil icon while Gemini is taking notes. These notices do not replace an announcement in the invitation.
How accurate are AI meeting notes?
Mostly as accurate as the audio. Crosstalk, weak connections and several people on one microphone degrade the transcript, and the summary inherits its mistakes. Always check names, figures, dates, owners and decisions by hand.
Do we need a works agreement?
If your German company has a works council, it has co-determination rights under Section 87(1) no. 6 of the Works Constitution Act for technical systems intended to monitor behaviour or performance. Legal experts recommend covering transcription tools in a works agreement.
How long can we keep recordings and transcripts?
Only as long as you need them for the stated purpose. Set retention periods in advance, keep audio and text separate and delete the audio once the reviewed minutes exist. The periods belong in the information you give participants.
How to get started with AI meeting notes
- 01
Write a template
Decide which sections every set of minutes contains and who approves them.
- 02
Settle the legal side
Agree the legal basis, participant information, retention periods and, where relevant, works council involvement with your data protection officer.
- 03
Test with one meeting
Try the tool in an internal meeting where everyone has agreed, and compare the draft with your own notes.
- 04
Prepare your team
Show your team when to ask first and which parts to check. We can help with AI training built around your own meetings.
AI meeting notes save you the note taking, not the responsibility. Ask first, use a fixed template and check the sensitive parts, and you get reliable minutes in a fraction of the time.
Where the information on this page comes from
- Section 201 StGB, violation of the confidentiality of the spoken word (gesetze-im-internet.de)retrieved 26 Sep 2026
- dejure.org: Section 201 StGBretrieved 26 Sep 2026
- Dr. Datenschutz: KI-Transkription im Meetingretrieved 26 Sep 2026
- Deutscher AnwaltSpiegel: KI-Transkription im Unternehmenretrieved 26 Sep 2026
- Section 87 BetrVG, co-determination rights (gesetze-im-internet.de)retrieved 26 Sep 2026
- General Data Protection Regulation (EUR-Lex)retrieved 26 Sep 2026
- gdpr-info.eu: Art. 28 GDPR, Processorretrieved 26 Sep 2026
- Microsoft Support: Recap in Microsoft Teamsretrieved 26 Sep 2026
- Nexacu: Copilot in Microsoft Teams, meeting summaries, recaps and action itemsretrieved 26 Sep 2026
- Tygart Media: Copilot in Microsoft Teams, meeting summaries and action itemsretrieved 26 Sep 2026
- Microsoft Support: Start, stop and download live transcripts in Teams meetingsretrieved 26 Sep 2026
- Microsoft Learn: Custom in-meeting notification for recording and transcriptionretrieved 26 Sep 2026
- Topedia: Customize consent notifications for Teams recording and transcriptionretrieved 26 Sep 2026
- Google Meet Help: “Take notes for me” in Google Meetretrieved 26 Sep 2026
- tl;dv: How to take notes with Gemini on Google Meetretrieved 26 Sep 2026
- Zoom Support: Using meeting summary with AI Companionretrieved 26 Sep 2026
- University of Colorado: Creating Zoom meeting summaries with the AI Companionretrieved 26 Sep 2026
- Regulation (EU) 2024/1689, AI Act (EUR-Lex)retrieved 26 Sep 2026
- Regulation (EU) 2026/1744, Digital Omnibus on AI (EUR-Lex)retrieved 26 Sep 2026
- European Commission: AI Literacy, Questions and Answersretrieved 26 Sep 2026
- ITMR: Digital Omnibus zur KI, was am AI Act verändert wirdretrieved 26 Sep 2026
- Bundesnetzagentur: KI-Kompetenzretrieved 26 Sep 2026

