AI meeting note takers for New Zealand teams.

How do the main AI note takers compare?

Seven tools cover most of what New Zealand teams use. Three are built into meeting platforms; four are standalone and work across platforms. As listed on 20 October 2026:

Fireflies, Fathom and Granola show prices with a bare “$”, so confirm currency and GST at checkout.

Built-in or standalone: which do you need?

Start with what you already pay for. If your meetings happen in one platform, its built-in note taker keeps notes inside your existing tenant, under admin controls you already run, with no extra vendor to assess. Standalone tools earn their place when meetings span platforms or the notes must feed a CRM.

Choose a built-in note taker if

• Most meetings run in Teams, Meet or a paid Zoom plan.
• You want notes stored with your other Microsoft 365 or Google Workspace data.
• You would rather not add another vendor to your privacy impact assessment.

Choose a standalone tool if

• Staff join client meetings on Zoom, Teams and Meet in the same week.
• You need notes pushed into a CRM such as HubSpot or Salesforce automatically.
• You need a language the built-in tools lack, and you have tested it.

Our comparison of Copilot vs Gemini covers the wider suite decision.

Do you need consent to record a meeting in New Zealand?

Assume you need to tell people. The Privacy Commissioner says recording someone without telling them may breach the Privacy Act, depending on who records and why. Recording on behalf of an organisation brings in the general collection rules, and “it might be unfair to record someone without telling them”. The Commissioner also notes that recording a private conversation you are not part of could, depending on the circumstances, be a crime. This is general information, not legal advice.

Three Privacy Act points apply to most business meetings:

• IPP 3. When you collect information from people directly, take reasonable steps so they know it is being collected and why.
• IPP 3A. Since 1 May 2026, information collected about someone not in the room, such as a client describing an employee, can carry its own notice duty.
• Offshore vendors and training. A standalone tool that trains on your meetings is using them for its own purposes. The Commissioner’s AI guidance expects a privacy impact assessment before an AI tool is used.

Vendors put the duty on you: Otter’s terms say users “must always ask for consent”, and Fathom asks you to hold “the necessary permissions and consents” before recording.

Should a note taker join external client calls?

Only with the client’s agreement. A bot appearing uninvited in a client’s meeting can damage trust, and the client may simply remove it. Teams, Meet and Zoom show participants that notes or transcripts are running, and Meet lets participants from the host’s organisation stop note-taking. Granola adds no bot at all, so nobody else can see it: disclosure is entirely your job.

• Ask in the invite: “We use [tool] to take notes; notes are stored in [place]. Happy to switch it off.”
• Never send a bot to a meeting you are not attending.
• Pause capture for salary, health, disciplinary or legal discussion.
• Share the notes with attendees so errors get corrected.
• Delete recordings on a schedule you have written down.

How well do they handle New Zealand accents and te reo Māori?

Only what vendors document can be said with confidence, and we have not tested any of them on te reo Māori or New Zealand accents. Teams transcription lists English (New Zealand) as a spoken language. Fireflies lists Māori among more than 100 languages. Google Meet’s notes support eight languages, one per meeting, and Granola, Fathom and Otter publish lists without te reo Māori. The Privacy Commissioner’s guidance records concern that systems built overseas may not work accurately for Māori.

Test before you rely on any of them:

1. Pick three recorded internal meetings with consent: one routine, one with heavy jargon, one where te reo Māori or strong regional accents are spoken.
2. Run each tool on the same recordings and keep a human-written action list for each meeting.
3. Count errors that matter: wrong owners, wrong dates, misspelt names and place names, invented decisions.
4. Decide per meeting type, not per tool. Our lesson on turning meeting notes into actions shows how to check an action table against its source.

What do they cost for a 10-person team?

Worked example at list prices on 20 October 2026, annual billing where offered.

The licence is rarely the real cost. Illustrative: ten people each spending five minutes fixing notes from four meetings a week is over three hours of staff time a week. Accuracy on your meetings matters more than a few dollars a seat.

Where AI note takers go wrong

• Invented commitments. Summaries can turn “we should look at that” into an action with an owner. Check against the transcript; see using AI for meeting preparation for the other end of the meeting.
• Language limits. Meet does not support several languages in one meeting, so bilingual hui need human notes or another tool.
• Free plans at work. Free tiers often carry training defaults that business plans do not.
• Everything recorded. Transcripts are discoverable records. Write retention rules into your AI use policy.
• Notes nobody uses. Pipe actions into the system where work happens, as in meeting actions into project systems, and train the team on what good notes look like through our AI training for business teams.

Written by the Levy Street team in Wellington. Last updated 19 October 2026.