Are AI Meeting Note-Takers Safe for Client Calls?

They can be, if tool, settings and consent are deliberate: a note-taker from your business meeting platform or a paid plan, clients told at the start, notes shared only internally and kept for a set period. The real risks are bots that join calls automatically, record people who never agreed and email summaries to everyone on the invite.

Consent rules differ between legal systems. Some let one party to a call record it, others need everyone's agreement, and professional confidentiality duties can go further. So the safe default for client calls is to announce the note-taker every time, offer to switch it off, and switch it off if anyone objects. That habit meets the stricter rules in most places, though regulated professions should check their own guidance.

Follow me on Instagram@sagnikteaches

Where a client call goes once a note-taker hears it

A single 30-minute call with a client can produce five separate things, and each has its own risks:

Connect on LinkedInSagnik Bhattacharya

1. The audio or video recording. Some tools keep it, some discard it after transcribing. Among clinical scribes, for example, Heidi says it doesn't keep audio, while Nabla discards audio and keeps transcripts for 14 days by default.
2. The transcript. Every word, including the aside about another customer's bill or the door code a client read out.
3. The AI summary and action list. Short, confident, and sometimes wrong.
4. Copies sent to people. Auto-share settings can email the notes to everyone on the calendar invite, inside and outside your business.
5. Vendor use. Otter's privacy page says it de-identifies user data before training its models. Other vendors differ, and defaults change.

Most of the trouble small firms run into comes from items 4 and 5 rather than the AI itself. The transcription is usually good; it's where the output goes that causes the problems.

Subscribe on YouTube@codingliquids

Built-in recaps versus a bot that joins the call

There are two broad kinds of note-taker. Built-in ones are part of your meeting platform: Microsoft Teams intelligent recap, Google Meet's "Take notes for me" and Zoom's meeting summary. Third-party bots, such as Otter's, join the call as an extra participant. The difference matters for client calls.

Built-in tools have one structural advantage: the recording and notes stay inside the platform and account your business already manages, under the retention and admin controls you already have. A third-party bot adds another company that holds your client conversations. That isn't automatically worse, and specialist tools are often better at summaries, but it's one more supplier to vet. If you're choosing between tools on features and price, AI meeting note-takers compared for small teams sets them side by side, and the platform-specific settings are covered in Copilot in Teams meetings: recaps and privacy settings and Zoom's summaries and privacy settings for small teams.

A commercial cleaning firm's tender call that leaked

This scenario is illustrative, but it's built from the most common failure. A commercial cleaning company with three account managers runs about 22 client calls a week with facilities managers: contract reviews, tender clarifications and complaint calls. Each account manager has an Otter Business seat at $19.99 a month billed annually, about $60 a month for the three.

The set-up looked sensible, but two defaults did the damage. Auto-join was set to every calendar event with a video link, and auto-share sent the notes to all guests on the invite. After a tender clarification call, the client's procurement lead dropped off and the two account managers stayed on for five minutes to discuss their margin and how low they could go. The bot was still recording. The notes, margin discussion included, went to everyone on the invite, which included the client.

They changed four things the same afternoon:

• Auto-join moved from every meeting with a link to meetings selected by hand. Internal calls and anything sensitive now start without a bot.
• Auto-share now goes to the account manager only, who sends a checked summary to the client if one is wanted.
• Internal debriefs happen in a separate, unrecorded meeting, never in the tail of a client call.
• Retention is set by a monthly clean-up: transcripts older than 90 days are deleted unless linked to an open complaint. At 22 calls a week, that caps the archive at roughly 280 transcripts instead of an ever-growing pile of more than 1,100 a year.

Note what didn't change. The tool stayed, the price stayed, and the summaries are as useful as before. The risk came entirely from settings nobody had chosen on purpose.

What to say at the start of a client call

Consent works best when it's routine, short and gives the client a real choice. A script that account managers, estimators and office staff can all use (illustrative):

"Before we start, I use an AI note-taker so I can focus on
the conversation rather than writing. It records and
transcribes the call, the notes stay with us, and we delete
them after 90 days. Are you happy for me to use it today?
If not, that's absolutely fine, I'll switch it off."

Add a line to the calendar invite too: "We may use an AI note-taker on this call; tell us if you'd prefer we didn't." That gives the client notice before they join, which matters in places that require everyone's consent.

When a client says no, switch it off visibly and take handwritten notes. Don't argue the case for the tool. An illustrative HVAC installer found that about one client in ten declined, mostly facilities managers bound by their own employer's rules, and that nobody minded once the choice was offered. The ones who say yes also trust you more, because you asked.

Your privacy notice should say the same thing in writing, so a client who asks later can see what you do. A paragraph like this is usually enough (illustrative; adapt the details):

Call recording and AI notes
With your agreement, we may record and transcribe calls using
[tool] to produce notes of what was agreed. [Vendor] stores
these on our behalf. We delete recordings and transcripts after
[90] days unless they relate to an open complaint or claim.
We don't use them for marketing, and you can ask us not to
record at any time.

Keep the promise realistic. If you write 90 days, someone has to delete old transcripts every month, or a retention setting has to do it for you. A notice that says one thing while the tool does another is worse than no promise at all.

Settings that decide whether it's safe

Whichever tool you use, these are the settings to check, in order of how often they cause trouble:

• Auto-join scope. Otter's options range from every calendar event with a video link down to meetings you select manually. For client work, manual or host-only is safer.
• Auto-share scope. Google Meet's options run from all invited guests, including people outside your organisation, down to hosts and co-hosts only. Google also warns that "invited guests" means everyone on the calendar invite, not just those who attended, so a supplier who was invited but didn't join still gets the notes.
• Retention. Teams has a recording and transcript expiration policy; other tools rely on you deleting. Otter moves deleted conversations to a trash folder that empties after 30 days.
• Consent enforcement. Teams admins can require explicit recording consent, and Google Workspace admins can require participants to consent before note-taking, recording or transcription starts; Google leaves that setting off by default.
• Vendor training. Check whether the vendor uses your recordings or transcripts, even de-identified, to improve its models, and whether your plan lets you opt out.
• Integrations. Notes that flow automatically into a CRM or a task tool spread client conversations further. Turning meeting notes into tasks automatically shows how to pass on the actions without the whole transcript.

Picking the least risky option for your set-up

For most small firms, the safest note-taker is the one built into the meeting platform clients already join, because it adds no new supplier and follows the admin controls you already manage. A simple way to decide:

1. Google Workspace Business Standard or above? Start with "Take notes for me", set sharing to hosts only, and turn on the admin consent requirement if your client work is sensitive.
2. Microsoft 365 with Teams Premium or a Copilot licence? Use intelligent recap, and set a transcript expiry policy.
3. Clients insist on their own platforms, so you're on Teams one day and Zoom the next? That's where a third-party bot earns its place, on a business plan with auto-join and auto-share locked down.

A quick sum shows the money is rarely the deciding factor. For three account managers, Otter Business costs about $60 a month billed annually. Microsoft 365 Copilot Business, which includes intelligent recap among much else, is $21 per user a month on annual billing, so $63 for the same three people. Google's note-taking comes with Business Standard, which the firm may already pay for at about $14 per user a month. The difference between options is a few dollars; the difference between a well-configured tool and a badly configured one is the leak in the example above.

Whichever you choose, the same questions apply to any new vendor: where recordings are stored, how long they're kept, who at the vendor can access them, and whether they're used for training. If the vendor can't answer those in writing, it isn't ready for client calls.

Calls that shouldn't be recorded at all

Some conversations are better with a notepad, however good the tool:

• Security details. A locksmith taking a call from a customer after a break-in will hear about damaged locks, alarm codes, when the house is empty and where spare keys are kept. None of that should sit in a transcript on a third party's servers.
• Payment card details. Never let a client read a card number on a recorded call. Send a payment link instead.
• Disputes and legal matters. A call with a client's solicitor or insurer about a claim is better minuted by a person, because the words may be scrutinised later.
• Staff matters. Disciplinary, grievance and health conversations need a different level of care from a client catch-up.
• Health information. Unless you use a tool built for clinical settings and your professional rules allow it.

Phone calls need the same care

Plenty of trades businesses do most of their client talking by phone, not video. Many business phone systems and apps now offer call recording with AI summaries, and the same rules carry over. An illustrative removals firm taking about 120 inbound calls a week switched on AI call summaries to capture move dates, inventory notes and access details. The sensible version of that set-up has three parts:

• An automatic announcement at the start of every recorded call, such as "Calls are recorded and summarised to help us get your move right. Tell us if you'd rather we didn't." Callers who object are moved to an unrecorded line or called back.
• Outbound calls from staff mobiles aren't recorded unless the member of staff announces it, because the automatic message only plays on the main line.
• Card payments are never taken on a recorded line; the caller gets a payment link by text instead.

The firm also found one thing the summaries did better than people: they captured access details such as parking restrictions and lift sizes that staff used to forget to write down. That's the benefit worth protecting, and it only lasts if customers trust how the recordings are handled.

Checking the summary before anyone relies on it

AI summaries are fluent, and fluent errors are the dangerous kind. The summary below comes from an illustrative roofing survey call, with the problems marked:

AI SUMMARY: Roof survey call, [client first name]

- Customer agreed to proceed with full re-roof at $14,500.
  [WRONG: customer said they'd "think about it over the
  weekend"; no agreement was given.]
- Work to start Monday 12th.
  [WRONG: Monday was the earliest possible date mentioned.]
- Scaffolding included in price.
  [CORRECT]
- Customer mentioned neighbour's roof also needs work.
  [DELETE before sharing: third-party information.]

Two of four points were wrong in the direction that would cause a dispute: the summary turned a possibility into a commitment. Before any summary goes to a client or into a job record, have the person who ran the call read it against their memory, correct commitments, prices and dates, and strip out anything about people who weren't on the call. It takes about two minutes per call and avoids the "but your own notes say we agreed" argument.

Vendors change defaults, so recheck twice a year

Settings you chose carefully can shift under you. A real example from this year: SimplePractice's Note Taker, a clinical tool priced at $35 per clinician, has opted new users in by default to keeping de-identified transcripts since 16 June 2026. Existing choices weren't the issue; new accounts simply started with a different default. Legal pressure is shaping the market too. A class action filed in August 2025, Brewer v. Otter.ai, alleges that Otter's notetaker recorded meeting participants without their consent and used their voices to train its models. Otter's response was that nobody should be recorded without their knowledge or permission, and that its terms put the responsibility for notice and consent on account holders.

That last point is the one to take away. Whatever the vendor's defaults, the business that invites the bot into a client call carries the consent duty. Put a six-monthly check in the calendar: reread the vendor's privacy page, confirm auto-join, auto-share and retention are still set as you left them, and make sure new staff use the consent script. It's twenty minutes of work that keeps a genuinely useful tool from becoming the thing a client remembers you for.

Note-takers on client calls: quick answers

Do I need everyone's consent to record a client call?

It depends on where the people on the call are. Some legal systems let one party record, others require everyone's agreement, and sector rules can add more. The practical answer is to announce the note-taker at the start of every client call, offer to switch it off, and do so if anyone objects. Ask a legal adviser if you work in a regulated profession.

Is a free AI note-taker plan safe for client calls?

Usually not the best choice. Free and personal plans tend to have fewer admin controls, and some vendors use de-identified recordings to improve their models by default. For client calls, prefer the note-taking built into your paid business meeting platform, or a business plan where you control auto-join, sharing and retention, and check the vendor's training policy.

What should I do if a client's AI bot joins my call?

Treat it like an unexpected person in the room. Ask whose it is and whether the notes will be shared with you. If the call will cover pricing, security details or anything confidential, you're entitled to ask for it to be removed, and most platforms let the host remove participants. Note in your own records that the call was recorded by the client.

Further reads

• Can Lawyers Use AI Note Takers in Client Meetings? — The stricter view from a profession with confidentiality duties.
• AI Note-Taking Tools for Therapists: A Confidentiality Checklist — A confidentiality checklist from the most sensitive kind of call.
• How Much Does an AI Note-Taker Cost per Seat? — What each option costs per seat once you've picked the safe set-up.
• How to Review AI Call Transcripts for Quality and Compliance — How to check transcripts and summaries before relying on them.
• Where Is Your Data Stored When You Use AI Tools? — Where recordings and transcripts are stored once the call ends.
• How to Update Your CRM Automatically After Every Sales Call — Sending call notes into your CRM without oversharing.
• How to Stop AI Tools Training on Your Business Data — The exact training switches in ChatGPT, Claude, Gemini, Copilot and Perplexity, plus the hidden AI features most owners forget to check.
• Shadow AI: Is Your Team Using AI Without Telling You? — How to find the AI tools your staff use without telling you: an amnesty survey, five afternoon checks, and a keep-move-stop rule for each thing you find.
• Should You Tell Customers You Use AI? A Disclosure Guide — When to tell customers you use AI and when you needn't, with a decision table for 11 everyday uses and wording you can copy for chats, quotes and emails.
• Best AI Tools for Small Event Planning Businesses — Eight AI tools for small event planning businesses, ranked by the hours they give back, with prices, first uses and the two to handle with care.
• A Wedding Planner's AI Workflow From Enquiry to Final Timeline — One illustrative planner's year with AI, stage by stage: consultation notes, proposal, suppliers, guests, confirmations and the wedding-day timeline.
• AI Mistakes That Put Couples Off Your Wedding Business — Eight ways AI makes a wedding supplier look careless or fake to couples, each with an example from real-looking enquiries and the fix.
• Are AI Tools Built for Coaches Worth Paying For Over ChatGPT? — What coaching platforms, AI clones and note-takers add over a $20 ChatGPT plan, a year of each costed, and the platform-closure risk to plan for.
• AI Session Notes and Follow-Ups for Coaches: A Weekly Workflow — A week-shaped routine for coaches: capture sessions with consent, turn transcripts into your own notes, send same-day follow-ups and review clients on Friday.
• AI Note Takers for Financial Advisers: Compliant Meeting Records — How advice firms set up AI note takers so every file note is checked, consented to, retained properly and able to survive a compliance review.
• AI Acceptable Use Policy for a Small Professional Firm — A clause-by-clause checklist and one-page sample policy for accountants, lawyers and consultants who handle confidential client work.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Otter.ai privacy and security page and help pages on Notetaker auto-join and auto-share; Google Meet Help (Take notes for me); Microsoft Learn (recording and transcription overview for Teams); reporting on Brewer v. Otter.ai (Computerworld, August 2025); vendor facts on SimplePractice, Heidi, Nabla, Zoom and pricing from vendor pages checked September 2026.