Yes, you should tell clients when an AI notetaker is present. Say it before the meeting begins, explain what the tool does, and ask for consent in a way that gives the other person a genuine choice. That practice protects trust even when you are still checking the legal requirements for your location and the client's location.
Do not hide the tool behind vague language such as "we use technology to improve service." A client deserves to know whether the meeting is recorded, transcribed, summarized, or analyzed. The article on writing a client case study with AI makes the same broader point: useful output starts with clear permission and reliable source material.
The sentence to use
Say: "I use an AI notetaker to help me capture decisions and follow-up items. It may record or transcribe this meeting. Are you comfortable with that, or would you prefer that I take notes myself?"
That sentence names the purpose, acknowledges recording or transcription, explains the benefit, and gives the client an alternative. Adjust it to match what your tool actually does. Never say it only takes notes if it also creates a transcript or stores audio.
Pause after asking. Silence is part of consent. If the person asks where notes go, how long they are kept, who can access them, or whether the tool is used to train a model, answer from your actual policy. If you do not know, stop and find out before continuing with the tool.
Why the answer changes
Recording and consent rules can vary by jurisdiction and by the people involved. A meeting with participants in different states or countries can create extra uncertainty because more than one set of rules may matter. The Reporters Committee's recording guide by state is a useful starting point, but it is not advice about your business, meeting, or client agreement. That is the same question behind Is a Chamber of Commerce membership actually worth the money, which walks through it in detail.
There is no universal sentence that removes every legal question. Disclose early, ask clearly, document the preference, and avoid recording when a participant declines. If the conversation involves sensitive financial, health, employment, legal, or confidential business information, use a higher standard of care and ask qualified counsel how your process should work.
Consider where each person is sitting, where your business is located, and what your contract says. A client may have an organizational policy stricter than the minimum law. Respect the most restrictive practical requirement you know applies instead of looking for a technical loophole.
Consent is more than yes
A useful consent process answers five questions:
- What is the tool?
- What will it capture?
- Why are you capturing it?
- Who can access the output?
- What happens if the person says no?
Put the answers in your meeting invitation when appropriate, then repeat the short version aloud. Written notice is helpful, but do not assume a buried calendar sentence will be understood as permission. People may join from a different device or forward an invitation to someone who never saw it.
When a new participant joins late, pause and disclose again. If the meeting changes from planning to a sensitive discussion, ask whether everyone is still comfortable. Consent should follow the conversation.
Build a no-notetaker path
Your business should still function when a client declines. Prepare a manual note template with the purpose, decisions, owners, deadlines, and open questions. At the end, read back the key points and send a short recap for confirmation.
This alternative helps you avoid pressure. It also shows whether the AI tool supports a process or covers for the absence of one. If your meeting has no clear agenda and no reliable follow-up without a transcript, fix that foundation first.
You can use AI after the client has approved the material and your policy allows it. A human note can become a decision log or task list. Keep the source narrow, remove unnecessary personal data, and review the result before it reaches the client.
Protect information after the call
Access and retention deserve the same attention as disclosure. Decide who needs the transcript, where it is stored, how long it remains useful, and how you handle deletion requests. Do not let recordings sit in a shared folder because nobody chose a retention rule.
Ask the tool provider what settings you can control, but do not repeat a vendor privacy promise as if it were your policy. Your client agreement, internal access controls, and actual behavior need to match. If you change tools, revisit the disclosure because capture and storage may differ.
When a meeting produces material for a public story, get separate permission for that use. Consent to take notes is not automatically consent to publish a quote or result. The case-study process described in what review replies tell future customers gives you a better approval checkpoint.
A practical decision rule
If you cannot explain the notetaker in one clear sentence, do not turn it on yet. If a client asks a question you cannot answer, pause. If someone declines, switch to your manual process. If the discussion becomes sensitive, ask again. These rules are easy to follow because they depend on behavior, not assumptions.
AI notes can save time, but the relationship is worth more than a perfect transcript. Tell clients, ask directly, keep records purposeful, and get legal guidance when the situation is unclear. That is how you gain efficiency without asking trust to carry a secret.
Write the disclosure into your meeting checklist so it does not depend on memory. Put a visible reminder beside the meeting link, make the manual note option easy to choose, and record the client preference in the place your team already uses for meeting details. Do not create a new private spreadsheet that nobody will maintain.
When the meeting ends, tell the client what will happen next. You might say that you will review the notes, send decisions for confirmation, and delete material that is no longer needed under your policy. A clear closing gives the client another chance to correct a misunderstanding before it becomes part of a project record.