AI meeting notes for consultants

Consultants use an AI meeting assistant to capture client discovery calls, workshops and staff interviews precisely enough to build deliverables from, within the rules the client sets. Check two things first: where audio and transcripts go, and whether you can use the AI provider the client has approved or keep everything on your laptop. Natively stores meetings on your computer and lets you choose the speech and AI providers, or run both on-device.

By Evin JohnPublished 31 October 20266 min read

What makes client meetings hard to capture

Confidential material. Discovery calls cover financials, org changes and problems the client has not told its own staff about. Where that audio goes is a contractual question, not a preference.
Client rules on AI. Some clients have approved one AI provider, some ban cloud AI entirely, and some have not decided. Your tool has to fit whichever applies.
Staff interviews. Employees speak more freely when they trust that their words will not be attributed to them in a report.
Workshops. Ten people in a room, one laptop, and decisions made in the last twenty minutes when everyone is tired.
Write-up time. Turning a week of interviews into a findings section is often slower than holding the interviews.

Matching the setup to the client's rules

Natively has no bot and does not connect to Zoom, Teams or Meet; it captures your computer's system audio and microphone. What leaves your laptop depends on the providers you choose. Audio goes to the speech provider you pick, or nowhere with an on-device model. Prompts with transcript context go to the AI provider you pick, or nowhere with Ollama. Transcripts, notes, screenshots and the search index are stored locally, and API keys are encrypted by the macOS keychain or Windows DPAPI.

The client approved a specific AI provider. Enter your key for that provider (OpenAI, Claude, Gemini and others are supported) and choose a speech provider the client accepts. Requests go from your computer to those providers.
The client allows no cloud AI. Use an on-device speech model such as Whisper or Parakeet, a model running in Ollama, and local embeddings for search. Transcription, answers, notes and search then work without internet. Download the models before you arrive on site, and test the quality on a practice meeting: a model small enough for a laptop usually writes weaker notes than a large cloud model.
The client does not allow recording at all. Do not run it. No tool setting makes that acceptable.

A client discovery interview, start to deliverable

BeforeConfirm the client's AI policy in writing, ideally in the statement of work. Choose the matching setup and the Requirements Gathering template for scoping, or Advisory for an executive session.
Opening the meetingExplain that you transcribe with an AI tool, where the notes are stored and who will read them. For staff interviews, say how quotes will be anonymized and that you will close Natively if they prefer.
DuringListen and type short markers. When someone shares a process map or a spreadsheet on screen, take a screenshot and ask about it. In a workshop, type the name of whoever makes each decision, because voices sharing one microphone may not be told apart.
AfterReview the notes against the transcript. Replace interviewee names with codes before anything goes into a working document, because Natively writes names as they were spoken.
Into the deliverableExport notes as Markdown, HTML or PDF into the engagement folder. Avoid share links for confidential material unless the client agrees: they are stored on Natively's server.

Fictional example

Six staff interviews for an operations review

Setup: A consultant reviews order handling at a fictional regional distributor. The client bans cloud AI, so the laptop runs Whisper and an Ollama model, with interviewees coded S1 to S6.
Consultant asks: "Which interviewees described approvals as a cause of delay, and at what step?"
Natively: S2, S4 and S5. S2 and S5 describe credit approvals for new customers waiting on one manager; S4 mentions approvals for orders above a set value.
Consultant asks: "Did anyone say approvals are not the problem?"
Natively: S6 says delays come from stock data being a day old, not from approvals.
Outcome: The findings section separates two causes, quotes each interviewee by code from the transcript, and the stock-data issue becomes its own recommendation.

Consulting meetings mapped to Natively

What to use for each kind of client meeting, and the plan it needs
Meeting	What to use	Plan
Client discovery or scoping	Requirements Gathering template	Pro
Executive or steering advisory session	Advisory template	Pro
Recurring project check-in	Project Sync template	Pro
Staff interview	General mode notes, interviewees coded	Free
Workshop in a room	Microphone capture and live transcription	Free
Findings across interviews	Ask across past meetings	Free
Client's approved AI provider	Your own API key for that provider	Free
No cloud AI allowed	On-device speech, Ollama and local embeddings	Free

What to use for each kind of client meeting, and the plan it needs. Scroll sideways on a small screen.

Templates are part of Pro; the one-time 30-minute free trial covers a single client call. The privacy and provider choices above are all on the free plan. The private meetings with local AI page covers the local setup in more depth.

Setting up for an engagement

Install on the laptop you bring to the clientmacOS 12 or later, or Windows 10 or 11 (x64).
Configure providers per the client's policyEither the approved cloud providers with your keys, or on-device speech, Ollama and local embeddings.
Decide on statisticsTurn off usage statistics in Settings if the client wants outbound traffic kept to a minimum. App analytics events, license checks and update checks still go out when online, so check that this is acceptable.
Run a test meetingCheck transcript accuracy on the client's accents and terminology, and the speed of a local model on your hardware.

What Natively does not do for consultants

It does not enforce the client's policy for you. Which providers you configure decides where data goes.
It does not anonymize transcripts or notes; removing names is your job.
It does not record video, and it does not join meetings you are not attending.
Its fully local mode is not zero network: see the note above on statistics, analytics events, license and update checks.

Questions people ask

Can I use Natively when my client does not allow cloud AI?

Often, yes. With an on-device speech model, a model running in Ollama and local embeddings, transcription, answers, notes and search work on your computer without internet. When online, Natively still sends usage statistics (which you can turn off), app analytics events, license checks and update checks, so describe it to the client in those terms.

Can I use the AI provider my client has approved?

Yes, if Natively supports it. You enter your own key for providers such as OpenAI, Claude or Gemini, and prompts with transcript context go from your computer to that provider. Audio goes only to the speech provider you choose, or nowhere with an on-device model.

Where are client meeting notes stored?

Transcripts, notes, screenshots and the search index are stored on your computer, and API keys are encrypted by the macOS keychain or Windows DPAPI. Share links are the exception: they are stored on Natively's server until you turn them off.

Where to go next

Natively captures the call audio on your own computer, so no bot joins the meeting. Notes, recaps and answers come from the transcript.

Facts on this page were last checked on 31 October 2026. Product details come from the Natively app itself; if something here is out of date, tell us at hello@heynatively.com.