← Back to Blog

Short answer: Dictate meeting minutes in three passes. Type the structured fields before the meeting, speak the decisions and motions in short complete sentences as they happen, then dictate the narrative summary immediately afterwards while the discussion is still fresh. Speak the sentences, type the numbers and names.

Minutes are the slowest writing job in most organizations, and the reason is not that the writing is hard. It is that the person taking them is trying to do two things at the same time: follow a live discussion closely enough to know what actually happened, and turn that discussion into formal prose at typing speed. Those two tasks compete for the same attention. Something has to give, and usually it is the record.

That is why so many minutes read like a list of headings with nothing underneath them. "Discussed budget. Discussed hiring. Motion passed." Six weeks later nobody can reconstruct what was decided or why, and the only person who knows is the one who was too busy typing to remember.

Speech removes the bottleneck. Most adults type somewhere around 40 words per minute and speak at 130 to 150, which means the cost of capturing a full sentence instead of a fragment drops by roughly two thirds. That change in cost is the whole point. It is not about producing minutes faster. It is about producing minutes that are actually usable, because the specific version finally costs less effort than the vague one.

Minutes are not meeting notes, and the difference matters

Before any technique, be clear about which document you are producing. Personal meeting notes are for you. They can be fragments, arrows, question marks, half-thoughts. If that is what you need, the method in our guide to taking meeting notes by voice on Mac is the right starting point and this post is more structure than you need.

Minutes are different. They are an institutional record. Somebody who was not in the room will read them, possibly years later, possibly in a dispute. In many boards, nonprofits, HOAs, works councils and public bodies they carry procedural weight: they get moved, seconded, corrected and approved at the following meeting, and once approved they are the official account of what happened.

That has three practical consequences for dictation:

Everything below is built around that split: speak the parts that are prose, type the parts that are data.

Set up before anyone joins

Minutes have a skeleton that never changes, and building it live is wasted effort. Ten minutes before the meeting, open your template and type the fixed parts: organization name, meeting type, date, time, location or platform, the chair, the minute taker, and the agenda items in order. Paste the agenda straight from the invite if you have one.

Then put your cursor under the first agenda item and leave it there. On a Mac with Voice Keyboard Pro you hold a hotkey, speak, and release, and the text appears at the cursor in whatever app you are already using. There is nothing to install into your minutes template, no plugin for your document editor, no admin approval to request. That matters more than it sounds for anyone taking minutes in a managed corporate environment or in a browser-based board portal, where getting anything installed is a multi-week request and getting nothing installed is a download.

One setup check worth doing once: if you take minutes inside a board portal or any web app that logs you out, do not write long passages directly into the live field. Dictate into a plain scratch document, then paste in blocks. Session timeouts eat unsaved text, and voice makes you produce more text per minute, which means a timeout costs more than it used to.

The three passes

Pass one, before: the skeleton

Structured fields typed, agenda in order, cursor parked. Two minutes of work that saves ten later.

Pass two, during: decisions in short complete sentences

This is the discipline that makes dictated minutes work. During the meeting you are not writing minutes. You are capturing decision-shaped sentences, one at a time, in the gaps.

The gaps are real and you will find them once you start looking. Someone pulls up a slide. Two people are agreeing at length. The chair asks whether anyone else wants to speak. Each of those is four to eight seconds, which is enough for one complete spoken sentence and nowhere near enough to type one.

Capture these, in this order of priority:

  1. Decisions. What was agreed, in a full sentence, with the subject named.
  2. Actions. Who is doing what, by when.
  3. Motions. The proposition, the mover, the seconder, the outcome.
  4. Deferrals. What was explicitly pushed to a later meeting, and to when.
  5. Positions on record. Any objection or abstention someone asked to be noted.

Notice what is not on that list: the discussion itself. Resist capturing the back and forth. It is the most tempting material and the least useful in minutes.

Pass three, immediately after: the narrative

The most valuable ten minutes of the whole process happen in the ten minutes after the meeting ends. Stay in your seat, put the cursor under each agenda item in turn, and talk the summary out. One or two sentences per item covering what was considered, what was decided, and anything that needs to be picked up next time.

You are not remembering wording here, you are remembering substance, and substance decays fast. A summary dictated ten minutes after the meeting is a different document from one typed the next morning, and everyone who has done both knows it.

How to speak a motion

Motions are the one place where paraphrase is genuinely a problem, and they are also the hardest thing to capture live because they are often reworded twice before the vote.

The reliable pattern is to wait. Do not try to capture the motion as it is being drafted out loud. Wait until the chair states the final form for the vote, which is exactly when the room goes quiet, then speak it in one pass using a fixed structure:

Moved by [name], seconded by [name], that [the proposition as stated]. Carried.

Speak it as one continuous sentence rather than word by word. Continuous phrases are recognized more accurately than isolated words, because the surrounding words provide the context that resolves similar-sounding options. If you are not getting clean results on longer utterances, our guide on how to speak clearly for dictation covers why over-articulating makes accuracy worse rather than better.

Vote counts are a different matter. Type those. "Four in favour, one against, one abstention" is four numbers in a sentence, and number errors are silent: a wrong digit looks exactly as confident as a right one, and nothing in the sentence flags it. If your body records the counts, type them and read them back before you move on. Our guide to dictating numbers and dates explains the cases where numbers do dictate reliably, and the ones where they do not.

Speak the sentences, type the values

The rule that keeps dictated minutes clean is a boundary, not a technique. Some content is prose and belongs to your voice. Some content is data and belongs to your keyboard.

Dictate: decision summaries, discussion summaries, rationale, action item descriptions, motion wording as stated, next steps, chair's remarks in substance, anything that reads as a sentence.

Type: vote tallies, dollar and budget figures, percentages, dates in numeric form, times, docket or resolution numbers, member ID numbers, addresses, email addresses, and the attendance list.

Attendance deserves its own note. It is a list of proper nouns with no sentence context around them, which is the single worst case for any transcription engine. Names read from a list have no neighbouring words to constrain the guess. Build your attendance roster in the template once, then check names off rather than dictating them fresh every meeting.

The names problem, and the fix

Anyone who takes minutes regularly hits the same wall: the same eight or ten names appear in every set of minutes, and they are the words most likely to come out wrong. Board members, committee names, your organization's internal acronyms, the project code names, the external firms you work with. They are a small fraction of your total words and a large fraction of your corrections.

This is not an accuracy problem you can fix by speaking better, because it is not a mishearing. Your colleague's surname genuinely is a rarer word than the common word it gets replaced with, and general-purpose transcription has no reason to prefer it. Trying harder does not help.

Voice Keyboard Pro's Smart Vocabulary is the direct answer to this. It is a personal dictionary with replacement rules: you add the names, the acronyms, the committee titles and the recurring terms your organization uses, and they stop coming out wrong. For a minute taker this is the difference between a five-minute cleanup pass and a thirty-second one, every single meeting, forever. Add names as they appear rather than trying to load the whole roster at once.

When you are also a participant

The genuinely hard version of this job is being the minute taker and a contributing member at the same time. You cannot dictate a decision summary while you are the one arguing for it.

For those meetings, capture is the problem, not writing. Meeting Mode on the Mac app runs during the meeting with speaker detection and produces AI notes afterwards, so the record of who said what does not depend on you being free to type. Calendar meeting detection means it can pick up that a meeting is starting rather than relying on you to remember. Our post on transcribing a meeting with speaker names and a summary on Mac walks through how the speaker labelling and summary side works in practice.

Two honest caveats. First, notes from a meeting are raw material, not minutes. You still do pass three: read them, decide what is decision and what is discussion, and dictate the formal record yourself. Nobody should publish machine output as an approved institutional record. Second, tell the room. Whether recording or automated notes are acceptable is a decision for your chair, your members and in some jurisdictions your local law, not for a software vendor. Some bodies require consent, some prohibit it outright, and some have a standing policy already written down. Ask before your first meeting, not after.

What does not belong in minutes

Dictation makes it cheap to produce text, and cheap text is how minutes get into trouble. Speech is looser than typing by nature, and the friction that used to stop you writing something questionable is gone.

Keep these out:

A simple safeguard: never send minutes in the same sitting you dictated them. Speech carries tone that text loses, and the sentence that sounded measured out loud can read sharply on the page. Let the draft sit, then read it once before it circulates. This is the same discipline that applies to any dictated document read by someone other than the author.

Doing it from a phone

Plenty of minutes get taken in rooms without a laptop: site visits, community meetings, anything held in a hall. The iPhone keyboard has a built-in mic button, so you can dictate into whatever notes or documents app you already use, and Voice Edit lets you speak a change to fix a line rather than trying to place a cursor precisely with a thumb. That second part matters more on a phone than people expect, because thumb-editing is where mobile capture usually falls apart.

For bodies that operate in more than one language, two-way translation while dictating covers 24 languages, which is useful when the discussion happens in one language and the record has to be circulated in another. Treat that as a first draft. Anything with procedural or contractual weight should be checked by a person who reads both languages.

A workflow you can repeat

  1. Ten minutes before: open the template, type the header fields, paste the agenda, park the cursor.
  2. During: dictate decisions, actions and motions in short complete sentences, in the gaps. Type vote counts and figures as they are stated.
  3. At the close: before anyone leaves, read the action items back to the room. Corrections cost nothing now and a great deal later.
  4. Immediately after: stay seated, dictate the narrative summary under each agenda item.
  5. Same day: read the whole thing once, out loud if you can, and fix what does not read as a record.
  6. Before the next meeting: add any name or term that came out wrong to Smart Vocabulary so it stops happening.

The first two meetings will feel awkward. Composing spoken prose is a distinct skill from composing typed prose, and it takes about a week of regular use before it stops feeling like performance. That curve is real and it is short.

What voice does not fix

Faster capture does not make you a better minute taker. If your minutes are vague because you are unsure what counts as a decision, dictation will produce vague minutes faster. Structure comes first; speed just makes good structure affordable.

It also does not resolve a chair who never states motions clearly, or an agenda nobody follows. Those are meeting problems. What voice fixes is the specific, narrow, extremely common problem of the record being thinner than the meeting was, because writing it down cost more than the person taking it could spare in the moment.

On privacy, since minutes routinely contain sensitive material: Voice Keyboard Pro's server stores only operational pings. No audio and no transcript content is kept there. Your minutes stay where you wrote them. Whether your body's material can be processed by any third-party tool at all is a question for your organization's counsel or IT policy, and that is genuinely their call rather than ours.

Frequently asked questions

Should minutes be a transcript?

Almost never. A transcript records everything said; minutes record what was decided. Bodies that require verbatim records usually say so explicitly in their governing rules, and those bodies typically use a dedicated transcription service rather than a minute taker.

Can I dictate straight into a board portal?

Yes, if the field accepts normal keyboard input, because text arrives at the cursor exactly as typing does. But dictate long passages into a scratch document first and paste them in, so a session timeout cannot take your work.

How do I handle a meeting where people talk over each other?

Do not try to capture it live. Note the agenda item and a one-line marker, then reconstruct in pass three. Crosstalk is hard for people and hard for software, and the honest answer is that the summary you write afterwards will be better than anything captured during it.

Is it acceptable to tell the room I am dictating?

Yes, and it is usually the right move. Dictating a decision summary out loud is also a free accuracy check: if you have it wrong, someone will correct you on the spot.

The bottom line

The reason minutes are thin is almost never that the minute taker did not care. It is that following a meeting and transcribing it simultaneously is two jobs, and typing forces you to do them at the same time. Speaking separates them. You capture decisions in the gaps at speech speed, then write the narrative while the meeting is still in your head.

Voice Keyboard Pro has a free tier with daily limits, and Pro is $4.99 a month or $34.99 a year. Try it on one meeting: type the skeleton, capture five decision sentences by voice, and dictate the summary before you stand up. Compare that document to your last set of minutes. The difference is not that it took less time. It is that it says what happened.

Keep reading