Four calls, four clients, one Mac that never sends any of it anywhere.

Tuesday's calendar has a 9:00 kickoff with a new retainer client, an 11:00 status call with a client three months into an engagement, a 1:00 call with a stakeholder who's difficult to get on the phone at all, and a 4:00 internal sync with the practice lead about all three. None of that gets rescheduled for the sake of writing anything up between calls — the write-up has to fit around the calendar, not the other way round.

8:52am — before the first name shows up on the call

The Mac is already recording before the client dials in: press record, join the call, and the two audio sources the Mac can hear — the microphone in the room and whatever the meeting app plays out loud — start as two separate tracks. Nothing about the call's participant list changes because of it; no name gets added and no permission prompt appears on the client's side.

9:00am — the kickoff, where most of what matters is said once

A new client spends the first call setting scope, and scope decisions have a way of getting revisited later as "I thought we agreed X" — usually from memory, usually slightly wrong on one side or the other. This call ends with a transcript sitting on the Mac and a short summary generated from it, both before the 11:00 call has even started.

10:40am — checking what was promised three weeks ago

Before the 11:00 status call, the actual prep isn't a slide — it's confirming what was promised on the last call, three weeks back, without trusting memory for the exact wording. A keyword search across past transcripts turns up the sentence directly: what was said, and when. That's a five-minute check against a folder of past recordings, not a rebuild of a call that already happened.

11:00am — the status call, and the commitment that needed checking

The client opens by asking whether the timeline discussed last time still holds. Having the actual sentence from three weeks ago, rather than a paraphrase, is the difference between "yes, and here's exactly what we said" and a vaguer answer that invites a longer negotiation than the question needed.

12:20pm — sending the recap before lunch, not after

The status call ends at noon. The recap the client actually reads goes out before lunch rather than at 6pm, because the summary is sitting on the Mac the moment the call ends rather than waiting on an evening's worth of other calls to get written up from memory. A short recap sent within the hour reads as attentive; the same recap sent the next morning reads as an afterthought, even though the content is identical.

1:00pm — the call with a stakeholder who's never on time

The third call is the hardest to schedule: a stakeholder who joins eight minutes late, from a different meeting that ran over, and who brings up the one open question the whole engagement has been waiting on. Recording starts at 1:00 regardless of who's actually on the call yet, so the eight minutes of waiting and the substantial part that follows are both on the same track — nothing about starting the recording depends on everyone already being present.

Between calls — the write-up that doesn't wait for evening

By 1:40 there are three transcripts and two summaries sitting on the Mac, and the useful moment to act on the stakeholder's pricing answer is now, not after the 4:00 internal sync. The proposal update goes out from the actual sentence the stakeholder said, not a recollection of it typed up between other things later in the day.

4:00pm — the internal sync, recorded the same way and read differently

The 4:00 call with the practice lead isn't a client call, and it gets exactly the same treatment — recorded, transcribed, summarised — because a consultant switching between "client mode" and "internal mode" note-taking habits is how details get lost at the boundary. The difference isn't in how the call is captured; it's in who reads the recap afterward and what they do with it.

End of day — one folder, four calls, nothing sent anywhere to produce any of it

By 5:00, four recordings sit in the same folder on the Mac: a kickoff, a status call, a stakeholder call, and an internal sync. Each has its own transcript and its own summary, produced on that Mac without any of the audio, the text or the summary being uploaded to generate them. Finding what was said in any of them next month is the same keyword search that worked for the three-week-old commitment this morning.

What today would have looked like with a shared calendar and no habit of pressing record

Not every day like this one runs cleanly. The habit that makes it work — starting the recording before the call actually connects — takes a few weeks to become automatic after years of relying on a bot that used to join calls on its own. A consultant switching off a bot-based tool for the first time typically forgets to press record at least once in the first fortnight, usually on the call that mattered least that week rather than the kickoff or the stakeholder call. That's a real cost of the switch, not a reason to pretend the switch is free — it's simply a smaller cost than a client noticing an unfamiliar name join a call they were told would be just the two of them.

Why four calls stay four files instead of one long recording

It would be simpler, mechanically, to leave the Mac recording continuously from 9:00 to 5:00 and sort out which minutes belonged to which client afterward. Nothing about the workflow works that way: each call is its own recording, started and stopped with the call itself, which means the Meridian kickoff is never accidentally searchable from inside the stakeholder call's transcript, and a client who asks for their own recording gets exactly their own call, not a slice cut out of a longer file that also contains someone else's conversation.

The proposal that went out the same afternoon

The stakeholder's answer on the annual discount at 1:00 could easily have sat until the next morning if writing it up depended on finding time between the 4:00 sync and the end of the day. Instead the updated proposal numbers went out at 2:15, while the stakeholder's exact phrasing on what they'd agree to was still sitting in a transcript rather than becoming "I think they said something about the annual number" by the time there was room to write it. A proposal that reflects what was actually said, sent the same day, is a small thing that compounds — the client remembers which consultants follow up fast and which ones need reminding.

What the practice lead actually reads from today

The 4:00 internal sync produces a summary meant for one reader, not three clients, and it reads differently because of that: three accounts in one short recap, cross-referenced against what each client actually committed to on their own call that day. The practice lead doesn't sit in on any of the three client calls, so the honest state of each account — not the version a consultant might unconsciously round up when reporting from memory at the end of a long day — depends entirely on whether the recap comes from an actual transcript or from whatever details survived four back-to-back calls.

What a bot-based tool would have added, and what it would have cost

Otter or Fireflies joining each of today's four calls as a named participant would have produced broadly the same transcripts and summaries, and for an internal-only calendar that's a reasonable trade. For calls with a new client and a stakeholder who's cautious about who's listening, an unfamiliar name on the roster is a small thing to explain on a call that isn't about the software. Granola avoids the bot specifically but is still a cloud subscription end to end — the call is still recorded, transcribed and summarised on a server neither the consultant nor the client chose together, at a running cost of $14 to $35 a month depending on the tier, against Otter's $8.33–$19.99-a-month tiers on annual billing. None of today's four calls needed a bot in the room or a server in the loop to produce the same recap.

The Friday review that makes the whole week checkable

Fridays, before the week closes out, the practice does a short review of every open commitment across all three accounts — what was promised, to whom, and by when. That review used to mean flipping through whatever notes survived each week's calls, some more complete than others depending on how busy that particular day had been. With a transcript for every call sitting in the same place, the Friday review becomes a search across the week's files rather than a reconstruction from partial notes, and a commitment that got mentioned once, in passing, on a Tuesday call is just as findable as one that made it into a formal recap.

The disagreement that got settled without an argument

Three weeks after today, the Meridian client disputes whether the scope was ever meant to include the go-to-market audit — a natural disagreement, since scope conversations happen once and get remembered slightly differently by each side afterward. The kickoff transcript settles it directly: the decision line from that first call's summary states plainly that scope covered the pricing review only. Neither side has to trust the other's memory of a conversation from three weeks back, because the actual sentence is still there, timestamped, exactly as it was said.

What today's workflow deliberately doesn't do

Nothing here files a call automatically against a client record, a project code or a billing system — there's no CRM integration, no practice management connection, and no shared workspace a colleague can browse without being sent the file directly. Each of today's four recordings lives on the Mac that made it, across up to three of a consultant's own machines on one licence, and the whole day's workflow runs on macOS only. A firm running structured engagements through a shared case system across a dozen consultants has a real reason to stay there — that's infrastructure this workflow doesn't attempt to replace.

The week this stops feeling like extra effort

The first week of this workflow feels like a habit still being built — remembering to press record, remembering to check the folder before a call rather than after. By the third or fourth week, checking a transcript before a call becomes as automatic as glancing at the calendar was before, and the moment it stops feeling like effort is usually the same week a client asks "didn't we discuss this already?" and the answer is a search, not a guess.

What it costs, what it runs on

Coii AudioNotes is $19, paid once, covering three of a consultant's own Macs, with a 30-day trial needing no card and no account. It runs on macOS 13 Ventura or later, Apple Silicon or Intel — the same setup used for all four of today's calls. The full case for a consultant's notetaker covers the argument in more depth than a single day can, and the comparison with Otter and comparison with Granola cover what each alternative actually does beyond today's four calls. For finding what was said weeks later, searching old meetings and writing minutes from a transcript both walk through the mechanics this page only showed in motion, and a round-up of private meeting recorders covers the wider field a confidentiality-first search turns up. The case for skipping the bot entirely, and how a user researcher's day runs against a different kind of back-to-back calendar, are both one click away.