Four client calls today, and a write-up owed after each one.

A consulting day rarely has room in it for the part that actually takes discipline: turning an hour of conversation into a page a client will read, before the next call starts and the last one's details start to blur into whichever came before it. Voice Memos will record the hour. It will not tell you which of the four calls a particular commitment came from, and it will not write the page.

What a bot-based notetaker adds to a client relationship

Otter, Fireflies and the other tools that put a named participant on the call solve the write-up problem, and for an internal team that never minds a bot on the roster, that is a reasonable trade. A client call is a different situation: a new client, mid-engagement, watching an unfamiliar name join the meeting and sit there for the whole hour is a small thing to explain on a call that is supposed to be about their problem, not the software recording it. It is also, for either tool, a call that is recorded, transcribed and summarised on that vendor's own servers before the consultant sees any of it — including Granola, which does not use a bot but is still a cloud subscription end to end.

Reading the Mac's own audio instead

Coii AudioNotes records the microphone and the system audio the Mac is already carrying during the call, as two separate tracks, and does the rest — transcription and the summary — on that same machine. Nothing joins the meeting, and nothing about the call is sent anywhere to produce the notes.

From an hour of conversation to a page worth sending

A language model bundled inside the app reads the transcript and turns it into a summary — decisions, action items, open questions — while the details are still accurate rather than half-remembered by the time a write-up gets written that evening. For someone billing four or five calls a day, that is the difference between a summary written from a fresh transcript and one reconstructed from memory an hour after the fact, after the details of call three have started running into call four.

Telling two voices apart without a feed from the platform

A bot inside a call separates speakers using the individual audio feed the meeting platform hands it for each participant. Reading the Mac's own microphone and system audio instead means that feed isn't available, so speaker separation works by voice print — matching each voice against the others in the same recording. The transcript above reads the same either way; a client and a consultant on a call are also, usefully, exactly two distinct voices for the matching to work against.

What it deliberately doesn't do

Worth stating plainly, since a consulting workflow often runs through several other tools already. There is no integration with a CRM, a practice management tool or a billing system — nothing here files a call automatically against a client record or a project code. There is no shared workspace a colleague on the same engagement can browse without being sent the file directly; each recording lives on the Mac that made it, across up to three of a consultant's own machines on one licence, not on a server other people can be granted access to. And it is built for macOS only, so a firm with analysts on Windows laptops will not get the same tool across the whole team from this one alone.

The confidentiality question underneath the write-up

Most consulting engagements start with an NDA, and what that document actually promises is that the client's words stay inside a small, specific set of hands. A cloud notetaker's own privacy page can describe where a recording goes and who can reach it, but that document exists somewhere the client never reads before the call starts, and the honest answer to "who else has this" is "a server operated by a company neither of us chose together." Recording, transcribing and summarising on the Mac that made the call sidesteps that question rather than needing a good answer for it — nothing about the call exists anywhere but the machine that was already in the room.

Who this genuinely isn't built for

A larger firm running structured engagements through a shared case management system, where a partner needs to browse an associate's call notes without being sent the file, gets real value from that infrastructure — value this tool does not attempt to replace. This is written for the version of consulting where the calls are mostly one person's, the write-up is mostly that same person's job, and the client's comfort with who's listening matters more than a shared dashboard. If the firm-wide workspace is what actually gets used every day, that's a reasonable reason to stay with a cloud tool built around it.

Setting it up is smaller than the decision to switch

The first launch asks for two macOS permissions — microphone access and permission to record system audio — the same prompt any Mac app requesting audio capture triggers, and that's the whole of it. No account to register, no workspace to configure, no calendar to connect before the first client call can be recorded. The larger part of moving away from a bot-based tool is usually not the software at all: it's remembering to press record instead of trusting a bot that used to join on its own, a habit that takes a couple of weeks of calls to fully replace.

Recording is a decision, not a default

There's a trade worth naming honestly in the other direction, too. A bot that auto-joins every scheduled call captures everything without anyone having to remember to start it, which is a genuine convenience for a consultant running a dozen calls a week and no consistent habit of hitting record. Moving to a tool that has to be started manually shifts that responsibility back onto the person on the call — a real cost worth weighing honestly against the confidentiality question, not something to pretend away because the rest of the argument favours switching.

The call where this actually matters most

Picture the call where a client is walking through something they'd rather not repeat in front of an unfamiliar name — a budget overrun, a vendor relationship gone wrong, a reorg they haven't announced yet. That's the call where a named bot on the roster, or the knowledge that the recording is sitting on someone else's server, actually changes what gets said out loud. Running that same call with nothing but the usual two names on the roster, and a recording that never leaves the Mac either of them is looking at, removes a variable that a sensitive conversation didn't need in the first place.

Why the transcript matters beyond the write-up itself

A summary is what gets sent to the client, but the transcript underneath it is what settles the argument three weeks later, when the client's recollection of what was agreed and the consultant's don't quite match. "I have it in the notes" is a weaker claim than "here's the exact sentence, at the timestamp it was said" — and the second one only exists if the call was actually recorded rather than typed from memory in the gaps between listening. That is a small thing on most engagements and occasionally the entire reason a disputed deliverable gets resolved without escalating.

The pattern across a full week of client work

A single call is where the write-up problem shows up first, but a full week of consulting makes a second problem visible too: which client said what, three engagements ago, when a similar question comes up again on a different account. A transcript sitting on the Mac, searchable later, is the difference between remembering roughly what was discussed and being able to check the exact wording — without any of it having passed through a shared search index other clients' calls also sit inside.

What actually changes across a full engagement

Nothing about how a client call runs changes — same agenda, same platform, same conversation. What changes is what a consultant has available three weeks into an engagement, when a client references something from the kickoff call and the accuracy of the response depends on whether that call is still just a memory or an actual transcript sitting on the Mac, searchable in seconds rather than reconstructed from whatever notes happened to survive.

Who this genuinely isn't built for, said a second way

A boutique or solo practice where the calls are the whole of the record — no shared CRM, no team of associates needing the same file — is exactly who this is built for. A larger firm that has already invested in a practice management platform with shared client histories has a real reason to stay there, not because this tool does anything worse on a single call, but because the value of that platform comes from the sharing this tool deliberately doesn't do.

The first week after switching

The first few client calls after moving off a bot-based tool tend to feel slightly unfamiliar for one specific reason: the habit of watching the participant list to confirm the bot joined before starting to talk. That habit fades within a week or two, and what replaces it is simpler — press record, run the call, and find the transcript and recap waiting on the Mac once it ends, the same outputs the bot used to produce, from a tool nobody on the call ever had to notice.

What it costs, what it runs on

Coii AudioNotes is $19, paid once, for 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. Against Otter's $8.33–$19.99-a-month tiers or Fireflies' comparable subscription, a solo consultant lands at $19 total rather than somewhere near $200 a year; against Granola's $14-a-month Business tier, the price gap is smaller but the underlying difference stays the same — Granola is still a cloud subscription end to end. Full comparisons here, here and here. The broader case for a Mac-native meeting notes app and what an AI notetaker without the bot actually changes both cover ground this page doesn't repeat, and a round-up of private meeting recorders covers the wider field a confidentiality-first search usually turns up. What skipping the bot itself changes on a client call, and what happens when the connection at a client's office isn't reliable, are both covered in offline transcription for Mac.