The client meeting ends at 3:00. The partner walks out at 3:02.
The meeting after the meeting starts at 3:03 and runs until roughly 3:40, and it is attended by one person: him, alone, at his desk, trying to reconstruct what just happened.
He writes the recap email. He digs up the two things he promised to send. He makes a note in the file, or means to. He forwards something to the associate with the subject line "see below" and no other context, which will generate three follow-up questions tomorrow. He tries to remember whether the client said the closing was the 14th or the 18th. He books the follow-up meeting, or forgets to, and it slips two weeks.
Thirty-seven minutes. At a partner's loaded rate — say $180 an hour for a firm billing $400 — that's about $110. Do it four times a week and it's $23,000 a year, per partner, spent on remembering.
That's not the expensive part.
The expensive part is what didn't happen
The recap email he sends is fine. It is not, however, the recap email he would have written if he'd had a transcript in front of him, because he's writing from memory, forty minutes after the fact, having already had another conversation in between.
So he misses the aside — the one where the client mentioned, almost in passing, that they're thinking about a second location next year. That's the whole ballgame. That's the expansion, the new entity, the financing work, the lease review, the eighteen months of engagement that would have followed. It went by in nine seconds between two other topics, and it did not make it into the recap, and it will not come up again for a year.
Every professional services firm loses revenue this way, continuously, and none of them can measure it. It is the least visible and most expensive leak in the business.
What the machine is actually good for here
Not "AI notetaker." That's the commodity version and it's not what's interesting.
What's interesting is the gap between what was said and what got acted on.
A system that hears the meeting can do several things a tired human at 3:03 cannot. It can produce an accurate record of what was actually said, including the parts nobody flagged as important. It can extract commitments — every instance of someone in the room saying they'll do something, by when, with who owns it — and turn them into tasks that appear in the right place without anyone typing. It can draft the recap in the partner's own register, with the specifics correct, ready for a ninety-second edit rather than a thirty-minute compose. It can update the client record with the facts that changed. And it can surface the thing the partner would have missed: the client mentioned a second location. Nobody responded to this. Do you want to follow up?
That last capability is worth more than everything else on the list combined, and it's the one nobody demos.
The trust question, up front
You cannot put a machine in a confidential client meeting without answering four questions, and if a vendor won't answer them plainly, don't buy from them.
Where does the recording go, and for how long. Who else can see it. Does it train anything. And what does the client have to consent to.
For a firm with any kind of privilege or confidentiality obligation, these aren't compliance checkboxes — they're the whole design constraint, and they should shape the architecture from day one, not get bolted on after. Sometimes the right answer is that the audio never leaves your environment. Sometimes the right answer is that certain meeting types are simply off-limits, and the system knows which ones by calendar tag. There is no version of this that works if your clients aren't comfortable, and pretending otherwise is how firms end up with a very awkward conversation.
Get this right and it stops being an obstacle. We've seen clients react to "we record and structure our meetings so nothing you tell us gets dropped" as a mark of seriousness, not surveillance. But you have to have earned the right to say it.
What the partner does with the forty minutes
This is where the whole argument lands.
The reclaimed time isn't the return. Forty minutes a day of a partner's time is nice, but it's not what makes this worth doing. What makes it worth doing is that the partner walks out of the meeting and, instead of sitting down to reconstruct it, walks down the hall and has the conversation he's been meaning to have with the associate about the case. Or picks up the phone and calls the client he hasn't talked to in five months. Or goes home at 5:30, which, for a fifty-person firm losing good people to burnout, is not a soft benefit — it's a retention strategy with a hard dollar value.
The relationship work is the work. It's the only part of professional services that can't be commoditized, and it's the part that gets crowded out first, every single time, by the administrative residue of the last conversation.
Take the residue away. Let the people who are good at people be with people.
The meeting after the meeting should be a machine's job. The meeting before the next one — the one where you actually think about the client — should be a human's, and right now almost nobody has time for it.