What You Give Away When You Send a Customer the Full Meeting Record

Sending them the notes covers two very different things

Somebody on the account team says they will send the customer the notes. Half the room hears a short summary of what was agreed. The other half hears the whole thing, the full text of the call. Both get called sharing the notes, both feel like the same generous gesture, and the question of which one is actually going out almost never gets asked out loud.

The two documents behave differently the moment they leave your side. A recap is short, written by you, about what happened. A record is long, written by the meeting itself, about everything that happened. The first is a claim your customer can agree with or correct in one line. The second is material, and material gets used.

This piece is about the second one. The recap is straightforward good practice and most teams send it less often than they should. The record is where teams tend to land in one of two bad places: oversharing because it feels like openness, or refusing to share anything at all because it feels vaguely risky, with no clear reason behind either choice.

A recap is a move in the conversation. A record is raw material.

Herbert Clark and Susan Brennan describe the target that two people in a conversation are actually aiming at. They call it the grounding criterion: the contributor and his or her partners mutually believe that the partners have understood what the contributor meant to a criterion sufficient for current purposes. Getting there takes two steps. One person presents something, and the other accepts it by giving evidence they understood it. The paper is blunt about the arithmetic: it takes both phases for a contribution to be complete.

A recap sent the same afternoon is the first step. It says, here is what I think we agreed. The customer reads it and either says yes or writes back one line fixing the part you got wrong. That correction is the second step, and it is the entire point of the exercise. You now both hold the same version of the meeting, and you got there cheaply, which is the thing Clark and Brennan call the principle of least collaborative effort: participants try to minimize the combined work from the start of a contribution to its mutual acceptance.

A transcript has no second step attached. Nobody reads five thousand words of dialogue and writes back to confirm which parts they took as commitments. Sending it establishes nothing mutual. It moves the whole job of interpretation onto a reader, and it leaves the result invisible to you, which is the part that hurts later.

Speech is built on the assumption that it fades

The same paper lists the constraints a medium puts on this process. Two of them explain why a meeting record is an awkward thing to hand over. On reviewability: speech fades quickly, but in media such as email, letters, and recorded messages, an utterance stays behind as an artifact that can be reviewed later by either of the partners, or even by a third party. On revisability: in face-to-face and telephone conversation, most self-repairs have to be done publicly.

Read those two together. Everybody in a live conversation is speaking without the option to fix a sentence privately before it lands, and they speak that way because the words are going to fade. Under time pressure people produce what the paper calls improper utterances: a phrase started and abandoned, a number said and then walked back, a provisional wording floated to see whether it survives contact. The paper's own transcribed example is a speaker saying, we must ha, we're big enough to stand on our own feet now.

Clark and Brennan's comment on this is the line worth carrying around. Working out what a speaker is and is not ultimately committed to, they write, is no easy task. That is said about the person sitting across the table with the full benefit of the room, the pauses, and the face. A shared record hands the same job to somebody who was never there, months later, with none of it.

Tone goes first, and the sender is the last to notice

Justin Kruger, Nicholas Epley, Jason Parker and Zhi-Wen Ng ran five experiments on this for the Journal of Personality and Social Psychology in 2005. In the second, sixty Cornell students worked in pairs, twenty-nine pairs surviving the analysis. Each person picked ten statements they thought would be easiest for their partner to read correctly as sarcastic or serious, then either typed them or read them onto tape. Partners who heard the statements got nearly three quarters of them right. Partners who read the same statements scored at a rate the authors describe as indistinguishable from chance.

The finding that matters for a shared record is what people believed about themselves. Readers were no less confident than listeners, 89 percent against 91 percent, while performing far worse. The third study, run across 154 pairs of University of Illinois students in email, voice-only and face-to-face conditions, found the same shape: participants expected to land 88.8 percent of the emotions and tones they were trying to convey and actually landed 70.4 percent. On the comprehension side the email condition predicted 89.3 percent and hit 62.8 percent, against 73.9 percent face-to-face and 73.3 percent voice-only. Being friends rather than strangers made no difference to any of it.

That study is about email and not about transcripts, and the gap is worth naming rather than glossing. What it isolates, though, is the variable a transcript also removes. Face-to-face and voice-only produced no difference between them, which the authors read as evidence that the damage runs through intonation and vocalization rather than gesture and expression. A transcript of a call is the words with the intonation taken out. That is a reason to expect the same problem, and it is reasoning rather than a measurement.

The practical version is small and specific. The line you said lightly, to take the edge off a hard number, arrives flat. The hedge you wrapped around a date arrives as a date. And because the confidence effect sits on the sender's side, nobody on your team reviewing the file before it goes out will flag any of it, since the thing that fails is precisely their sense of how their own words read.

The person who reads it is not the person who was on the call

The contact you have a warm relationship with is not the only reader once a file exists on their side of the wall. Clark and Brennan name the third party directly in their definition of reviewability, and in practice the third party has a job title. A procurement lead assembling a file before a renewal. A successor account manager six months after your contact moves on. A legal team building a timeline. A director who was not invited to the call and now has all of it.

The timing does more damage than the audience. A long record gets read closely at the moment somebody has a reason to read it closely, and that moment is usually the one where a relationship has already gone sideways. The reader arrives looking for the sentence where your side said something that did not happen. In ninety minutes of talking there is nearly always a sentence.

What is in there that you did not mean to send

Work through an actual transcript of an external call and the inventory is longer than people expect. Your own colleagues talking to each other, including the two minutes before the customer joined. A number somebody said before anyone had checked it. Your reasoning about pricing, which is a different object from your price. The customer's own remarks about a competitor, or about a colleague of theirs. Anything that was visible on a shared screen, if your capture includes screens.

Two of those carry more weight than the rest. Internal reasoning about pricing tells a buyer where your floor is, in your own words, which is far harder to walk back than a number in a spreadsheet. And a customer's offhand comment about one of their own colleagues, handed back to them inside a document that colleague may eventually read, quietly withdraws a courtesy you never meant to withdraw.

When sending the record is the right call

There are real cases, and they have something in common. A working session where both sides were building the same artifact, and the record is the artifact, and everybody knew that while they were talking. A participant asking for their own words back, which is a request worth honoring by default. A relationship where both sides agreed up front to work from a single base text, which is normal in some regulated settings and in long technical integrations.

One more, with a caveat attached. If a dispute is live, the full record may well help you, and at that point the decision stops being yours alone and belongs with whoever handles that for your company. None of this is legal advice.

What all of those share is that the record was expected. When everybody in the room understood that the transcript was going to be a shared document, they spoke like it, and the document is a fair representation of what they were willing to put in writing. The case to be careful with is the retroactive one: a record made under the assumption that it would stay inside, sent outward later because it seemed like a helpful thing to do.

A recap that actually does the work

Send it the same day, while the customer can still check it against their own memory of the room. Keep it to what a person will read on a phone: what was decided, what is still open, and what each side owes by when. Write it in your own words rather than pasting a generated summary through unchanged, because the words you pick are the claim you are making and you should want to have picked them.

Then ask for the correction, plainly. A closing line along the lines of, if any of this does not match what you heard, tell me and I will fix it. That sentence is the acceptance step from the grounding research, dressed in ordinary language. It turns a document you wrote into a version you both hold, and the customer who writes back to fix one detail has done you a considerably larger favor than the customer who replies with thanks.

Keep your internal reasoning out of it. A recap covers what two companies agreed. Why your side was willing to agree to it stays on your side.

Capturing and sharing are two separate decisions

Quiet capture removes no obligation to be clear with the people in the room about what is being kept. That holds regardless of what you later do with the record, and it is the first decision, made before the meeting starts.

Sending the record outward is a second decision, and it does not inherit permission from the first. Somebody who was fine with you taking notes on a call has not thereby agreed that the full text of what they said travels to their own procurement team or to a third company. If they did not know it was being captured at all, sending it outward makes the original problem substantially worse rather than resolving it.

Where Driffle fits, and where it does not

Driffle captures a meeting from your own computer, with no bot joining the call and nothing to install on the other side, and it never auto-joins, auto-records, or runs in the background. Audio is transcribed in real time and discarded, so what exists afterwards is text and the notes you wrote, rather than a recording. Notes are visible only to their author until they are explicitly shared, and workspace admins see what has been shared into shared folders rather than private notes.

Sharing is one click to Slack, email, Notion, or a CRM, and that is worth saying out loud in an article like this one. One click is about the right amount of friction for handing a teammate a note and the wrong amount for a decision about what leaves the company, because the product has no idea who is on the other end of that channel.

Three limits stated plainly. Driffle does not decide what is safe to send outside your company. A record being searchable inside your team is a different thing from a record being sendable, and nothing about the first implies the second. And none of this is legal advice or a compliance position.