Journal / Technology

Technology

When the Meeting Is Recorded by Default

Automatic transcription inverted the default: meetings are now on the record unless someone objects. That changes what gets said, what your company holds, and what you should decide before the next call.

10 min readResolution

Until recently, recording a meeting was a deliberate act: someone pressed a button and everyone knew. Automatic transcription inverted that. The tool joins the call on its own, announces itself through a notification nobody reads, and the meeting becomes searchable text more or less permanently.

Nobody decided on that change. It arrived as a default setting.

This is not an article about which tool is best. It is about three things that change when recording becomes the default, and the decisions worth making before the next call.

What changes in behaviour

This is the effect the category discussion ignores, and it is the one that matters most if you depend on meetings to sell.

People stop thinking out loud. The useful part of a diagnostic meeting is precisely the half-formed hypothesis, the complaint about an internal process, the remark about the current supplier. None of that is said with the same freedom when a record exists. You still have the meeting; you no longer have the information.

The conversation becomes more formal and less informative. The mechanism is the same one the behavioural literature describes for observation generally: knowing you are observed changes the behaviour being observed. Applied to a commercial meeting, it means the perfect transcript is of a worse conversation.

Silence changes function. The pause that produced elaboration becomes a recorded pause, and the effect described in Silence Is the Cheapest Question weakens considerably.

quality of the conversation quality of the record how permanent the record feels →
fig. 01 — the record improves as the conversation degrades

What changes in liability

This part is dry and worth the attention.

Consent rules are not uniform. Requirements for recording a conversation vary by jurisdiction, and an international call can involve more than one at the same time. Where personal data is processed, the applicable data-protection regime brings obligations on legal basis, purpose and retention. This is the one area where getting it wrong is expensive, and it is worth checking against your own jurisdiction rather than against an article.

A transcript is a document. A sentence said off the cuff and captured verbatim becomes something that exists, is searchable, and can be requested. "I think we could deliver in two weeks," in an exploratory conversation, carries very different weight once it is written down with a timestamp.

Retention without a policy is infinite retention. Most tools keep transcripts by default and indefinitely. That means your company holds an archive of what clients said, with no disposal schedule and nobody responsible for it.

Third parties in the room. When you bring the tool to a meeting, you are processing data belonging to people who did not choose that tool and almost certainly did not read its terms.

What we recommend

Not "use X, avoid Y." These are decisions worth more than the choice of vendor:

  1. Write the policy on one page. When you record, when you do not, how long you keep it, who can access it. Without that, your actual policy is the default setting chosen by whoever installed the tool.
  2. Do not record the first diagnostic call. It holds the information you most need and the information that most reliably disappears under recording. Take notes by hand. Record from the second meeting, when the conversation is about scope and dates — which is exactly where verbatim capture helps.
  3. Ask with your voice, not with a notification. One sentence at the start — "I'm using automatic transcription, is that alright?" — settles consent and, more usefully, returns to the other person the knowledge that they are on the record. If they hesitate, you have just learned something.
  4. Set short retention by default. Ninety days covers nearly every real use. A longer archive is liability without a counterpart.
  5. Treat the transcript as a draft. Errors in names, numbers and figures are common, and the written version acquires an authority it has not earned. What counts is the summary a person reviewed.
  6. Give the tool an owner. Same logic as the subscription audit: a tool that processes client data with no owner is a problem with a date on it.
internal execution record search beats memory first diagnostic call do not record and say so out loud scope and terms record, ask first literal wording helps here
fig. 02 — the decision by meeting type

What the category genuinely solves

To be fair to the technology: for internal meetings, team alignment and handoffs between people, automatic transcription is a clear improvement. Nobody goes back to writing minutes by hand once they have full-text search over everything that was agreed.

The error is applying the same default to the conversation whose value lies in being unprotected. Internal execution meeting: record it. A conversation where someone needs to admit a problem: do not, and say that you are not.

That distinction is the entire policy. The rest is vendor selection.

Further reading

Books that shaped this article, including the ones we disagree with. Where a work is popular rather than peer-reviewed, we say so.

Shoshana Zuboff — 2019
Long, polemical, and the clearest available account of how default data collection becomes normal without anyone agreeing to it. Read critically.
Edgar H. Schein — 2013
On the conditions under which people tell you something true. Worth reading precisely to see what a permanent record removes.
Erving Goffman — 1959
Being observed changes the performance. Goffman worked this out before recording was cheap; the argument survives the technology.

Resolution is a participant in the Amazon Services LLC Associates Program. As an Amazon Associate we earn from qualifying purchases — at no additional cost to you. Affiliate links never determine what appears on these lists: several of these books are here specifically because we think they are wrong in an instructive way.

References & notes

  1. Consent and data-protection requirements for recorded conversations vary by jurisdiction and change; check the rules applying to you rather than relying on any general article.
  2. The behavioural effects described here are drawn from our own commercial practice.
  3. On observation altering the behaviour observed, see the discussion in Baseline First.

Corrections are published inline and dated. Write to us if something here is wrong.

// weekly dispatch

One email. Every Tuesday.

The week's analysis, one tool we actually tested, and one behavioural pattern worth practising. Unsubscribe in one click.

// no spam · no resale · ~4 emails a month