What Copilot does to meetings

Recaps changed attendance and the meeting record before any organisation decided they should, and both changes need a deliberate answer.

Meeting behaviour changed before anyone approved the change. Nobody held a workshop on whether attendance should become optional or whether an automated summary should count as the record. Both happened anyway, in the first few weeks, driven by individuals making sensible local decisions. The organisational consequences arrived later and landed on people who had not been consulted.

Attendance became optional and nobody signed that off

The behaviour spreads quickly because the first version of it is genuinely good. Someone is double-booked, declines, reads the recap, and loses nothing. Within a month, a meeting that used to have nine people in it has four, and an invite list of fourteen.

Some of that is a gain. Plenty of people were in meetings for no reason, and a reliable summary is a better use of their forty minutes than sitting on mute. I would not undo it.

The problem is that reading is not attending. In a status update, the recap is a fine substitute for being there. Where the decision is actually made, it gives you the conclusion without the reasoning. Three people argue, change their minds, and land somewhere none of them started. The person who reads about it afterwards inherits an action they did not shape. Six weeks later somebody says that is not what we meant, and they are right, and so is the person who read the summary.

A team I worked with hit this repeatedly before anyone named it. Their fix was blunt and it worked: every recurring meeting got labelled a briefing or a decision meeting. Briefings: read the recap, do not attend, invite whoever you like. Decision meetings, six people maximum, attendance expected, no proxy by summary. The alternative was a slow argument about accountability conducted entirely by email.

If a meeting is worth an hour of six people’s time, it is worth deciding whether reading about it afterwards is the same thing.

The recap is a summary of what was said

It is not a summary of what was agreed. Those are different documents and only one of them is a record.

A decision log is a deliberate act. A person with authority states the decision, the owner and the date, and writes it down in a place that survives. A recap is a faithful account of a conversation, including the parts that went nowhere. Where a meeting ended ambiguously, a good summary reports the ambiguity, so two people can leave with opposite understandings and both cite the same document. I have watched exactly that argument, and the summary was not wrong. It was doing its job, just not the job people had quietly assigned it.

The correction is thirty seconds long and belongs at the end of the meeting, not after it. Chair says: what did we decide, who owns it, by when. Somebody types those three things while everyone is still in the room. Then the recap becomes what it is good at: context, tone, what was raised and by whom. The decision log stays a human artefact with a name against it.

There is a verification cost either way, and it is the number left out of the benefits case. Count the checking, or do not claim the saving.

Recording a meeting is processing your colleagues’ data

The ICO’s guidance on monitoring workers, last updated on 16 June 2026 and carrying a notice that it is under review because of the Data (Use and Access) Act, sets out the tests. You need a lawful basis. You must tell workers about monitoring “in a way that is accessible and easy to understand”. A DPIA is required before any processing likely to result in high risk. And a proportionality point: “Just because a form of monitoring is available, does not mean it is the best way to achieve your aims.”

Transcription of every internal conversation is available. That does not make it proportionate. The honest question is not whether you may, but what you would say if asked why you needed a searchable transcript of a one-to-one.

Transcripts and recaps are records, and records are disclosable. Microsoft’s compliance surface for Copilot includes Purview audit and eDiscovery, which is the right answer for governance and also means these artefacts turn up in a dispute. Write the retention rule before there are fifty thousand of them, because retention decided retrospectively is made under pressure by a lawyer who was not in the room.

UK policy is also moving towards more consultation, not less. The Department for Business and Trade’s consultation on workplace monitoring technologies opened on 8 July 2026 and closes on 30 September 2026, and one option on the table is a legislative duty to consult and negotiate with workers on monitoring technologies. No such duty exists today. Deciding your meeting-recording position by talking to staff now costs a fortnight and removes a future rewrite.

Where an AI system takes part in a conversation with people rather than summarising it afterwards, the Article 50 obligation to tell people they are dealing with an AI system has been in force since 2 August 2026, unless that would be obvious to a reasonably well-informed, observant and circumspect person. Worth knowing before someone puts an agent into a customer call.

What I would write down this week

Three lines, and none of them require a project.

Label every recurring meeting a briefing or a decision meeting, and let the label govern attendance. Nominate a person, not a tool, to record decisions in the last two minutes. Publish the retention period for transcripts and recaps, and the basis on which meetings are recorded, in language a person can read without help.

The tooling is not the hard part and never was. The hard part is that summarising a meeting well removes the pressure that used to force people to conclude one.

CRAIG STANLEY

Written 8 August 2026 in the North East of England. If something here is wrong, tell me and I will correct it on the page rather than quietly.