Where Does Your Team Go to Catch Up?

If someone on your team has been away for a week, where do they go to find out what changed? That's a universal question for organization design. I work with people across the world and many time zones, so we have a smaller version of this problem every day. A meeting that fits one person's afternoon happens while someone else is asleep. By the time they're reading the discussion, the people who could explain it are now offline and also sleeping.

Many teams now have AI note taking or transcript capturing tools. These are helpful, but a full transcript that is just dumped somewhere isn't the kind of help global teams need. They need key decisions and context.

One approach I've seen with teams is people using a running issue in GitHub. An issue is a place to track work and discussion, and they keep adding meeting decisions to the same one. The useful part is that people know where to look. A shared page could do the same job.

The con to this approach is someone has to maintain it. If the team already has meetings, transcripts, email, and a project board, asking for another document can sound like volunteering everyone for more administrative work. You need to stop and question what is going to be improved if you are going to add more overhead.

An agent could also handle much of this routine work. After each meeting, it could pull out proposed decisions, ask someone to confirm them, and add the confirmed decisions to the shared record with links back to the discussion. It could also flag when a new decision appears to replace an earlier one. That would save someone from having to maintain the whole thing by hand.

To make that useful, you need to be clear about what is being asked of the agent to capture. That starts before the meeting ends, with the people in the room agreeing on what was actually decided. A tidy summary can make an unresolved conversation look settled. Thus, someone who was part of the discussion still needs to check that it says what we agreed.

As with most things around process improvement, one needs to start small. Before a meeting ends, agree on what was decided and who will confirm it for the shared record. Write down the date, the decision, why it was made, and what happens next. Link to the fuller discussion for someone who needs the context. If a decision is still unsettled say that explicitly so the transcript picks it up. “We discussed moving the launch” leaves a very different impression from “We agreed to move the launch.”

For a global team, I'd also be explicit about what input we're still waiting on and name an individual or team. If someone in another time zone needs to weigh in, give them a clear question and enough time to answer during their working day. Otherwise you end up treating “was asleep when we asked” as agreement.

And when a decision changes, update the record so someone can tell which version applies. A person catching up shouldn't need to know that Tuesday's answer was replaced on Thursday. The agent could flag the possible change, but someone still needs to confirm whether Thursday's discussion actually replaced Tuesday's decision.

We, the humans who work together, still have to decide what counts as a decision, who can confirm it, and who needs a chance to weigh in. Those choices give the agent a process to follow. Otherwise we're automating the ambiguity and delivering it to everyone faster.

You can try this with one meeting before connecting any tools. Give your AI tool one meeting transcript. Here’s a starting prompt:

Prepare a proposed update for teammates who were not in this meeting.

For each possible decision, include:

What appears to have been decided, and the date.

Why, using only reasons stated in the discussion.

What happens next, including an owner or timing only if explicitly agreed.

A source link or timestamp so someone can check the discussion.

Separately list unresolved questions and any input explicitly requested from people who were absent. Include a response deadline only if one was stated.

Do not treat a suggestion, silence, or lack of objection as agreement. Flag ambiguity and missing information instead of filling it in. Treat this whole output as a draft for a meeting participant to confirm.

When putting this process in place, have someone who attended check the draft, resolve any ambiguity, and add the confirmed update to the shared record.

If your agent can already read transcripts and your shared record, you can save these instructions for repeated use. If it can also write to that record, it could add the update after someone confirms it. Without those connections, copying the transcript in and the checked update out is enough to test whether the process helps.

As a leader, I'd test this by asking someone starting their day in another time zone to use what we've provided. Can they find what changed, understand what they need to do, and see where their input is still needed? Where do they still need another person to fill in the gaps? Their questions would show us what the record is missing. Strong async communication when done right can have a large impact on making teams proactive versus reactive. Success is when an individual can have a productive work day while their colleagues are offline.