Building an async-first meeting culture

The commonly cited figure is that upwards of 40% of meetings could be replaced by asynchronous communication. Treat it as an order of magnitude rather than a measurement: it comes from surveys of employees judging their own calendars, which is a reasonable signal and not a controlled study. What makes it credible is that it matches what people find when they audit their own week — the meetings that survive scrutiny are a minority, and the rest are information transfer wearing a calendar invite.

The real argument is not the percentage. It is the arithmetic of interruption. A one-hour meeting with eight people does not cost eight hours; it costs eight hours plus the fragmentation it creates in eight days. A meeting at 2:00 PM does not consume 2:00 to 3:00, it consumes the ninety minutes before it, which are too short to start anything that would be interrupted. Async communication costs its readers only the time they spend reading, and they choose when.

There is a real cost on the other side, and pretending otherwise is why async initiatives fail. Async is slower per exchange — a decision that would take eight minutes live can take two days in a thread — and it is worse at disagreement, ambiguity, and anything with feelings in it. Async-first means defaulting to async where it is better, not everywhere.

The dividing line is whether the meeting needs simultaneous presence for its actual purpose, or only for its convenience.

The bottom four are worth defending explicitly. Teams that go async enthusiastically tend to cut them first, because they are the hardest to justify on efficiency grounds — and then discover eighteen months later that nobody knows each other.

GitLab's all-remote handbook is the most detailed public documentation of an async-first culture, and its central claim is not about tools. It is that async only works if the written record is genuinely complete, because anything decided in an unrecorded conversation immediately re-privileges the people who were in it. Four practices carry most of that:

• Write proposals, not questions. "What should we do about X?" produces a thread. "I propose X, because Y — objections by Thursday, otherwise we proceed" produces a decision. The default-to-yes clause is what converts asynchronous discussion into asynchronous deciding.
• Set response-time expectations explicitly. Async fails when it means "reply whenever," because nobody can plan around that. 24 hours for normal items, named channel for anything faster, and a rule that genuinely urgent things are a phone call.
• Default to public channels. A decision in a DM is invisible; the same decision in a channel is searchable by the person who joins in March.
• Record the meetings you do hold. A meeting that must be live is still an information source for everyone who was not there, and a recording plus a summary makes non-attendance safe. This is the piece that lets you shrink invite lists without shrinking the audience.

The transition rarely goes from ten meetings a week to four in one move. What works better is concentrating the meetings that remain, so the async time is contiguous rather than distributed as unusable twenty-minute crumbs.

• Meeting days and maker days. Pick two days for synchronous work and defend the rest. A week with three meeting-free days is qualitatively different from a week with fifteen scattered free hours, even when the totals match.
• A team-wide no-meeting day. The reason to make it collective is that an individual no-meeting day fails the moment someone schedules something important on it. A shared one has a norm behind it.
• Core hours, if you span timezones. A four-hour overlap for the meetings that must be live, with everything outside it async by construction rather than by policy.

The mechanics of protecting the resulting blocks are in protecting focus between meetings.

Every practice on this page rests on the same foundation: what happened in a live conversation has to be available to people who were not in it. Where that fails, async culture inverts — attendance becomes the only reliable way to know what is going on, and meetings grow rather than shrink, because being in the room is the safe choice.

Most teams try to close that gap with volunteer note-taking, which is inconsistent by nature: the meetings least likely to be written up are the informal ones where the interesting decisions get made. Earkeep closes it differently. It records mic and system audio continuously while it runs and transcribes locally on your device, so every live conversation leaves a searchable record without anyone assigned to produce one — no bot to invite, no start/stop to remember, which also means the ad-hoc call that turned into a decision is covered. Meetings are defined afterwards by selecting a span on a timeline, or automatically from a connected calendar. Audio is processed only in memory and never written to disk, transcripts are plain files, and an agent can turn one into the written update the async channel needs — see meetings after the fact and agents and MCP.

Async is a decision about which meetings exist; meeting workflows covers making the survivors produce outputs, and time management for meeting-heavy roles covers defending the time this frees up. The written standup formats referenced above are in standup and retrospective formats.

Download Earkeep free for macOS so the people who weren't in the room can still find out what was decided.