Meeting workflows

The failure mode is rarely a bad meeting. It's a good meeting whose decisions dissolve the moment everyone opens their next calendar invite. Harvard Business Review's widely cited analysis of executive calendars found that senior leaders now spend an average of 23 hours a week in meetings, roughly double what executives reported in the 1960s, and the same research reported that most surveyed professionals rated a substantial share of the meetings on their calendar as unproductive. Doodle's annual State of Meetings research has separately estimated the direct cost of poorly run meetings to businesses in the tens of billions of dollars a year once salaried time is counted.

None of that is an argument against meeting. It's an argument for treating what happens in the last five minutes, and the first hour after, as part of the meeting itself.

1. Pick the format before you pick the time. A status update, a decision meeting, a brainstorm, and a 1:1 each need a different structure, and most calendar chaos comes from running all four with the same 30-minute default. Decide which one you're running and name it in the invite title, so everyone shows up calibrated to what "done" looks like.

2. Capture the record, without making capture someone's job. The classic failure is delegating notes to whoever's least senior in the room, which means the person with the least context decides what's worth writing down. The fix is removing the choice: either type live decisions into a shared doc as they happen, or let the conversation be captured passively and pull the decisions out afterward. Earkeep is built for the second path: it records mic and system audio continuously in the background and transcribes it locally, so a meeting doesn't need anyone assigned to take notes, and it doesn't need a bot announcing itself in the call either. See always-on capture and calendar connection.

3. Assign every decision and commitment a named owner, on the spot. "We should follow up on that" is not an output; "Priya follows up with finance by Thursday" is. The single highest-leverage habit in any meeting is refusing to let a decision or action item leave the room unattached to a name and a rough date. This is cheap to do live and expensive to reconstruct later.

4. Close the loop before the room empties, or as soon after as possible. A decision that isn't written down within the day has a good chance of being remembered differently by two different people a week later. The last-five-minutes ritual, and what to do when nobody ran it, is the whole subject of Never leave a meeting empty-handed.

Some meeting types live or die on whether last time's context survives into this time. 1:1s are the clearest case: a 1:1 that resets to "so, how's it going?" every week is a different meeting than one that opens with "last time you were stuck on the API contract, where did that land?" 1:1s that compound covers how to structure a recurring one-on-one so it accumulates instead of resetting.

Retros and user interviews have their own version of the same problem. A retro that doesn't reference what came out of the last retro is theater; a series of user interviews that doesn't get cross-referenced for a recurring theme wastes the second and third interview's real value, which is confirming or contradicting the first. In both cases, the fix is the same one that runs through this whole page: a durable, searchable record of what was actually said, not a summary someone wrote from memory two hours later.

For user interviews specifically, the note-taking tradeoff is sharper: a researcher who's typing is a researcher who's not making eye contact or following an unexpected thread. Passive capture lets the interviewer stay present and pull quotes and themes out afterward instead of triaging them live.

A workflow this dependent on follow-through needs time to execute it, and back-to-back meetings are the most common way that time disappears. A few concrete habits show up repeatedly in research on meeting-heavy calendars:

• Default to shorter than the calendar app suggests. Every mainstream calendar tool defaults to 30 or 60 minutes; almost no meeting needs exactly that. Cutting a recurring meeting from 30 to 25 minutes, or 60 to 50, guarantees a buffer between back-to-back calls where the closing ritual can actually happen.
• Batch meeting-heavy days against focus-heavy days. Clustering meetings into two or three days a week, rather than spreading them evenly, protects the remaining days for the work the meetings were about in the first place.
• Let recurring meetings earn their recurrence. A standing weekly meeting that's drifted into a status readout everyone could get from a written update is a tax on the calendar, not a workflow. Reviewing standing meetings quarterly is one of the few calendar interventions with no real downside.
• Protect the five minutes after a meeting, not just the meeting itself. If the calendar books back-to-back with zero gap, the closing ritual either doesn't happen or happens hours later from memory, which defeats the purpose.

More on both: protecting focus between meetings and back-to-back meeting survival.

None of the workflow above requires Earkeep. It requires a format, a record, named owners, and a closing ritual. What Earkeep changes is how expensive the "record" step is: instead of assigning someone to take notes, or trusting memory, or typing during the conversation instead of participating in it, the record already exists because Earkeep was listening the whole time, entirely on your device, with the audio itself processed only in memory and never uploaded anywhere.

That's most useful for roles whose whole job is follow-through across meetings: see Earkeep for product managers for stakeholder commitments and customer discovery calls, Earkeep for engineering leads for decisions made in refinement and ad-hoc huddles, and Earkeep for people managers for the 1:1 continuity problem specifically.

If you only fix one thing this week, fix the closing ritual: Never leave a meeting empty-handed walks through agenda-to-output patterns, a decision-log template, and what to do when you realize, ten minutes after the call ended, that nobody wrote anything down. If your bottleneck is specifically 1:1s losing context week over week, go straight to 1:1s that compound.

Want the record to exist automatically, without anyone assigned to take it?