How to run meetings that actually decide things

You walk out of a one-hour meeting feeling productive. Two weeks later, three of the eight attendees recall a different decision than you do. The work splits across two threads. The fix takes another meeting. Multiply by every recurring sync your team has, and you start to understand why the annual cost of one weekly meeting can hit five figures while delivering almost nothing.

The problem isn't meetings. The problem is meetings that don't decide. After looking at hundreds of recurring meetings — through transcripts, action item lists, and post-mortems on shipped vs. dropped initiatives — the same five things separate meetings that produce decisions from meetings that produce hangovers.

Why decisions evaporate

Before the framework, the diagnosis. There are four common failure modes, and any one of them is enough to dissolve a decision that everyone in the room believed they made:

1. No agenda. Not "no Google Doc". No shared model of what the meeting needs to resolve. Without it, conversation drifts toward the most vocal participant's concerns.
2. No owner. The decision is made in a room of equals. When everyone is responsible, no one is.
3. No scribe. Decisions get spoken, not written. A week later, two attendees remember "we agreed to ship in Q3" and one remembers "we agreed Q3 was our target if testing went well".
4. No recap. The meeting ends when the calendar block ends. Nobody confirms back what was decided, who's doing what by when. Whatever was in the air stays in the air.
5. No record. Three months later, no one can find what the team decided. The decision didn't just evaporate — it never got written down where future-you would look.

Notice these aren't tooling problems. You can do all five with a notebook and a thirty-second discipline at the end of every meeting. You can also fail at all five with the most expensive meeting platform on the market. The discipline is the work; the tooling just makes it cheaper.

The five-step framework

Here's what high-functioning teams do — explicitly or by accident — every time a meeting needs to produce a decision.

1. Send the agenda the day before

Not the morning of. The day before. The point is to give people enough time to say no to the meeting. If the agenda is "discuss the redesign", a thoughtful invitee should be able to reply: "I have nothing to add to that — drop me from the invite". You want that reply. The wrong people in a decision meeting is the cheapest way to ruin it.

An agenda that produces decisions has three traits:

• It's framed as questions to answer, not topics to discuss.
• Each item has a target outcome: "decide", "align", "inform".
• It's short enough to fit in the calendar invite description, not a separate doc.

Three lines beats three pages. "Decide which CDN we'll use for the asset migration. Align on the rollout window. Inform everyone of the post-launch on-call rotation." Done.

2. Name the decision owner up front

Before the discussion starts, the chair says: "For the CDN decision, X is the owner. We'll discuss, but X makes the call." This sounds undemocratic. It isn't — it's the opposite. It tells everyone in the room that disagreement is welcome now, because once the meeting ends, the decision is final and X is on the hook.

The alternative — "we'll decide as a group" — produces consensus theatre. Everyone nods. Two people leave unconvinced. They re-litigate the decision in Slack the next day. The decision drifts. The owner pattern kills that drift before it starts.

3. Designate a scribe — not the chair

The chair is steering the conversation. Asking them to also write decisions in real time is asking for both jobs to suffer. Pick a scribe at the start of the meeting. Their entire role is:

• Capture each decision as a single sentence in the active voice.
• Capture each action item with an owner and a due date.
• Capture each open question that didn't get resolved (these become next meeting's agenda).

Notice what's missing: there is no "summary of what was discussed". Discussions are noise; decisions, actions and open questions are signal. Capturing the signal is faster than capturing both, and the signal is what you'll need a month from now.

4. The 60-second recap

In the last minute of the meeting, the scribe reads back the list out loud. Decisions, action items, open questions. Word for word. People object — that's the point. "Wait, I didn't agree to ship by the 15th, I said the 15th was a stretch goal" — caught in the room, in 30 seconds, instead of next sprint planning.

This is the highest-leverage minute in the meeting. It's also the one most teams skip because the calendar block is ending and someone needs the room. Schedule the meeting for 50 minutes, not 60. Use the last ten for the recap and you'll claw back the next two weeks.

5. Write decisions where future-you will look

The recap goes somewhere durable. Not a Slack thread that disappears below the fold. Not a Google Doc nobody bookmarked. Somewhere your team's "system of record" lives — wherever you check when you ask "what did we decide about X?".

For most teams the right answer is one of:

• A dedicated channel or page in your wiki, one entry per meeting, with the date and the decision list.
• Tickets in your tracker, one per action item, linked back to the meeting.
• A meeting-platform that captures this automatically (more on this below).

The format matters less than the discipline of always the same place. If decisions sometimes go to Confluence, sometimes to Slack, sometimes to email, you don't have a system of record — you have an archaeology problem.

Where tooling helps

Three of the five steps — scribe, recap, record — are mechanical. Anyone can do them. But humans are bad at doing mechanical things consistently for a year, especially when the meeting goes long and the calendar block is ending.

This is the gap modern AI meeting tooling fills. PeopleWorks Meeting, the platform we build, captures the audio, transcribes it in real time across 125+ languages, and uses an in-meeting co-pilot to surface decisions, action items and open questions as the conversation happens. The scribe role doesn't disappear — somebody still confirms accuracy at the end — but the cognitive load drops to almost nothing.

The recap becomes a 60-second review of what the AI captured rather than a frantic typing exercise. The record is automatic: every meeting ends with a structured note (decisions, actions, questions, mind map) saved to your workspace, searchable across sessions and projects.

None of this matters, of course, if you skip steps 1 and 2. No tool will fix a meeting that didn't have an agenda or an owner. The mechanical 3-4-5 is the part that benefits from automation; the strategic 1-2 is the part that benefits from discipline.

Try it tomorrow

For the next meeting on your calendar, do three things:

1. Reply to the invite asking the organiser for the three questions the meeting is supposed to answer. If they can't list them, propose dropping the meeting.
2. If the meeting happens, ask in the first minute: "who owns each decision?" Watch the room reset.
3. At minute 50, even if you're not the chair, say: "let's recap quickly — what did we decide?" Read back what you heard. Record it.

Three small interventions. They feel awkward the first time. By the third meeting, your team starts asking the questions for you. By the tenth, you start to notice which recurring meetings have decisions and which have only discussion — and you can start cancelling the latter.

And if you want a sharp reminder of why this matters: run your most painful recurring meeting through the meeting cost calculator. Multiply by 52 weeks. The number tends to focus the mind.

---