The Meeting Problem | Oatmeal Blog

Otter.ai has transcribed over 1 billion meetings.

Let that sink in. One billion conversations. 50 billion minutes of people talking to each other in conference rooms, Zoom calls, and huddles. Roughly one meeting for every eight humans on Earth.

This is either a triumph of productivity tooling or a symptom of something deeply broken. I think it’s both.

The numbers

Microsoft’s 2025 Work Trend Index paints a picture:

3x more meetings than 2020. The pandemic didn’t cause this—it accelerated something that was already happening.
57% of the workday spent on communication: meetings, email, chat. 43% left for actual creation.
275 interruptions per day. An interrupt every two minutes during core work hours.
68% of workers say they don’t have enough uninterrupted focus time.

The heaviest meeting users—the top quartile—spend 7.5 hours per week just in meetings. That’s almost a full workday, every week, sitting in rooms (real or virtual) talking.

And it’s getting worse, not better. Meetings after 8pm are up 16% year-over-year. Sixty percent of meetings are ad-hoc. One in ten gets scheduled at the last minute.

The cognitive cost

Here’s what the productivity stats miss: meetings don’t just consume the time you’re in them.

Gloria Mark at UC Irvine found that after any interruption, it takes an average of 23 minutes and 15 seconds to fully refocus on deep work. But meetings aren’t ordinary interruptions—they engage your prefrontal cortex heavily with social processing, decision-making, and multitasking.

After a draining meeting? Recovery can take 45 minutes or more.

Do the math. A one-hour meeting doesn’t cost you one hour. It costs you the meeting plus the recovery time on both sides. A 60-minute call in the middle of your afternoon might actually consume 2+ hours of productive capacity.

Multiply that across an organization. A company of 1,000 knowledge workers, each in 10 meetings per week, each meeting costing 30 minutes of hidden recovery time—that’s 260,000 hours of lost productivity per year. Per thousand employees.

The experimental evidence

Some companies have tried to fight back.

Shopify implemented “meeting-free Wednesdays” and canceled all recurring meetings with more than three people. Result: time in meetings dropped 33%. The company cut 76,000 hours of meetings for the year and estimated a 25% increase in completed projects.

MIT Sloan studied companies that reduced meetings by 40%. Productivity increased by 71%. Not because people worked more hours—because they felt more autonomous and less drained.

A study across 76 companies (25,000 employees) found that just one no-meeting day per week improved autonomy, communication, engagement, satisfaction, and reduced stress. Productivity went up. Micromanagement went down.

The pattern is clear: fewer meetings = more done.

So why can’t we stop?

The meeting trap

Meetings persist because they solve a real problem: synchronization.

When you’re working on something complex with multiple people, you need to align. What are the goals? Who’s responsible for what? What decisions need to be made? These conversations are genuinely valuable.

The issue is that we’ve conflated synchronization with communication. We schedule meetings for things that should be async documents. We hold standing meetings because “we’ve always had them.” We invite 12 people when 3 would suffice because excluding anyone feels political.

And critically: we have no memory.

Most of what’s said in meetings evaporates. The person who took notes captured maybe 10% of the discussion. The action items got lost in someone’s Notion page. Three weeks later, you have the same conversation again because nobody remembers the conclusion.

Meetings proliferate because decisions don’t stick.

The memory layer

This is what got us interested in meeting transcription in the first place.

Not productivity optimization. Not “AI note-taking.” Memory.

What if you could actually remember what happened in a meeting? Not someone’s interpretation. Not bullet points. The actual conversation. Searchable, referenceable, connected to the decisions that came out of it.

The value isn’t in the transcript itself—it’s in what the transcript enables:

You can skip the meeting and catch up later (async by default).
You can search “what did we decide about pricing?” across six months of conversations.
You can stop holding status update meetings entirely—record once, share the recording.
You can settle disputes about what was actually said.

Transcription doesn’t fix the meeting problem. But it changes the math on which meetings you actually need.

What we’re building

Oatmeal started as a question: what if meetings had perfect memory?

Not cloud-processed memory that lives on someone else’s servers. Not a bot that joins your call and makes everyone self-conscious. Just… your meetings, transcribed locally, stored on your machine, searchable.

The hypothesis: if meeting content is captured and accessible, you can have fewer meetings. Async becomes viable. “Let’s just get on a call” becomes “let me check what we discussed last time.”

We’re early. The product is simple—press Cmd+Shift+M to record, get a transcript and AI summary. But the direction is clear: make meetings memorable so you can have fewer of them.

The goal isn’t zero meetings

I want to be clear: the goal isn’t to eliminate meetings. Some conversations genuinely need to happen in real-time. High-stakes negotiations. Creative brainstorms. Difficult feedback.

But the goal is fewer, better meetings. Meetings that happen because they need to, not because we forgot what we decided last time. Meetings where everyone’s present because the content matters, not because the calendar invite exists.

The meeting problem is a memory problem. And memory is something software can solve.