Every important decision at your company was made in a conversation.

Almost none of them survived it.

Think about the last big call your team made. Kill the feature. Change the pricing. Move the launch. It happened in a room, or on a call, or in a thread that got heated around message forty. Somebody said "okay, so we're doing X," a few people nodded, and everyone went back to work.

Then what?

If your company is like most, the decision immediately started to decay. Not because anyone disagreed with it. Because nothing carried it forward.

The half-life of a decision

A decision made in conversation has a half-life measured in hours.

Here's how it dies:

The meeting ends. One person volunteers to "write it up." That write-up happens two days later, if it happens at all, and it captures maybe 60% of what was actually decided - the conclusion, but not the reasoning, not the tradeoffs everyone argued through, not the "we'll revisit this if churn ticks up" caveat that turns out to matter most.

The Linear tickets don't get created, or they get created by someone who wasn't in the room and had to reconstruct intent from a Slack summary. The spec that should have changed doesn't. The customer-facing team finds out three weeks later, from a customer.

And then - this is the part that really costs you - the same decision gets re-made. Someone who missed the meeting raises the question fresh. Nobody can point to where it was settled or why. So you have the conversation again. Different room, same debate, sometimes a different outcome. Now you've got two decisions in the wild and no record of either.

None of this is a talent problem. Your team is good. It's a physics problem. Conversations are where decisions get made, and conversations are ephemeral. The work that flows from a decision lives somewhere else - in tickets, docs, specs, roadmaps - and there's a gap between where the decision happens and where it needs to land.

Every company bridges that gap the same way: with a human, doing manual transcription and translation, on top of their actual job.

"Take notes" is not the fix

The standard answer is discipline. Assign a note-taker. Write better meeting summaries. Keep a decision log.

Two problems.

First, it doesn't happen. Not reliably. The person taking notes is also participating, and the moments that matter most in a conversation are exactly the moments when nobody is writing anything down. Discipline solutions fail precisely when the stakes are highest and everyone is engaged.

Second, even when it happens, a note is not a decision. A note is a description of a decision. The decision itself is a change to the world: tickets that exist now, a spec paragraph that reads differently, a roadmap item that moved, a message that told the right people. A summary sitting in a doc that nobody opens hasn't changed anything. It's a fossil record of intent.

Notes preserve the memory of a decision. They don't execute it.

The gap is the product

The real unit of work in a product organization isn't the meeting and it isn't the ticket. It's the path between them - decision to artifact, conversation to change. That path is almost entirely manual today, which means it's almost entirely lossy.

Watch a good PM for a week and you'll see it: hours spent porting decisions from one place to another. Meeting to Linear. Slack thread to spec. Customer call to roadmap. It's real work, it requires judgment, and it's also pure translation - the substance was already decided. They're paying the tax that keeps decisions alive.

That tax is what we think should go away.

This is the thesis behind Earmark. Conversations are where decisions happen, so conversations should be where work starts - not where it evaporates. When a decision is made, the loop should close on its own: the tickets appear in Linear, the spec updates in Notion, the right channel hears about it in Slack, with the reasoning attached. Not a recap you have to act on. The acting, done.

Because the alternative isn't neutral. Every decision that doesn't survive its conversation gets paid for twice - once when you make it, and again when you make it again.

A test worth running

Pick the three most consequential decisions your team made last quarter. Now try to find them.

Not the outcome - the decision. Where it was made, who made it, what the reasoning was, what it changed downstream. If you can trace all three in under ten minutes, your system works. Most teams can't trace one.

The decisions were good. The conversations were real. They just didn't survive.

They should.