I read Lloyd's "The Coordination Tax" last night and couldn't stop thinking about one number: 88%.
A CTO mapped how she spent a month. Architecture reviews, hiring decisions, strategic technical calls — the stuff you'd assume IS the job — accounted for 12% of her time. The other 88% was coordination overhead. Team syncs. Vendor negotiations. Budget reconciliation. Roadmap alignment.
Lloyd's thesis is that AI agents eliminate this tax by collapsing the distance between decision and execution. The three-person company will outcompete you — because it operates at 4% of the headcount with equivalent output.
I agree with the diagnosis. But I think the prescription misses something.
The coordination tax isn't overhead. It's the whole job.
If you're a manager, coordination IS your job. You exist to maintain context across people, projects, and time. You're the human buffer between a decision made in February and its consequences in June. You're the one who remembers that Sarah is frustrated about PR review times, that the checkout outage in March had the same root cause as November's, that Dana starts Monday and needs the team docs.
The three-person company doesn't need managers. The other 99% of companies do. The real question isn't "can we eliminate coordination?" It's "why does coordination cost so much?"
The friction isn't in the coordinating. It's in the remembering.
Context loss — you had a great conversation with a direct report three weeks ago. They mentioned feeling stuck. You meant to follow up. You forgot.
Tool tax — you want to track a task for someone. This requires opening Jira, creating a ticket, filling in fields. The task takes 10 seconds to describe. The tool takes 90 seconds to feed. You start skipping the tool. Now things fall through cracks.
Knowledge decay — your team made a decision about the on-call rotation six months ago. The reasoning is buried in a Slack thread. Someone new asks why. Nobody remembers. The decision gets relitigated from scratch.
Every "productivity tool" promises to fix this. Most of them make it worse. They solve coordination problems by creating new ones. Every tool demands feeding — structured data in, dashboards out. The tool becomes the job.
The interface is the bottleneck.
I've been building Manager — a thinking partner embedded in Slack. The thesis is simple: you're already having the conversations. The conversations should be the input.
When I mention that Sarah is frustrated about PR review times, Manager captures that as a private observation. When I prep for my 1:1, it surfaces that note alongside her open tasks and recent context. I didn't open a separate app. I didn't fill in a form. I thought out loud, and the thinking became structured knowledge.
When my team makes a decision — "we're keeping PagerDuty, dropping OpsGenie, going follow-the-sun" — Manager saves it as a knowledge doc with the reasoning. Six months later, when someone asks why, the answer exists. It didn't require anyone to write documentation. It required a conversation.
This is a fundamentally different model from every project management tool I've used. Jira, Asana, Notion, Linear — they all start from structure and ask you to fill it in. Manager starts from conversation and builds the structure for you.
What I learned from 30 PRs in a day.
I run a small team. We shipped 38 pull requests in a single day last week — not because we're superhuman, but because our AI coding agents handle implementation while we focus on judgment. The coordination tax between me and an AI agent is near-zero. I describe what I want. It explores, writes the code, opens a PR. I review the output.
But here's what surprised me: the hardest part wasn't the coding. It was the context. Making sure each agent had the right background. Remembering what we decided yesterday. Tracking which PRs were related. Knowing what to work on next.
The coordination tax didn't disappear when we added AI agents. It concentrated. All the context that used to be distributed across a team now needs to live somewhere accessible — and that somewhere can't be another tool you have to feed.
The companies that win.
Lloyd is right that the three-person company has a structural cost advantage. But the bigger insight is what happens to the companies that still need ten, fifty, or five hundred people.
In those companies, every manager carries a coordination tax. And the size of that tax is directly proportional to how much friction exists between thinking and knowing — between having an insight and being able to act on it six weeks later.
The companies that win won't just be the smallest ones. They'll be the ones where management overhead shrinks to judgment. Where the manager spends 60% of their time on decisions that require their specific accumulated context — not 12%.
The coordination tax doesn't have to be the whole job. But only if the tools stop adding to it.