Clément Sauvage has a blunt answer to the question every tech leader is asking this year: if AI agents make each person several times faster, why don’t companies ship several times faster? Because the work was never the whole job. The rest is people agreeing with each other, and that coordination cost doesn’t shrink when the code writes itself. It lands, as it always has, on your calendar.
Agents speed up the work, not the agreeing
In “Agents are here to stay. How can companies really go faster?”, Sauvage argues that “human sync” (alignment, reviews and meetings) has always been the real bottleneck, and that agents may make it worse. He leans on two old laws of software engineering:
- Amdahl’s law. If half of a team’s time goes on coordination and agents make the other half ten times faster, the whole thing only gets about 1.8 times faster. Coordination then takes roughly 90% of the time that’s left, and no amount of extra agent speed gets you past 2x.
- Brooks’s law. Communication channels grow much faster than headcount: 5 people have 10 of them, 10 people have 45, and 50 people have 1,225.
The twist is that agents raise each person’s output without raising their ability to align with everyone else. More output means more to review, more to discuss and more to agree on. His conclusion: small, autonomous teams, and broad roles like the Product Engineer, who owns a feature from the idea to measuring its impact, are best placed to turn agent productivity into real speed. He admits the maths is illustrative. You don’t really need the maths to believe it, though. You only need to open your week view.
Where the coordination cost shows up
Every one of those channels eventually becomes a calendar event: the stand-up, the review, the “quick catch-up” that is never quick. For someone on one team, that’s a crowded calendar. For the people we build for (freelancers, consultants and anyone with more than one client or job), it’s several crowded calendars, owned by organisations that can’t see each other. Client A’s review gets booked on top of client B’s planning session because neither calendar knows the other exists.
That is the problem availability mirroring solves. When you’re busy in one connected account, Calendrz blocks that time in all the others, without sharing what the meeting is about. It won’t cancel a single meeting (we’d love to, but that’s above our pay grade), but it stops the meetings you can’t avoid from colliding.
What we learnt building Calendrz with agents
The article matches what we see from the builder’s side. Calendrz is built by a very small team: our founder, Liviu Tudor, working with AI agents. Brooks’s law barely applies to a team that size, so almost all of the speed the agents add turns into shipped work. A few habits make that possible:
- Specs instead of meetings. Bigger features start as written plans. Our Smart Scheduler was planned and shipped in three documented phases, each saying what lands, how it’s switched on and what must be true before the next one starts. An agent can read a plan; it can’t read the room.
- Specialists with clear lanes. We use separate agents for the backend, the frontend, the database and GitHub, each with its own written rules. They never need a meeting to align, because the boundaries are written down.
- Review without a review meeting. Every pull request gets an automated AI code review when it’s opened, before a human looks at it. And every night an agent reads the last 24 hours of production logs and, when it finds something worth fixing, opens a pull request with the fix. Dozens of small fixes have shipped that way, each one reviewed on its own.
None of this means the human disappears. Liviu still decides what to build and approves what ships. But the decision takes minutes, not a week of calendar invites. We believe AI rewards velocity: companies that still wait for a committee to vote are going to be overtaken by small, agentic teams that write things down and get moving. Sauvage’s Product Engineer is, more or less, a description of how Calendrz gets made.
What to take from it
- Measure your coordination, not just your output. If agents doubled your output and your calendar doubled too, you’ve moved the bottleneck, not removed it.
- Write it down. A clear spec replaces a surprising number of meetings, for people and agents alike.
- Protect the calendars you have. If you work across several organisations, the meetings you can’t cut shouldn’t be allowed to land on top of each other.
Amdahl’s law was written about computers in 1967. It’s a little humbling that it now describes our Tuesday afternoons so well 🙂
Juggling more than one calendar? Connect your calendars to Calendrz and let auto-blocking keep your meetings from double-booking you.







