A few years ago I worked at a company that was flush with venture capital. Product was moving slower than the board wanted, so we did the obvious thing. We hired. Aggressively. We brought in engineer after engineer, mostly junior, and stood up new teams around them.
You can guess what happened. Despite that we had more software engineers on the team, our velocity actually slowed.
This wasn’t a mystery. Fred Brooks wrote it down in 1975:
“Adding manpower to a late software project, makes it later.” — Frederick P. Brooks Jr., The Mythical Man-Month: Essays on Software Engineering
I’d read the book. I learned the lesson anyway, the hard way, watching it happen in real time. The bottleneck was never the number of hands. It was the communication overhead. More people meant more meetings, more Slack threads, more design docs reviewed by more reviewers, more time spent reaching consensus before anyone could ship anything. And the onboarding cost on the new hires (especially the junior ones) ate into the senior engineers we already had.
Right now everyone is talking about how AI makes individual employees more efficient. Faster code. Faster spreadsheets. Faster contracts. Faster marketing copy. It’s a real story and worth telling.
But it’s the small one.
The bigger story is what happens to the size of teams.
I co-founded a company called Mobility Places. We have two full-time employees and a handful of 1099 contractors. If we weren’t good at leveraging AI, we’d be at eight to ten full-time engineers right now. That’s not a guess. That’s me looking at the work we ship every week and counting the heads it would have taken to ship it three years ago.
Brooks’s math hasn’t changed. Communication overhead still grows quadratically with team size. What changed is per-person output. Work that took eight people now takes two. The optimal team size for a given output dropped by a factor of three to five, and most companies haven’t repriced their headcount.
Engineering is the cleanest case today, because the AI coding tools are evolving fastest, but the same compression is starting to hit sales ops, support, and finance.
There’s a catch. This only works if you have the right people.
I worked with a new graduate at a previous company who is the archetype I have in mind. Fantastic social and technical skills. Curious about everything. Not risk averse. He’d take on projects with literally no prior experience and just figure it out. He took ownership without being asked.
That’s a high-agency individual. In a fifty-person engineering org, this kind of person is a nice-to-have. They make the org better, but they’re a luxury. The org runs with or without them, because there’s always someone else to ask, another meeting to schedule, a senior to defer to.
In a two-person company, they’re table stakes. There is no one else to ask. If you can’t make the call yourself and execute on it, the work doesn’t get done.
The other thing AI changes is the role itself. Roles are blurring. A software engineer who’s also a product manager. A salesperson who’s also a marketing person. A finance person who is also a legal person. AI fills in whichever side of the role the person is weaker on. You don’t need to be deep in two domains. You need to be deep in one and willing to wade into the other with AI helping you.
If you’re a founder or a head of ops or an exec right now, the question worth asking has changed. It’s not “how do I make my team more efficient with AI.” That’s the small story again. The question is: what’s the smallest team that could run this company if I had the right people?
That number is a lot smaller than your current headcount. I’d bet on it.
If I went back to that VC-flush company today, I wouldn’t restart the hiring spree. I’d pause hiring and spend much more time evaluating the strengths and weaknesses of the people already on the team. Quality of engineer over quantity.
The companies that effectively leverage AI and maintain small teams are going to run circles around the big, bloated legacy ones.