Jake Van Clief (@lostandlucky) posted a reel this week about The Mythical Man-Month. Worth a watch. Here’s what I’d build on top of it — because there’s a piece that usually gets skipped.
What Brooks actually said
Frederick Brooks published The Mythical Man-Month in 1975 (the research goes back to the early ’70s). The headline finding: adding more people to a late software project makes it later. Not because people are bad at their jobs. Because coordination overhead grows faster than output. Every new team member doesn’t just add capacity — they add communication paths. The work multiplies. The throughput doesn’t keep up.
Companies are running straight into this problem with AI agents right now. Stand up a swarm — one agent to research, one to write, one to review, one to route — without a clear structure underneath it, and you get faster confusion. Not faster results.
The piece that gets skipped
Brooks wasn’t just writing about headcount. He was writing about structure. The fix for a late, overcrowded project wasn’t fewer people — it was clearer roles, cleaner handoffs, and someone who actually owned the architecture of the work itself.
That’s where I see teams get stuck with AI. They buy the tooling, spin up the agents, and then hit a wall — because nobody built the structure the agents are supposed to operate inside. The agents are capable. The map doesn’t exist yet.
No amount of prompting fixes a missing map.
What getting it right early actually looks like
Before you scale, answer these: What is each agent responsible for? What’s the source of truth when two agents surface conflicting information? Where does a human need to be in the loop, and why? How does work get handed off cleanly so nothing falls between the cracks?
These aren’t AI questions. They’re operations questions. They’re the same questions good information architects and workflow designers have been asking for decades. The tooling is new. The discipline isn’t.
Getting these sorted before you build saves you from rebuilding later — and rebuilding later, when agents are already embedded in real processes, is expensive and slow. Brooks would recognize the pattern immediately.
If you’re in this right now
I’ve been doing structural work — information architecture, workflow design, content operations — for a long time. Some of it was called web development. Some of it was called marketing operations. What you’re calling AI implementation today involves the same underlying questions.
If your team is figuring out where the agents go and the structure isn’t solid yet, I can help you work through it. Reach out — I’d rather help you get ahead of the problem than pull you out from under it later.
JCody Ventures — video, drone, networks, marketing, and the structure underneath all of it. Get in touch.
Written with AI assistance. Every idea, opinion and experience here is mine. I just thrive on strategy and execution more than eloquent wordsmithing.