What the Evidence Actually Says
Part 3 of 4: Your Best People Aren't the Point
This is Part 3 of a 4-part series: Your Best People Aren't the Point. Catch up: [Part 1 — The Disappearing Team](https://theevolveproject.substack.com/p/the-disappearing-team) | [Part 2 — The Wrong Problem](https://theevolveproject.substack.com/p/the-wrong-problem)
If we're massively underweighting systems, what does the evidence look like when you actually examine it?
It turns out this question has been answered — repeatedly, across completely unrelated fields. And the answers converge on something most leaders would rather not hear.
Start with biology. Termite mounds are among the most sophisticated structures in nature — temperature-regulated, ventilated, architecturally precise. No termite understands the plan. There is no plan. Each insect responds to chemical traces left by others in the environment. The intelligence isn't in the termite. It's in the structure of interactions between termites. Ant colonies, immune systems, slime moulds — this pattern, intelligence embedded in the system rather than the agent, shows up across billions of years of natural selection. Not because evolution "arrived at an answer." Because organisms that distributed intelligence into their environment kept surviving, and organisms that didn't, didn't.
That's interesting. But it's biology. You could dismiss it as a metaphor.
Harder to dismiss is what happened in a 1993 economics experiment. Two economists wanted to understand what drives market efficiency. So they did something radical: they replaced human traders with "zero-intelligence" agents — algorithms that made completely random bids and offers. No strategy. No learning. No memory. No skill whatsoever. They expected human traders to dominate.
The result: when these random agents operated inside a specific market structure — buyers and sellers competing on both sides, nobody able to spend more than they had — they produced outcomes that were nearly as efficient as markets run by experienced human traders.
Read that again. Random agents. No intelligence. No training. No development programmes. Just the right structure. And that particular market worked.
The result is stunning — and it's specific. It's a finding about a particular kind of market with particular rules. It doesn't prove that any random agent in any well-designed system produces great outcomes. What it does show is that the right structure can do far more of the heavy lifting than we instinctively credit. If you're a founder obsessing over your next hire while ignoring the rules your team operates under, this should give you pause.
Behaviour is a function of the person and the environment. Kurt Lewin put it as an equation — B = f(P, E) — in the 1930s. Not just the person. Both variables. We've spent a century investing almost exclusively in P while barely touching E.
Picture someone living in Framingham, Massachusetts in the late 1980s. She's not thinking about her weight. She's thinking about her kids, her mortgage, the commute. Her closest friend — someone she has dinner with, someone whose house she knows the layout of — has slowly, quietly, put on weight over the past two years. Nothing dramatic. Just the way portions expanded, the way walks got shorter, the way second helpings became normal. She begins to follow the same gradient. Not because she decided to. Not because she was told to. Because the ambient signals around her shifted — what was offered, what was normal, what went unremarked. Now follow the chain one link further: her friend's weight change was itself shaped by someone she barely knows, a coworker's husband she met twice at parties. He changed. Her friend changed. She changed. She has no idea any of this happened. When Nicholas Christakis and James Fowler analysed three decades of data from the Framingham Heart Study, they found this transmission in the numbers. If your friend becomes obese, your own risk rises 45%. A friend of a friend — someone you may have never met — still raises your risk 10%. The effect runs to three degrees of separation, and it holds for smoking and happiness too. Christakis and Fowler argue the network structure itself is shaping outcomes — that the pattern is too consistent, too directional, and too persistent to be explained by people simply gravitating toward similar others. Others have pushed back, arguing that shared environment and selection effects account for much of what looks like transmission. The debate is real. But even the conservative reading leaves something significant on the table: your best people didn't develop their habits, their standards, or their ambitions in isolation. They were shaped, to a degree none of us has fully reckoned with, by connections they can't even name.
Three fields. Three independent lines of inquiry. One correction: we have been systematically underweighting the environment side of Lewin's equation. Not that individuals don't matter. They do. But the system they operate in matters far more than we've been willing to admit — and we've been pouring resources into the smaller variable while ignoring the larger one.