The Minimal Viable Organisation (a Cowboy Bebop lesson)

Spike Spiegel doesn’t run weekly standups. Jet Black doesn’t file progress reports. Faye Valentine doesn’t check the shared calendar before she moves. And yet the Bebop crew functions — rough, imperfect, often chaotic, but it functions. Four mismatched people on a ship in permanent financial freefall, capable of coordinating in action without ever formalizing their coordination.

This isn’t a narrative quirk. It’s a proof of concept.

What the enterprise replaced with process

Anthropologist Robin Dunbar established that the human brain can sustain genuine trust relationships with a limited number of people — roughly five for the strongest bonds, around fifteen for active cooperation. Beyond that, coordination demands infrastructure: rules, roles, tools. Process isn’t a managerial choice — it’s a cognitive prosthetic. Large organizations need it to exist at all.

The problem: the prosthetic became the product. Meetings to prepare for meetings. Reports nobody reads. Summaries summarizing summaries. A growing fraction of collective energy spent not on doing, but on coordinating to do.

The AI now embedding itself in organizations absorbs exactly that layer. Automated scheduling, meeting synthesis, generated reporting — these are coordination tasks, not production tasks. And when they disappear from the collective cognitive load, something is revealed in the negative space: the structures that had nothing else to offer.

Trust doesn’t scale — that’s its advantage

Mary Parker Follett, in the early twentieth century, was already saying what scientific management refused to hear: power is not a resource distributed top-down, it’s a capacity that emerges between people when they work together. She called it power-with, as opposed to power-over. Authentic coordination doesn’t come from authority or process — it comes from relationship.

The Bebop crew has no clear functional hierarchy. Jet is the de facto captain, but Spike only acknowledges that when it suits him. Faye acts for herself right up until she acts for the others — without saying so. Ed operates according to a logic nobody fully deciphers. What Follett would see in this crew isn’t disorder: it’s emergent coordination, built on genuine mutual knowledge rather than assigned roles.

That kind of trust doesn’t scale. That’s precisely its competitive edge.

A large structure can reproduce the form of a small team — squads, pods, autonomous cells. It rarely reproduces the substance. Because small-scale trust is built over time, through conflict resolved, through failure shared. It cannot be decreed.

What AI amplifies, and for whom

Those who know me know this isn’t a recent discovery. Long before AI entered the organizational conversation, I’ve held a simple conviction: the teams that actually work are small, stable, and built on a trust that makes process optional. That’s the core of how I think about organization, and it’s from there that I watch what technology changes — and what it doesn’t.

Coordination AI — the tools that plan, synthesize, prioritize, remind — doesn’t invert that conviction. It confirms it and gives it velocity. For a core team, it frees up time without removing what makes the team effective. It doesn’t replace the relationship: it strips away everything that was cluttering the relationship. What only people can do — read a situation, hold under pressure, decide together without manufactured consensus — remains intact. And there’s more room to do it.

For large organizations, the equation is different. Strip out the coordination layer without rebuilding the trust layer and you’ve removed the skeleton without developing the muscle. What remains is a structure with the appearance of agility and the inertia of an institution.

Four people is enough — or never enough

The Bebop doesn’t win, in the sense that an organization “wins.” The crew misses bounties, betrays each other, loses themselves, finds each other. What they never lose — even in the darkest episodes — is the capacity to coordinate when it actually matters, with zero infrastructure to support it.

The question AI poses to organizations isn’t “how do we automate coordination?” It’s more radical than that: when coordination is automated, what holds your teams together?

If the answer is “process,” you have a problem technology won’t solve.