Zero-Human Companies
Most companies still hire someone the second a problem shows up twice. That used to be the only move. It isn’t anymore, and the gap between the two models is about to get embarrassing.
5dive’s bet: the companies that win the next decade won’t have more employees. They’ll have better agents. We’re building the platform to run them.
the old loop isn’t lazy, it’s structural
Hiring isn’t a failure of imagination. It’s the cheapest way to handle work you can’t fully spec in advance.
So companies run the loop. Notice recurring work. Hire someone to absorb it. Backfill when they leave. Every iteration eats months and adds variance.
Two people in the same role do the work differently.
The same person does it differently in week one and week fifty.
Knowledge walks out with attrition. Process docs drift from reality the day they’re written.
None of this is anyone’s fault. It’s what happens when the unit of execution is a human and the spec lives in a doc instead of code.
What dissolves the loop is an executor that takes an underspecified job, fills the gaps with judgment, and comes back tomorrow having gotten better at it.
That executor is now real.
humans decide, agents execute
“Zero-human” is a deliberately uncomfortable phrase. It’s not a prediction that humans disappear from companies. It’s a description of which roles get reshaped first.
The roles that survive intact are the ones where the value is the decision. What to build. Who to serve. How to price. When to pivot. What to refuse. Decisions need taste, accountability, and a body to put behind them.
The roles that get absorbed are the ones where the value is the execution. The recurring tickets. The report nobody reads carefully. The onboarding email written from a template. The moderation queue. The daily reconciliation. The support reply that just needed somebody to actually look at the logs.
Most companies have a much higher ratio of execution to decision than they admit. Shift it and the same headcount produces several times more output. Refuse to shift it and you keep paying for variance and management overhead you didn’t have to.
The clearest proof of this in our own product is support. We don’t run a help desk.
A Claude agent lives on every customer’s server. It’s the first responder, with the logs, files, and deploy history a ticket queue would have to ask for. We wrote about that in There Is No Support Email.
The second-order effect is the part that matters: every gap that would have become a ticket becomes a better default instead. The company shape compounds.
why now: the bottleneck moved
For years the bottleneck on agentic work was AI quality. Models could write code, just not reliably enough to ship. Draft email, not reliably enough to send. Read logs, not reliably enough to act on what they found.
That bottleneck moved.
Claude, Codex, and Gemini are now capable enough that the limiting factor isn’t the model. It’s what the model is allowed to touch.
A model in a chat tab can’t watch a deploy overnight. A model in a sandbox can’t reach the database that needs the migration. A model that loses its filesystem when the laptop sleeps can’t pick up where it left off.
The missing layer is infrastructure. Every agent needs its own machine, its own identity, its own credentials, its own way to pick up tomorrow what it dropped yesterday. We made that case in Your Agent Should Have Its Own Server.
Some teams will also want a shape on top — roles, budgets, approval points, an audit trail. That’s the governance layer above the runtime, optional for most.
Once the runtime is in place, the zero-human company stops being a thought experiment.
the rule
If a piece of work has to be done more than once, don’t write a job description for it. Define it once, give it to an agent with the tools to do it, put a human in front of the decision points.
That’s what a zero-human company looks like in practice. It’s not a future-tense claim. It’s an organizational shape that’s now buildable, and the companies that build it first will look strange to the ones that didn’t.
5dive is the platform for the agents that run inside that shape. The bet is there will be a lot of them. And the companies running them won’t need to grow headcount to grow.