The idea of a company that runs itself has been a staple of business writing for decades, usually filed under someday. It is worth looking at it again, because the someday has partly arrived. There are factories that produce through the night with nobody on the floor, and there are operations where software handles most of the customer contact. The fully humanless company is still rare, but the pieces of it are real and running today.

What near total automation looks like

The clearest examples come from manufacturing. FANUC has run lights out production in Japan since 2001, with robots building other robots at around fifty units per shift and some lines operating unmanned for up to thirty days at a stretch. The lights are off because no one needs to see. The Philips shaver plant in Drachten runs on roughly 128 robots with about nine people on quality assurance, which is close to humanless on the line itself.

These are not science projects. They are ordinary products, made at scale, with almost no one present during production. The work that remains is oversight and maintenance, not assembly. That pattern, machines doing the volume and a small crew watching over them, is what near total automation actually looks like in practice.

The software version

In digital businesses the same shape is forming without the robots. Klarna's AI assistant took on the equivalent of 700 full time agents within a month of launch, resolving issues in under two minutes and cutting repeat contacts by a quarter. Across the wider economy, Gartner expects around 15 percent of routine work decisions to be made autonomously by 2028, up from none in 2024. Step by step, functions that used to need a room full of people are being handled by systems.

Put the pieces together and you can sketch a company where production, support, and a good share of operations run with very few people in the loop. It is no longer a thought experiment. It is a set of tools that already exist, waiting to be assembled.

The ceiling is real

It would be a mistake to read this as a straight line to zero employees. The limits are showing up just as clearly as the progress. Klarna spent 2024 leaning hard into automation, then said in 2025 that it had gone too far in some areas and began reopening human support roles, noting that good human help is worth paying for. The market corrected the hype in real time.

The wider numbers carry the same warning. Gartner expects more than 40 percent of agentic AI projects to be cancelled by the end of 2027, often because the costs rose, the value was unclear, or the controls were missing. Automation hits a ceiling wherever work needs judgment, accountability, or trust. Those are exactly the things that are hardest to hand to a machine, and they do not disappear just because the rest of the work did.

The digital native version

Some of the clearest examples are companies that were built thin from the start. The messaging app WhatsApp served hundreds of millions of users with around 55 employees when Facebook bought it in 2014, and Instagram had just 13 people when it sold two years earlier. Neither was humanless, but both showed that a tiny team standing on top of well built systems could carry a load that once would have demanded thousands of staff. That model has only become easier to copy as agents and automation have matured.

The practical lesson is not to chase a headline employee count of zero. It is to ask, function by function, how much of this could run on its own, and then to build toward that honestly. Production might go almost all the way. Support might go most of the way until the hard cases. Strategy and relationships stay with people. The shape that results is lopsided, with machines carrying the volume and a small human core carrying the judgment, and that shape is already showing up across very different kinds of business.

How to think about it

The useful framing is not humanless versus human. It is figuring out how far down the automation path a given business can sensibly go, and what the remaining people should do once it gets there. For a factory that might be almost the whole line. For a law firm or a hospital it is a much smaller share. The right level is different for every company, and getting it wrong in either direction is costly.

So the humanless company is real, but it is also misnamed. Even the most automated operations keep a small group of people doing the work that machines cannot, which is watching, judging, and taking responsibility for the result. The future that is arriving is not one with no people in it. It is one where far fewer people carry far more, and where the human job is the part that was always the hardest to automate.