Floor cleaning is close to an ideal robotics task: repetitive, unpleasant, mostly nocturnal, on flat surfaces, with a clear definition of done. Autonomous scrubbers and sweepers have accordingly spread through shopping centres, airports, hospitals, logistics halls and large offices, usually without press releases, and often as the first robot a facilities team has ever operated. The labour they replace or extend is chronically short and turnover is high, so the business case is rarely the obstacle.
The obstacles are environmental in the literal sense. A cleaning robot works in a space designed entirely for humans and shared with them: cables left across corridors, chairs pushed out from desks, pallets parked overnight, wet floors from other trades, glass walls that confuse sensors, lifts it must summon and doors it must ask to be opened. Every one of these is a decision the machine either handles or escalates, and the difference between a deployment that sticks and one that quietly ends is how much a night supervisor has to babysit it. Vendors have learned to instrument this: robots report where they were blocked and why, and the useful conversation with the customer becomes about the building's habits rather than the machine's capabilities.
Integration has followed the same route as other building systems — lift and door interfaces, floor-plan management, and reporting that a facilities manager can show to a client to prove that a floor was cleaned when the contract said it would be. That last item turns out to be a selling point in its own right, since cleaning contracts have long been disputed on the basis of nobody being able to prove anything.
The lesson generalises beyond floors. Service robots that succeed indoors are the ones whose failure mode is a polite stop and a message, not an incident, and whose vendors treat the building's people as part of the system to be designed around rather than as an inconvenience to be trained out.