The case for the human shape
Buildings, tools, shelves and workstations are designed for people. A robot with two arms, hands and legs could in principle work in them without the site being rebuilt, and move from one task to another with new software instead of new hardware. Advances in AI models that learn from demonstration have made that idea much more credible than it was a few years ago.
Where they are today
Most real deployments are pilots with large manufacturers and logistics companies. The tasks are narrow and repetitive: moving totes from one place to another, loading and unloading containers, handing parts to a line. Several vendors now offer robots on subscription for this kind of work, often with the vendor's engineers on site.
Common limits in 2026:
- Speed. Many humanoids complete tasks more slowly than a trained worker.
- Runtime. Battery life limits shifts; swapping or charging has to be planned.
- Reliability. Hands, balance and perception still fail in ways that need a person to step in.
- Remote operation. Some demonstrations and some working deployments rely on a person controlling or supervising the robot remotely. Ask how much.
- Payload. Most carry modest loads, well below what a forklift, a cart or a purpose-built arm handles.
Safety and standards
A robot that balances on two legs can fall, and it can lose balance when it loses power. Wheeled and fixed robots do not have this hazard, and the standards are still catching up. ISO 10218 covers industrial robots and their cells, and ANSI/A3 R15.08 covers industrial mobile robots, but neither was written with legged humanoids working beside people in mind. Standards work for this class of robot is under way.
Until it matures, most pilots keep humanoids in separated or controlled areas, with a documented risk assessment and a clear plan for stopping the robot. OSHA's General Duty Clause and lockout/tagout rule apply here as they do to any machine. Read robot safety before any pilot.
How to run a sensible pilot
- Pick one task with a clear measure: totes moved per hour, trays loaded, errors.
- Measure the baseline with people or existing equipment before the robot arrives.
- Keep it contained. A fixed area, a fixed time window and trained staff nearby.
- Track interventions. Log every time a person had to help, remotely or in person.
- Set a decision date. At the end, compare cost per unit of work against the baseline and against a purpose-built alternative.
Questions to ask a vendor
| Topic | Ask | Why |
|---|---|---|
| Task | Which exact task will it do, and how many times a day? | A demo of a different task tells you little |
| Rate | Units per hour against a person doing the same task | Early humanoids are often slower than people |
| Autonomy | Share of the time it runs without a remote operator | Teleoperation hides in many demos |
| Uptime | Hours between interventions, runtime per charge, charge time | A robot that needs rescuing every hour costs a person |
| Safety | Standard evaluated to, and behavior when it falls or loses power | Balance loss is a hazard wheeled robots do not have |
| Exit | Cost and notice to end the pilot | Keep the downside small |
Costs people miss
Pilot fees, vendor engineering time, site changes for charging and safety zones, staff time spent supervising, insurance, and the opportunity cost of a project team. Most humanoids are offered on subscription; see robots as a service for the contract terms to check, especially exit terms and what happens to your data.
If you are a small business, the useful move in 2026 is to watch the pilots of larger companies, and to put robots to work now where purpose-built ones already pay back: cleaning, warehouse transport and picking and delivery and serving. The service robot directory and the cobot directory list what is on the market.
Before you sign up
- Name one task with a measurable output, and a baseline.
- Compare with a purpose-built robot or conveyor for the same task.
- Ask how much of the work is done under remote operation.
- Get the safety standard and the risk assessment in writing.
- Keep the pilot short, contained and cheap to exit.
- Set a decision date and the numbers that would justify going further.