The price of a capable humanoid robot has fallen faster than almost anyone predicted, and the demos are genuinely impressive. Boards are starting to ask the obvious question: should we get one?
It's the wrong first question. A humanoid is not a gadget you buy; it's a capability you operate. The organisations that will get value from embodied AI are the ones that treated it like hiring a new kind of worker — with a role, a manager, guardrails and a plan — rather than like buying a very expensive appliance.
Before anyone signs a purchase order, six questions separate a productive deployment from a very costly ornament.
What job, exactly? "A humanoid" is not a use case. "Moving totes between packing stations on the night shift" is. The narrower and more repetitive the task, the better the odds. Robots earn their keep on the dull, physical, high-frequency work humans dislike — not on the flexible, judgement-heavy work that looks easy to us and is fiendishly hard for a machine.
Why a humanoid, and not something simpler? Human form is expensive. It's worth paying for only when the environment is built for humans and can't easily be changed — stairs, existing workstations, tools designed for hands. If a fixed conveyor or a wheeled unit would do the job, that's usually the smarter buy. Choosing the human shape should be a deliberate decision, not a default driven by how the demo looked.
Who owns it after go-live? This is where most robotics investments quietly fail. The technology arrives, everyone is excited, and then no one is accountable for keeping it running, retraining it when the task changes, or handling the days it doesn't behave. A humanoid without an owner becomes a very photogenic paperweight within a quarter.
What happens when it's wrong? Physical automation has physical failure modes. Before deployment you need clear answers on safety around people, what the robot does when it's uncertain, how a human intervenes, and who is responsible when something goes wrong. This isn't box-ticking — it's the difference between a capability you can scale and a liability you have to hide.
How does it fit the work around it? A robot is one node in a process. If the steps upstream and downstream aren't ready to hand off to it and receive from it cleanly, the robot just becomes a bottleneck with better marketing. The integration — not the hardware — is where the value is made or lost.
How will you know it worked? Pick the number before you buy: throughput, cost per unit, hours returned to people, error rate. Decide what success looks like in advance, or you'll end up justifying the purchase after the fact with a story instead of a result.
None of this is a reason to wait on the sidelines — embodied AI is going to reshape whole categories of physical work, and the organisations that learn to operate it early will build an advantage that's hard to copy. But learning to operate it is exactly the point. The winners won't be whoever buys the shiniest robot first; they'll be whoever built the readiness — the role, the ownership, the guardrails, the integration — to make it actually work.
That readiness is precisely what we assess and build. Whether it's a digital e-Employee or one embodied on a physical humanoid, the questions above are the ones we help organisations answer before the money is spent — so the robot that arrives has a job, a manager, and a number to hit.