No Scale Without Trust

I have spent a large part of my professional life in marketing, logistics and innovation. That background strongly shapes how I look at humanoid robotics today.

In logistics, you learn very quickly that a technology is not successful because it works once in a demonstration. It is successful when it works every day, under pressure, during peak volumes, with different people and in different environments. Marketing taught me something similar from another perspective: credibility is earned through experience, not through promises.

That is why I believe the next big challenge in humanoid robotics is no longer simply technological capability. It is reliable deployment.

For the last few years, the industry has focused heavily on what humanoids can do: walking, manipulating objects, navigating complex environments and learning new tasks. The progress is impressive. But as we move from demonstrations and controlled pilots into warehouses, factories, hospitals and public environments, the questions become much more practical. What happens when a sensor fails? What happens after a software update? What happens if connectivity is interrupted? Who is responsible if the behaviour of the robot changes? And what happens when the robot encounters something nobody anticipated during testing?

Anyone who has worked in logistics knows the difference between a pilot and real operations. A pilot can be protected. You can control the environment, prepare for exceptions and intervene when something goes wrong. Daily operations are different. A warehouse does not stop because a robot has a bad day. A customer still expects the parcel to arrive. Employees need to know that the technology around them is predictable and safe.

This is why safety and governance are part of the business case, not simply compliance topics. A company will not deploy hundreds or thousands of humanoids because the technology is fascinating. Management needs confidence that the systems are reliable. Employees need to know that they can work safely alongside them. Insurers need to understand the risks. Regulators need evidence that risks are being managed responsibly.

You can scale production without trust. But you cannot scale deployment without trust.

My background in marketing also makes me cautious about how easily trust is claimed. If we tell people that humanoids are safe, we need evidence. If we say a system is reliable, we need to understand how that reliability was tested. If we say an AI-enabled robot is under control, we need to know what happens when its software or behaviour changes.

Trust has to be engineered before it can be communicated.

One of the biggest challenges is that an AI-enabled humanoid is no longer a fixed product. The hardware may remain unchanged while software, perception models and AI capabilities continue to evolve. The robot standing in a warehouse may look identical after an update, but its behaviour may have changed considerably.

That raises an important question: when does an update become significant enough that the system needs to be reassessed?

We cannot certify a robot once and simply assume that we are dealing with exactly the same machine five years later. We need to think increasingly about continuous assurance throughout the lifecycle.

Safety is not a component. Safety is a lifecycle.

Cybersecurity is part of this discussion as well. In traditional IT, a security incident may result in lost data or disrupted services. With a humanoid robot, a cyber incident can have physical consequences. These machines move, interact with people, carry loads and receive software or AI updates. The traditional separation between IT security and operational safety therefore becomes increasingly difficult to maintain.

The same applies to responsibility. If something goes wrong, is the manufacturer responsible? The AI developer? The software provider? The integrator? The operator? In reality, responsibility will often be distributed across the ecosystem.

From logistics, we already know how important clear accountability is across a chain. A parcel may move through many systems and organisations, but the customer expects one thing: that it arrives. Humanoid robotics will face a similar challenge. The person working next to the robot does not care which supplier wrote a particular piece of software. They expect the complete system to be safe.

This becomes even more relevant because humanoid robotics is global from the beginning. A robot might be manufactured in China, contain components from Japan or Europe, use AI developed in the United States, be integrated in Germany and finally operate in Switzerland. From a logistics perspective, that is nothing unusual. Global supply chains have worked this way for decades.

But it means that safety cannot stop at national borders either.

Europe, China and the United States will continue to develop different regulatory and standardisation approaches. I do not believe they need to become identical. What matters is that they are compatible enough to create a common understanding of safety, testing, evidence and responsibility.

We do not need identical regulations globally. We need interoperable safety principles.

At the same time, standards should not become a brake on innovation. Humanoid robotics is moving too quickly for regulation to prescribe every technical solution. Engineers need room to develop better systems. But there also has to be a clear baseline in areas where failure can have serious consequences.

My position is pragmatic: outcome-based wherever possible, prescriptive where failure is unacceptable.

Innovation should happen above the safety floor, not instead of the safety floor.

When people talk about scaling humanoid robotics, the conversation often focuses on manufacturing capacity: who can build 10,000 or even 100,000 robots? That is an important question, but from an operator's perspective another question matters just as much: how do we safely deploy those robots in thousands of different real-world environments?

Every site is different. Every process is different. Every workforce is different. Every integration is different. That is where the real work of scaling begins.

This is also where I see the role of the Humanoid Robotics Association Switzerland. Our goal should not be to slow down innovation, but to bring manufacturers, operators, integrators, researchers and safety specialists together early enough that deployment requirements are addressed from the beginning rather than added afterwards.

I am very optimistic about humanoid robotics. Coming from logistics, I can immediately see the potential in repetitive tasks, difficult environments and labour-intensive processes. But years of working with customers, operations and new services have also taught me that technical potential alone does not create adoption.

Companies scale solutions they can rely on and control. Markets reward technologies that consistently deliver what they promise.

The humanoid that ultimately succeeds may therefore not be the robot that can do the most impressive things. It may be the one companies are comfortable putting into daily operation.

No Scale Without Trust.

Watch the interview

In this interview ahead of my keynote in Munich, I discuss why safety and governance are becoming critical for humanoid robotics, how different markets are approaching regulation and standardisation, and why I believe the industry needs to move from one-time certification towards continuous assurance.