Wandelbots: Humanoids Are the Bridge to Intelligent Automation

Humanoid robots are currently attracting a lot of attention because they make it easy to envision the next phase of automation. However, their true significance does not lie in the fact that they look like humans. Rather, their significance lies in the fact that they represent a broader shift toward the adoption of intelligent automation: systems that can understand the physical world, make decisions, and act within it—with significantly greater adaptability than traditional automation solutions.

The humanoid design can be useful in certain environments, particularly where infrastructure and workflows are already tailored to human movements and human labor. But what truly sets them apart is not their physical form. The real advantage lies in the layer of intelligence that can be applied to many different types of robots, machines, and industrial processes in the future.

I’ve been in the manufacturing industry long enough to have witnessed several waves of technology that held enormous promise. When I first started working in the field of industrial automation, Industry 4.0 and digitalization dominated the discussion. Every factory was supposed to become smarter. Every machine was supposed to be connected. Every process was supposed to generate data. Dashboards, digital twins, predictive maintenance, cloud platforms, and connected production systems were presented as the building blocks of a smarter industrial future.

Much of this development was real. Manufacturing became more connected, more measurable, and, in many cases, more efficient. At the same time, the industry learned an important lesson: technology does not automatically solve the underlying complexity of production processes. If a process is poorly understood, managed inconsistently, or riddled with hidden edge cases, digitization does not magically make it simpler. Sometimes it merely makes the complexity more visible.

With AI, we now face a similar moment. The language has changed, but the ambitions seem familiar to me. Once again, we’re talking about flexible systems, adaptive automation, intelligent decision-making, and software that reduces the need to plan and develop every detail in advance. This time, however, the promise is not just that machines are connected to one another, but that they can understand, reason, and act.

This is a powerful idea, and I am convinced it has merit. Nevertheless, it also warrants a certain degree of caution.

In manufacturing, “flexibility” has always been one of the most appealing buzzwords in automation. It conveys the idea that a system can handle variations, accommodate uncertainties, and continue operating even when the real world doesn’t behave exactly as expected. The challenge is that flexibility can also become a way to postpone difficult process decisions. Instead of clearly defining all special cases, stabilizing workflows, or eliminating the root causes of variability, we sometimes hope that a more flexible layer of technology will absorb the chaos for us.

I’ve often seen a variation of this in industrial camera systems, even before the current AI era. Image processing systems were frequently viewed as a universal solution because they appeared flexible. A camera gave the impression that it could detect almost anything, especially in situations where it was difficult to define the right sensor. In practice, many of these applications could have been implemented with simpler, cheaper, and more reliable solutions: a simple sensor, a mechanical guide, a fixture, a photoelectric sensor, or a small process change. Nevertheless, customers were often willing to pay a premium for the expected flexibility, even if it ultimately created more complexity than benefit.

It is precisely this experience that shapes my view of humanoid robots today.

Humanoids are fascinating because they embody flexibility in its most visible form. They look as if they could enter a world created by humans and navigate its complexity. They suggest that in the future, we will no longer need to redesign processes for automation, but rather that we can integrate automation into existing processes.

This is precisely why the current interest in humanoids is significant. And this is precisely why we must be cautious when assessing what we are actually looking at.

Humanoid robots will not succeed simply because they are the best robots. In many manufacturing environments, they are not. Industrial robotic arms are faster, more precise, more robust, and easier to justify economically when a task is clearly defined. Autonomous mobile robots are usually better suited for material handling. Specially developed automation solutions remain the obvious choice when processes are repetitive, stable, and can be optimized in an economically viable way.

So when manufacturers consider humanoids today, the key question, in my view, is not whether they are better than industrial robots. In most cases, they solve a different problem. The real question is whether we are entering a new phase in the adoption of intelligent automation, in which adaptability becomes so valuable that it justifies an entirely new type of machine.

Why Humanoids Are Back in the Discussion

Humanoid robots have long been among the grand ambitions of robotics. For decades, however, they were mainly found in research labs, technology demonstrations, and science fiction. They were impressive to watch, but it was hard to imagine them as reliable industrial tools. The gap between a controlled demonstration and a real production environment was simply too wide.

That is now beginning to change.

Major automakers are now testing humanoids in real production environments. BMW has tested humanoid robots from Figure at its plant in Spartanburg, including for tasks related to handling sheet metal parts within the production process. Mercedes-Benz has also announced a partnership with Apptronik to explore the use of Apollo humanoid robots in manufacturing logistics. This includes, for example, supplying parts to production lines and inspecting components.

However, these examples should not be overinterpreted. They do not mean that humanoids are ready to replace existing industrial automation, nor that factories will be filled with versatile robotic workers in the near future. Rather, they demonstrate that manufacturers are beginning to explore a new category of automation. This category focuses less on optimizing a single task and more on adaptability across many different tasks—all within existing environments.

This distinction is crucial, because traditional automation works best when the environment is designed for the machine. Fixtures, conveyor systems, safety systems, robot cells, tools, and process flows are designed so that the robot can repeat a specific task with a high degree of reliability. This model is extremely powerful and will continue to play a central role in manufacturing in the future.

But it has its limits. Not every process is stable enough. Not every factory can be retrofitted. Not every use case justifies a dedicated automation project. Humanoids open up a different possibility. Instead of forcing manufacturers to adapt their environment to the robot, they point to robots that can work in an environment already designed for humans.

That is precisely where the opportunity lies.

The Inconvenient Truth: Humanoids Are Often the Worse Robots

The current hype surrounding humanoids can easily give the impression that the human body is the ideal design for manufacturing. It is not.

When it comes to welding a car body, grasping components from a known position, transporting material from A to B, or palletizing boxes at high speed, there are generally better machine designs. A humanoid brings with it a level of complexity that many industrial robots completely avoid. It must maintain balance, perceive its environment, navigate, manipulate objects, deal with uncertainties, and work safely around people. Each of these capabilities increases costs, risks, and engineering effort.

That’s why I don’t believe the future of humanoids should be viewed as a classic story of displacement. Industrial robots aren’t going away. Fixed-installation automation isn’t going away. Specialized machines aren’t going away. Wherever speed, precision, repeatability, and throughput are the decisive factors, specialized automation will continue to be superior.

The area where adaptability creates more value than maximum efficiency will be more interesting.

Many industrial environments are not perfectly optimized. Tasks change. Products vary. Workplaces continue to evolve. Labor availability fluctuates. Infrastructure ages. The surrounding process is often not stable enough to economically justify a dedicated automation solution, even if the actual task appears simple at first glance.

This is precisely where humanoids become interesting. They may not be the best machine for a single task. However, if they can be flexibly deployed for many different tasks, the economic calculation changes. This is particularly relevant for manufacturers exploring Robot-as-a-Service models. In this context, a robot’s value is measured not only by how efficiently it performs a single process, but also by how effectively it can be reused across different applications.

In this sense, the economic promise of humanoids does not lie in maximum performance.

It lies in strategic flexibility.

The factory was built for humans

Manufacturing environments are full of assumptions about humans.

Tools are positioned at human working height. Workstations are designed for human reach. Doors, handles, carts, containers, stairs, inspection points, and maintenance areas reflect the fact that humans have been the fundamental unit of physical labor for more than a century.

Many processes are still performed manually today, not because manufacturers prefer manual labor, but because automating them would require a comprehensive redesign of the environment. A permanently installed robotic cell might be able to handle the task, but only after changes to the layout, safety systems, tools, fixtures, software integration, and process workflows. For stable, high-volume production processes, this investment may be worthwhile. For variable work, it often is not.

This is one of the strongest practical arguments in favor of humanoids. They offer a potential path to automation without having to redesign everything related to automation.

In theory, a humanoid can move through the same areas as a human, use similar tools, interact with existing workflows, and take on tasks in areas where fixed automation would be too expensive, too disruptive, or too inflexible.

This does not make humanoids fundamentally better.

It makes them strategically different.

Their value lies in their compatibility with a world designed for humans.

The Generalist Trap

Much of the attention in humanoid robotics is currently focused on fully autonomous humanoids that can reason, understand the world, and perform tasks they have never encountered before. This is understandable. It’s a fascinating vision and fits perfectly with the current momentum surrounding large AI models and world models.

From a manufacturing perspective, however, there is a risk that something important is being overlooked.

For more than a hundred years, the manufacturing industry has done almost the opposite. It has channeled broad human capabilities into narrower roles, clearly defined responsibilities, repeatable processes, and specialized skills. The same applies to large parts of our education and training systems. We generally do not train people to be able to do everything. We train them to perform valuable work in specific contexts, with specific tools, under specific conditions, and with specific responsibilities.

This specialization is not a lack of imagination.

It is one of the reasons why modern manufacturing works.

An experienced machine operator, technician, welder, machinist, quality inspector, or maintenance engineer is not valuable because they can perform every conceivable task. Their value stems from their ability to reliably perform a relevant set of tasks within a specific environment. This value is based on technical expertise, contextual understanding, judgment, and experience.

This is precisely why the idea of making machines more versatile than humans should be viewed critically. Many of today’s automation systems are successful precisely because they focus on one or a few tasks. This focus is what makes them reliable. Now we expect new robotic systems to become general enough to handle unknown tasks, changing environments, unclear instructions, and unusual physical situations—situations that even humans often need training to manage.

Perhaps there will eventually be a future in which robots can reliably accomplish this.

For manufacturing, however, this is probably not the most practical starting point.

A more realistic path lies somewhere in between. Not a permanently installed machine that can perform only a single movement. But also not a fully autonomous humanoid designed to handle every unknown situation like a universal employee. The sensible middle ground is a machine that performs exceptionally well in a select group of valuable tasks while operating within clearly defined parameters—such as location, available resources, safety requirements, tools, process variations, and economic benefits.

This is exactly how the introduction of intelligent automation will likely unfold in practice.

Companies will not adopt humanoids simply because they can theoretically do everything. They will deploy them when they can reliably perform something valuable. Then, perhaps, something else. And subsequently, a growing number of related tasks within the same environment. Value is created through the gradual development of useful capabilities, not through the expectation of general intelligence from day one.

This applies to humans just as much as it does to machines.

A factory becomes more adaptable when humans learn several relevant skills, machines take on several valuable tasks, or both work together to contribute to a more resilient overall system.

Manufacturers should seek precisely this balanced middle ground.

Acceptance Might Be More Important Than We Think

There is yet another reason why humanoids receive so much attention. And this one is not purely technical in nature.

People react differently to robots that look and move like humans. We intuitively tend to attribute human characteristics to humanoid machines. We instinctively interpret where they are looking, what they might do next, and how they interact with their environment. As a result, they often seem more understandable than other forms of automation, even when the underlying technology is significantly more complex.

This is important because introducing new technologies into manufacturing is never just an engineering task.

It is always also a human and organizational challenge.

Operators, managers, safety officers, unions, and works councils play a decisive role in determining whether a new technology is accepted. A robot that appears relatable might encounter less resistance than a machine that seems alien or inscrutable, even if the other option is actually the more technically efficient solution.

At the same time, the humanoid form can give rise to new concerns. If a robot looks like an employee and performs similar tasks, people might perceive it as a more immediate threat to jobs than a traditional machine. The symbolism is stronger, and the emotional reaction may be as well.

That is why acceptance must not be viewed as a minor issue.

For humanoids to achieve commercial success, manufacturers must clearly communicate what these systems are designed for, what added value they create, how they are controlled, and how they integrate into the existing workforce.

Mechanical performance will be important. However, trust, perception, and communication could become just as important.

The Real Story Is Software

The most important question in humanoid robotics is not who builds the best mechanical body.

The most important question is who develops the best software layer for physical work.

Recent advances in AI have transformed what robots are fundamentally capable of. Large language models and vision-language-action models are beginning to integrate perception, instruction processing, reasoning, and action. Research systems like Google DeepMind’s RT-2 have demonstrated how vision-language learning at the scale of the internet can be combined with robot control. At the same time, projects like NVIDIA’s GR00T point to the development of foundation models for humanoid thinking and humanoid capabilities.

This is significant.

However, it should not be confused with the operational readiness of fully autonomous systems in manufacturing.

Factories are demanding environments. They require reliability, safety, repeatability, traceability, and integration with existing systems. A robot that successfully completes nine out of ten tasks may be impressive in a demonstration. In production, however, that would be unacceptable.

Industrial automation is not judged by whether it works once. It is judged by whether it works predictably, safely, and consistently.

Skills Over General Intelligence

This is exactly where the discussion gets interesting.

The future of Physical AI will likely not consist of a single gigantic model that understands everything. A more realistic approach is a system of specialized capabilities coordinated by a higher-level thinking and decision-making layer.

A humanoid does not need to understand the world in the same abstract way as a human. It must perform useful industrial tasks.

It must be able to load a machine, transport a container, inspect a component, open a door, operate a tool, prepare materials, assist an employee, or handle a minor process deviation.

Each of these capabilities can become a skill. The higher-level decision-making system then determines which skill to use, when to deploy it, and how to respond to changes in the situation.

For manufacturing, this is a significantly more practical vision than waiting for artificial general intelligence.

This is because industrial work rarely consists of a single complex task. It consists of many smaller abilities that must be performed reliably, safely, and repeatably.

This is precisely why the future of intelligent automation might resemble a platform rather than a single intelligent robot.

From Humanoid Robots to Intelligent Automation Platforms

At this point, the discussion extends beyond humanoids themselves.

Humanoids are one manifestation of Physical AI, but they are not the only one.

Physical AI refers to software that can understand the physical world, make decisions, and act within it. This intelligence can be delivered via a humanoid robot, an industrial robotic arm, an autonomous mobile robot, a mobile manipulator, a drone, or a machine that may not even exist today.

The long-term opportunity does not lie in building the perfect robotic body. The opportunity lies in creating platforms that enable intelligence to be transferred between different types of robots, environments, and tasks.

This is precisely why humanoids are so important. And this is precisely why they should not be confused with the actual goal. They are a visible entry point into a much larger transformation. They attract attention because they seem familiar, look spectacular, and are easy to understand. 

The deeper transformation, however, lies in software-defined automation.

That is, automation that can be configured, adapted, and improved through software, rather than having to be developed from scratch for every new use case. For manufacturers, this is the strategically crucial point. The future may not be determined by whether a robot has two arms, two legs, or a human-like face.

Rather, it could be determined by whether physical work can be programmed, adapted, deployed, and orchestrated with the same flexibility that we already expect from software today. 

That is why the adoption of intelligent automation is more important than the humanoid itself.

The lasting value will come from making physical work easier to deploy, easier to adapt, easier to control, and easier to scale across real-world industrial environments.

Why Europe Can Play an Important Role

This transformation is particularly relevant for Europe.

The DACH region is already one of the strongest markets for industrial automation worldwide. Germany remains one of the leading nations in robotics, with a very high robot density and a significant share of Europe’s installed industrial robot base.

At the same time, European manufacturers face a structural challenge in the labor market. The shortage of skilled workers is increasingly becoming a limiting factor for growth and competitiveness.

According to the 2025/2026 Skilled Labor Report by the German Chamber of Industry and Commerce, 36 percent of the companies surveyed were at least partially unable to fill open positions due to a lack of qualified workers.

This combination is creating strong pressure to automate. However, Europe will likely not adopt Physical AI in the same way as other regions. European manufacturers place great emphasis on quality, safety, reliability, data ownership, and governance. The EU AI Act establishes a risk-based framework for trustworthy AI. At the same time, the EU Data Act strengthens access to industrial data as well as the rules governing its use.

As a result, AI in manufacturing will increasingly need to be explainable, controllable, and compatible with European concepts of data sovereignty. This could prove to be a strength rather than a limitation. Europe does not have to win by developing the most spectacular humanoid demonstration. Europe can win by creating trustworthy, high-quality systems that orchestrate physical work in real industrial environments. 

The most important companies may not be the ones producing the most viral robot videos. They could be the ones that make intelligent automation safe, reliable, autonomous, and truly useful for everyday factory operations.

What Will Happen by 2030?

By 2030, the discussion about humanoids could look completely different.

Today, the focus is on design. It attracts attention because it is visible, familiar, and emotionally compelling. However, as the technology matures, manufacturers will pay less attention to whether a robot looks human and more to what tasks it can reliably perform.

The first wave of interest is driven by design.

The second wave is driven by concrete use cases.

The third wave is driven by orchestration.

In other words, by the question of how different robots, capabilities/skills, AI systems, and human employees work together within a production environment.

Some of today’s hype will fade. The limits of the technology will become more apparent. There will likely be disappointing pilot projects, exaggerated promises, and use cases where humanoids simply don’t make sense. 

However, that does not mean that humanoids will fail. It simply means that the market is becoming more realistic. Humanoids will not replace every employee. 

They will not replace industrial robots. And they won’t make every factory fully autonomous. Their real contribution might be something else:

They could accelerate the adoption of intelligent automation by making the idea of adaptable physical labor easier to understand, easier to test, and easier to implement.

Humanoids Are a Bridge

I don’t believe that humanoid robots are the goal.

I believe they are a bridge. A bridge between human-created environments and software-defined automation.

They are relevant because factories were built for humans, because humans can intuitively understand them, and because they create a visible and practical entry point into Physical AI.

The goal is not a robot that looks like us. The goal is intelligent automation that can be deployed safely, reliably, and flexibly in the physical world. By 2030, manufacturers may no longer be buying “humanoids” as we think of them today.

Instead, they will procure adaptable physical labor, provided by whichever machine makes the most sense for the specific task, environment, and business case.

– – – – –

Further Reading

👉 www.wandelbots.com  

Photo: Wandelbots