The conversation around robotics has shifted. For years, the industry measured progress in terms of speed, payload, and precision. Those metrics still matter, but they are no longer sufficient on their own. As robots move out of locked cages and into shared spaces with people—factory floors, warehouses, and service environments—the criteria for what makes a good robot have expanded. Safety is no longer a compliance checkbox; it is a design philosophy, a competitive differentiator, and a prerequisite for deployment.
The source material points to several concrete developments that illustrate this shift. Take the Delta D-Bot series of collaborative robots. These units received the BIG SEE Product Design Award 2026, an honour that recognises not just aesthetics but also the integration of safety features into the product’s core design. Delta Electronics, a company known primarily for power management and smart green solutions, has positioned these cobots as examples of how safety can be embedded from the ground up rather than bolted on as an afterthought. The award itself is a signal: design juries now consider safety to be a legitimate and important aspect of product excellence.
Amazon’s fulfilment centres offer another lens. The company has been deploying AI and robotics systems such as Blue Jay and Project Eluna across its US operations. Blue Jay is tasked with handling repetitive tasks, which reduces the physical strain on human workers. Project Eluna provides operational insights, giving employees data-driven visibility into workflows. Both systems are described in the source material as tools that build safer and more efficient workflows for front-line employees. The key takeaway here is not just that Amazon is using robots—it is that the stated purpose of these systems is to improve safety and ergonomics, not merely to cut costs or increase throughput. The technology is being framed as a partner to the workforce, not a replacement.
The broader industry trend, as outlined in the source, is that companies and researchers are moving beyond prototypes and beginning to deploy humanoids in real-world settings. This is a significant milestone, but it brings with it a new set of challenges. Humanoid robots, if they are to compete with traditional automation, must match high industrial requirements in terms of cycle times, energy consumption, and maintenance costs. But beyond performance, they must also meet industry standards that define safety levels, durability criteria, and consistent performance on the factory floor. The source material is explicit on this point: as robots increasingly operate alongside humans, ensuring they operate safely is not just important—it is essential for the robotics industry.
What makes this particularly complex is the role of AI-driven autonomy. The source notes that AI fundamentally changes the safety landscape. Unlike a traditional industrial robot that follows a fixed path, an AI-driven robot can make decisions in real time. This makes testing, validation, and human oversight much more complex—but also more necessary. The intended use of humanoid robots makes this especially clear. These machines are designed to operate in human environments, which means they must be certified in line with ISO safety standards. The source does not specify which exact ISO standards apply, and we should not invent those details. What is clear is that compliance with established safety frameworks is a non-negotiable requirement for deployment.
There is also a broader ecosystem angle. NVIDIA is working with industrial software giants and robotics leaders such as ABB, Universal Robots, and KUKA. The goal is to integrate physical AI models and simulation tools into manufacturing lines. This collaboration enables the deployment of smarter robots, but it also raises questions about how safety is validated in a simulated environment versus the real world. Telecom providers like T-Mobile are also involved, as base stations evolve into edge AI platforms. This suggests that the infrastructure supporting robotics is becoming more distributed, which has implications for how safety monitoring and human oversight are implemented.
So, what should a robotics buyer, integrator, or facility manager look for when evaluating new systems? The source material suggests several criteria. First, look for evidence that safety is part of the design DNA, not an add-on. The Delta D-Bot award is one example of third-party recognition for this approach. Second, look for systems that are explicitly designed to work with humans, not just near them. Amazon’s Blue Jay and Project Eluna are examples of systems that augment human capabilities rather than replace them. Third, look for compliance with ISO safety standards. The source is clear that robotic systems need to be designed and certified in line with these standards. Fourth, consider the role of AI and autonomy. If a robot has decision-making capabilities, ask how those decisions are tested, validated, and overseen by humans. Finally, consider the ecosystem. Are the robot’s software and simulation tools integrated with established players? The NVIDIA collaborations with ABB, Universal Robots, and KUKA suggest that a connected approach is becoming the norm.
Practical steps
If you are responsible for deploying or managing robotic systems, the source material offers several actionable directions. These are not hypothetical recommendations; they are grounded in the developments described in the source.
First, prioritise safety as a design criterion from the outset. The Delta D-Bot series won a design award for its safety features, which indicates that safety can be a source of competitive advantage, not just a regulatory burden. When evaluating robots, ask vendors how safety is integrated into the hardware and software. Is it a matter of adding sensors and emergency stops, or is it embedded in the control architecture? The source suggests that modern collaborative robots are being recognised for holistic safety design, so your procurement criteria should reflect that.
Second, focus on human-robot collaboration models. Amazon’s approach with Blue Jay and Project Eluna is instructive. Blue Jay handles repetitive tasks, which frees human workers from monotonous and physically demanding work. Project Eluna provides operational insights, which helps employees make better decisions. The lesson here is that robots should be deployed to handle specific, well-defined tasks that are either dangerous, repetitive, or both. This not only improves safety but also enhances overall efficiency. When planning your deployment, map out which tasks are best suited for automation and which are best left to humans. The goal is not to replace people but to create workflows where humans and robots complement each other.
Third, invest in testing and validation. The source material is emphatic that AI-driven autonomy makes testing and validation more complex. If your robot has any degree of autonomous decision-making, you need a robust testing protocol. This should include simulation, as suggested by the NVIDIA collaborations with ABB, Universal Robots, and KUKA. Simulation allows you to test edge cases that would be dangerous or expensive to replicate in the real world. However, simulation is not a substitute for real-world testing. You need both. The source does not specify exact testing protocols, and we should not invent them, but the principle is clear: rigorous testing is a prerequisite for safe deployment.
Fourth, ensure human oversight is built into the workflow. The source states that human oversight is more necessary than ever when AI is involved. This means that even if a robot can operate autonomously, there should always be a human in the loop for critical decisions. This could be a supervisor who can intervene in real time, or it could be a review process that monitors robot performance and intervenes when anomalies are detected. The exact nature of the oversight will depend on your specific use case, but the principle is universal: autonomy does not mean abdication of responsibility.
Fifth, align with ISO safety standards. The source is clear that robotic systems need to be designed and certified in line with ISO safety standards. This is not optional. When you are selecting a robot, ask for documentation that demonstrates compliance. If a vendor cannot provide this, that is a red flag. The source does not specify which ISO standards apply to which robot types, and we should not guess. What we can say is that certification is a requirement, and you should verify it before deployment.
Sixth, consider the broader ecosystem. The source highlights collaborations between NVIDIA and major robotics players, as well as telecom providers like T-Mobile. This suggests that the robotics landscape is becoming more interconnected. When you choose a robot, consider how it will integrate with your existing infrastructure. Will it work with your current software systems? Can it be updated as new AI models become available? The source does not provide specific integration details, but the trend is clear: standalone robots are becoming less common, and connected systems are becoming the norm.
Seventh, plan for skills development. The source mentions changing skills demand and competing in an automation-driven economy. This implies that your workforce will need new skills to operate and maintain advanced robotic systems. This is not just about technical training; it is also about changing the culture of the workplace so that employees see robots as tools that help them work smarter, not as threats to their jobs. Amazon’s framing of Blue Jay and Project Eluna as tools that empower employees is a good model to follow.
Finally, monitor the landscape for new developments. The source material is dated 2025-06, and it references trends for 2026. The field is moving quickly. What is considered best practice today may be outdated in a year. Stay informed about new standards, new technologies, and new collaborations. The source mentions that companies and researchers are moving beyond prototypes to deploy humanoids in real life. This is a trend to watch, but it also comes with new safety challenges that will need to be addressed.
Common mistakes to avoid
The source material provides enough detail to identify several common pitfalls in the robotics industry. Avoiding these mistakes can save you time, money, and potentially prevent accidents.
One mistake is treating safety as an afterthought. The source material highlights that safety is essential for the robotics industry, and the Delta D-Bot award demonstrates that safety can be a design feature. If you wait until after a robot is deployed to consider safety, you are already behind. Safety needs to be part of the initial design and procurement process. This means asking tough questions early: What happens if the robot malfunctions? How does it behave in unexpected situations? What are the failure modes? If a vendor cannot answer these questions, move on.
Another mistake is assuming that human oversight is optional once a robot is certified. The source is clear that AI-driven autonomy makes human oversight more complex and more necessary. Certification is a baseline, not a guarantee. Even if a robot meets ISO standards, you still need to monitor its performance in your specific environment. The source does not specify how often oversight should occur, and we should not invent a schedule, but the principle is clear: ongoing oversight is required.
A third mistake is ignoring the human element. Amazon’s approach with Blue Jay and Project Eluna is designed to empower employees, not replace them. If you deploy robots in a way that alienates your workforce, you will face resistance and potentially failure. The source mentions changing skills demand, which suggests that training and communication are critical. Do not assume that employees will automatically accept robots. You need to explain the benefits, provide training, and address concerns.
A fourth mistake is focusing solely on performance metrics like cycle time and energy consumption. The source notes that humanoids need to match high industrial requirements, but it also emphasises safety and durability. If you optimise only for speed and cost, you may end up with a system that is unsafe or unreliable. The source does not provide specific numbers, and we should not invent them, but the trade-off between performance and safety is a real one that needs to be managed carefully.
A fifth mistake is neglecting the ecosystem. The source highlights collaborations between NVIDIA, ABB, Universal Robots, KUKA, and T-Mobile. This suggests that robots are becoming part of a larger network. If you choose a robot that is isolated from this ecosystem, you may miss out on updates, integrations, and best practices. The source does not specify which ecosystem is best, but the trend towards connectivity is clear.
A sixth mistake is underestimating the complexity of AI-driven autonomy. The source states that AI fundamentally changes the safety landscape. A robot that can make decisions is fundamentally different from one that follows a fixed path. If you treat an AI-driven robot like a traditional one, you will be unprepared for its behaviour. This means investing in more sophisticated testing, validation, and oversight. The source does not provide a specific framework for this, and we should not invent one, but the need for a different approach is evident.
A seventh mistake is ignoring ISO standards. The source is explicit that robotic systems need to be designed and certified in line with ISO safety standards. This is not a suggestion; it is a requirement. If you deploy a robot that is not certified, you are exposing yourself to legal and operational risks. The source does not list which ISO standards apply, and we should not guess, but the requirement for certification is unambiguous.
An eighth mistake is failing to plan for maintenance and lifecycle costs. The source mentions that humanoids need to match high industrial requirements towards maintenance costs. This suggests that maintenance is a significant consideration. If you do not plan for ongoing maintenance, you may face unexpected downtime and expenses. The source does not provide specific maintenance schedules or costs, and we should not invent them, but the need for a maintenance plan is clear.
Finally, a ninth mistake is assuming that what works in one environment will work in another. The source describes Amazon’s fulfilment centres, Delta’s collaborative robots, and NVIDIA’s industrial collaborations. These are different contexts with different requirements. What works for Amazon may not work for a small manufacturer. The source does not provide guidance on how to adapt these examples to different settings, and we should not invent it, but the need for context-specific evaluation is evident.
In summary, the path to safe and effective robotics deployment is not straightforward. It requires attention to design, compliance, oversight, and human factors. The source material provides a snapshot of where the industry is heading, but it is up to each organisation to apply these lessons to their specific circumstances. The one thing that is clear is that safety is no longer optional. It is the foundation upon which the future of robotics will be built.
Sources
Robots are getting smarter and more sophisticated. They also need to get safer
Published by Vigla Media OÜ (Estonia).