There is an unusual buzz around robotics. At trade shows Machine Design visited this year, the push to bring artificial intelligence into the physical world was unmistakable, with robotics emerging as one of its clearest expressions.

The momentum extends well beyond the exhibition halls. Robots are already deeply embedded in global manufacturing. The International Federation of Robotics reports that more than 4.28 million industrial robots were operating in factories worldwide in 2023, with annual installations exceeding 500,000 units for the third consecutive year. Meanwhile, growing investment in AI infrastructure and platforms is pushing technology companies to view robotics as the next interface between digital intelligence and the physical world.

To analyze what these signals mean for machine builders and automation engineers, Machine Design invited Jenny Shern, General Manager at NexCOBOT, to tackle a broad range of questions on the convergence of AI, robotics and industrial automation. Drawing on her experience in motion control, embedded computing and robotics platforms, Shern thoughtfully addresses topics spanning the influence of major technology companies on robotics ecosystems, AI-native development, interoperability, motion control, safety, engineering talent and real-world ROI.

READ MORE: Automation as Strategy: A3’s Jeff Burnstein on the State of Robotics Adoption, AI and Scaling

As Shern explains, scaling AI-powered robotics will require more than increasingly capable algorithms. The industry must bridge the gap between AI’s ability to perceive and reason and the industrial disciplines that have defined automation for decades. Those fundamentals (precise motion control, functional safety, open architectures and reliable performance at scale) remain the benchmarks by which AI-enabled systems will ultimately be judged.

The full transcript of our correspondence follows. It has been lightly edited.

Machine Design: What indicators convince you that Big Tech’s ambition is extending to robotics? What makes robotics a natural extension of Big Tech’s AI ambitions?

Jenny Shern: Recent acquisitions by companies like Meta and Amazon signal that Big Tech sees robotics as the next major platform for AI. These companies are not necessarily trying to build complete robots themselves, though. Instead, they appear to be building robotics “operating systems” with intelligence layers, software stacks and application ecosystems that robot manufacturers and machine builders can use.

Robotics is a natural extension of Big Tech’s AI ambitions because robots create a bridge between AI models and the physical world. Big Tech already has AI models, data infrastructure and deployment environments. What robotics adds is a way to apply those capabilities in factories, warehouses and service environments while generating even more real-world data for training, simulation and improvement.

MD: What does this influx of Big Tech capital and platforms mean for industrial adoption?

JS: Investments will accelerate the movement from traditional automation toward AI-enabled robotics. For industrial users, the biggest impact will be greater access to platforms, tools and ecosystems that make robotics easier to develop and deploy. However, industrial environments still require real-time control, functional safety, precision and reliability. This creates an opportunity for companies with deep expertise in motion control and safety-certified platforms to become essential ecosystem partners as robotics adoption scales.

MD: Manufacturers ultimately care about reliability, safety and ROI. What needs to happen before AI-powered robots move from compelling demonstrations to trusted industrial systems operating at scale?

JS: Before AI-powered robots can transition from factory floor demonstrations to trusted industrial systems operating at scale, they must evolve past flashy algorithms to meet strict production requirements like uptime and repeatability. This transition hinges on combining high-level AI perception with real-time motion control, safety-certified hardware and software designed for seamless human-robot collaboration.

READ MORE: Physical AI Hype vs Reality: Kung Fu Robots are Cool...But Should You Hire One?

Commercially, the financial equation must make sense over a 10-year lifecycle, meaning the amortized maintenance costs of these systems must fall significantly below human labor costs. Until companies see a clear return on investment (ROI), adoption will lag. Therefore, it is crucial to first anchor these technologies in high-volume, highly replicable application scenarios, such as automated warehousing, hotel housekeeping or robotic yard maintenance, where cost-efficiency is immediate.

Ultimately, scaling this technology requires bridging a critical ecosystem gap: Big Tech’s tendency to cling to proprietary, closed-loop systems clashes directly with the highly fragmented manufacturing sector’s reliance on established industrial standards like PLCs and EtherCAT, as well as emerging software frameworks like ROS 2. Until the robotics industry adopts standardized protocols akin to the PC or mobile industries, widespread industrial deployment will remain a distant goal.

MD: As AI assumes a greater role in perception, planning and decision-making, how is the role of robot controllers and motion-control platforms evolving? Does AI increase the importance of the control layer rather than diminish it?

JS: AI increases the importance of the control layer. AI can improve perception, planning and decision-making, but industrial robots still need controllers that can execute motion precisely and safely. The more intelligence we add to robotics, the more important it becomes to have a reliable foundation underneath it.

If a robot is only moving boxes, the precision requirements may be relatively low. But if it is inserting memory cards into a server board or performing tasks with extremely short cycle times, the robot needs specialized motion control. AI provides the “brains” behind the system, but the control platform ensures the robot actually performs the task as intended.

MD: Open ecosystems vs. closed stacks: In industrial automation, what are the advantages—and potential risks—of applying that philosophy to robotics platforms, controls and software architectures?

JS: Open ecosystems can lower the barrier to robotics adoption. Instead of requiring every robot builder or machine maker to develop the entire stack themselves, open platforms allow companies to build on existing AI models, controllers, safety systems and application layers. This helps developers focus on the use case rather than the kernel-level robotics infrastructure.

That “openness” still has to be paired with industrial discipline. Robotics platforms must meet requirements for safety, cybersecurity, reliability and interoperability. “Open ecosystems” should not be loosely defined as “integrated systems with unclear accountability.” The combination of openness with certified control, safety and security foundations will give open ecosystems a true advantage.

MD: One promise of AI-native robotics is reducing the engineering effort required to deploy and reconfigure systems. Where are you seeing the greatest opportunities to shorten commissioning, programming and integration timelines today?

JS: Quite frankly, the current reality has yet to see promised engineering time savings; training AI models remains an incredibly time-consuming process that demands hundreds of thousands of hours of high-quality video data to capture diverse human workflows and skills. Beyond the heavy data gathering, systems still require an extensive trial-and-error phase on the factory floor before achieving operational readiness.

However, there is true opportunity in what happens once these robust video datasets are successfully established. AI has the unique capability to rapidly train robots across a vast spectrum of complex tasks, from automotive and aerospace maintenance to healthcare support, in a fraction of the time. This shifts the paradigm away from traditional, rigid programming methods, where engineers are limited to coding robots for a single, specific task at a time, towards a significant long-term advantage in multi-task deployment and agility.

The biggest opportunity is in developing platform-based robotics. Open platforms can allow builders to use ready-made AI models, control layers and safety technologies rather than developing everything from scratch. This can shorten development timelines for humanoid, quadruped and mobile robot builders because they can focus on applications such as navigation, perception, voice interaction or task-specific workflows. In that model, the controller and safety platform handle the core robotics foundation, while developers build higher-level functionality on top.

READ MORE: Flexible Robots, Messy Worlds: Inside Siemens’ Push for Practical Industrial AI

MD: As robots increasingly need to interact with vision systems, AI models, AMRs, PLCs and cloud platforms, how important is interoperability becoming as a design requirement for machine builders and end users?

JS: Interoperability is rapidly becoming a critical, non-negotiable design requirement rather than a nice-to-have feature. This is driven by the reality that modern robotics systems are no longer isolated machines, but interconnected nodes that must seamlessly collaborate with AI models, AMRs, safety sensors, cameras PLCs and cloud platforms. To navigate this complexity, open ecosystems are essential, as no single company can build every component alone, and machine builders increasingly require platforms that seamlessly bridge multiple layers of the automation stack.

Achieving true interoperability requires designing highly efficient communication protocols modeled after human interaction; incorporating bi-directional reasoning and dynamic acknowledgement mechanisms to ensure machines don’t just transmit data but understand intent and confirm execution in real time. For this vision to scale across production environments, global standards organizations such as the ISA, IEC and IEEE must step in to formally define and unify these advanced protocol frameworks for the next generation of the robotics industry.

MD: If AI reduces the amount of specialized robotics programming required, how might that change the skills manufacturers look for in engineers, operators and system integrators over the next five years?

JS: As AI reduces the reliance on traditional, specialized robotics programming, we are witnessing the emergence of a massive, yet still undefined, professional market centered around robotic behavior training and data engineering. Over the next five years, the industry will shift from syntax-heavy coding toward a complex, large-scale engineering discipline focused on high-quality video data collection and curation.

This new breed of specialists will not only need to gather massive volumes of visual data but also execute sophisticated post processing and annotation—embedding precise metadata while ensuring strict compliance, which includes removing corporate branding, blurring human faces for privacy and filtering out non-compliant footage.

Furthermore, training even a foundational action like walking requires capturing an immense variety of unstructured environments, from flat surfaces and inclines to highly irregular terrain. Because every robotic behavior demands this level of exhaustive, real-world visual training across endless permutations, the path to fully training AI native robots remains a monumental and prolonged engineering endeavor.

MD: When you look at the convergence of AI, robotics and open automation, what do you think today’s machine builders are underestimating, and what should they be preparing for right now?

JS: Many machine builders are underestimating how difficult it is to move from traditional automation to AI-enabled robotics. They may assume that AI can replace the need for specialized control, safety and robotics expertise, but industrial systems still require years of hardware iteration, motion-control development and safety certification.

Machine builders should prepare now by adopting AI-capable hardware platforms, building or joining open ecosystems, and seeking partners that can provide the missing pieces around real-time control and functional safety. Traditional PLC-based systems may not provide enough performance for AI control and AI safety. Combining AI innovation with industrial-grade control, safety and interoperability is the move companies must make to stay competitive.