ff Venture Capital

John Frankel is a technology investor with a background in capital markets and technology development. He focuses on early-stage companies developing applied AI, robotics and other emerging technologies.

Beyond Traditional Automation

As an early-stage investor, I have spent the past quarter-century watching software, and then AI, eat the world. The next wave of disruption will not come from code alone. It will come from AI systems that can perceive, reason about and manipulate the physical world. Physical AI represents the convergence of AI, advanced robotics and sensor technology, and it is poised to reshape manufacturing more profoundly than any innovation since the assembly line.

Traditional industrial robots excel at repetition. Program them correctly, and they will weld the same seam or insert the same component millions of times with superhuman precision. Ask them to handle something unexpected, however, like a part that is slightly out of position, a new product variant or an obstacle in their path, and they fail.

Physical AI systems are fundamentally different. By combining computer vision, large language models, world models, reinforcement learning and sophisticated actuators, these machines can adapt to unpredictable environments and learn new tasks without explicit reprogramming. They perceive their surroundings in real time, reason about what they observe and adjust their actions accordingly. Or, at least, they are about to.

Companies like Figure AI and Tesla have demonstrated the potential of humanoid robots that learn manufacturing tasks through demonstration rather than traditional programming. Similar advances are emerging across specialized applications where AI-powered systems are already delivering value.

PlusOne Robotics illustrates this shift. Its warehouse robots can be guided by remote operators when they encounter unfamiliar packaging or damaged items, allowing the system to learn and improve continuously. The company's dual-arm systems process up to 3,300 parcels per hour for FedEx and DHL.

CivRobotics applies similar adaptability to construction surveying with autonomous robots that navigate complex terrain while maintaining sub-centimeter accuracy. Burro extends the concept to outdoor operations, where autonomous vehicles mow, tow, move, monitor and carry materials across farms and other challenging environments.

Expanding Manufacturing Possibilities

Physical AI enables manufacturing models that were not previously feasible. Micro-factories with minimal human oversight, rapid reconfiguration for small-batch production and lights-out facilities operating around the clock all become economically viable.

The technology does more than reduce costs. It expands manufacturing flexibility, supports new production models and increases operational reach. The factories of tomorrow will operate differently from those of the past because intelligent automation enables both efficiency and adaptability.

Building the Next Generation of Physical AI

We are at the frontier of what is possible. Current physical AI systems perform well in structured environments but continue to evolve for the complexity of real-world operations. Dexterous manipulation, handling fragile materials and recovering from unfamiliar situations remain active areas of innovation.

“Physical AI represents the convergence of AI, advanced robotics and sensor technology, and it is poised to reshape manufacturing more profoundly than any innovation since the assembly line.”

Training data remains one of the industry's biggest challenges. Unlike software AI, physical AI systems require real-world interaction data. The data is the moat. Simulations help, but the gap between virtual and physical environments—the ‘sim-toreal’ problem—remains a major bottleneck. Companies continue investing in simulation technologies and techniques that improve knowledge transfer into real-world applications.

Battery technology, edge computing and advanced sensors have historically constrained deployment, but those barriers are gradually easing. As the technology matures, opportunities are emerging across AI models, robotics hardware and applicationspecific systems.

The greatest potential, in my view, lies in combining advanced AI with purpose-built hardware for defined use cases. Fabrica illustrates this approach with its autonomous tile-grouting robot, which delivers fivefold productivity gains while allowing one operator to manage up to five robots. Companies that can also accelerate training and solve the sim-to-real challenge will help move the entire industry forward.

Capturing the Next Competitive Advantage

We are witnessing the early stages of a fundamental shift in how physical goods are produced. Physical AI will not replace all human workers. Manufacturing will continue to require human judgment, creativity and oversight, but these systems will reshape the nature of manufacturing work and the economics of production.

Companies that move quickly to adopt and integrate these technologies will gain meaningful competitive advantages. Those that delay risk falling behind competitors capable of producing higher-quality products with greater flexibility and lower costs.

Physical AI is no longer a distant possibility. It is already reshaping manufacturing. The question is no longer whether it will transform the industry, but which organizations will be best positioned to capture the value it creates.