In industrial markets, like agriculture, interoperability is essential across all AI-enabled equipment. Walk any commercial farm operation and you'll find a mixed fleet: tractors from different brands, implements from other manufacturers, machines that vary in age by decades.
Physical AI in Agriculture and Rugged Industries
Q&A with Nic Fischer, CTO | Agtonomy
Tell us about yourself and your role at Agtonomy.
I'm the CTO at Agtonomy, and honestly, robotics has been my whole career. After finishing my PhD in dynamics, systems and control at the University of Florida in 2012, I spent years building autonomous vehicles across every domain — underwater vehicles at Bluefin Robotics, drones at PRENAV, and then Serve, the delivery robot I helped build as Systems & Robotics Lead at Postmates, later acquired by Uber. I joined Agtonomy in February 2021, right at the founding, as Head of Autonomy. As we grew, I took on all of software in 2024, and then stepped into the CTO role in June 2025, where I now own our system architecture, technology direction, and the execution of our physical AI platform end to end.
What keeps me here is the same thing that's driven me across every platform I've worked on — ground, air, or sea: building commercial-scale autonomy that's genuinely useful in the real world, not just in a demo. That means pulling in the right lessons from other industries and building a team that can actually deploy and scale it.
What do you think is going to be the most impactful physical AI innovation in the next year?
The most impactful physical AI won't be humanoid. For all the hype, humanoids are still chasing a general-purpose dream, while purpose-built machines of every shape and size are already out doing real work. The biggest near-term opportunity is in off-road industrial markets: construction alone generates roughly $15 trillion a year, agriculture isn't far behind at over $12 trillion, and both face skilled labor shortages that get worse every season. The work in these industries is physical, repetitive, and increasingly impossible to staff — exactly the conditions where automation pays off first. These machines are in the field today, and over the next year I expect them to deliver more real-world impact than the entire humanoid category combined.
That's exactly why we're expanding beyond agriculture and into construction. Jobsites are facing the same labor math we've seen on farms, and repeatable tasks like grading, ditching, and material handling are a natural fit for the same physical AI platform. It's demand pulling us there, not the other way around — customers are asking for autonomy because there simply aren't enough skilled operators to hire.
Agriculture and other rugged industries are especially difficult to automate because robots have to operate in highly dynamic outdoor environments: mud, uneven terrain, lighting variability, and weather. What engineering tradeoffs have been most important when designing robotic systems that can maintain reliable perception and autonomy performance under those real-world conditions?
The most important tradeoff was deciding what to build and what to inherit. Industrial AI combines software, sensors, robotics, and AI and in this space the machines themselves are already excellent. OEMs have spent a century learning to build equipment that survives mud, dust, vibration, and heat, backed by dealer networks that keep it running. Rebuilding all of that from scratch would have cost us a decade of hard learnings and iterations. Instead, we put the brains into the brawn: sensors, compute, and autonomy software factory-integrate into proven platforms, and durability stays solved by the people who solved it best.
The second tradeoff is designing for the worst hour of the day, rather than the demo. Outdoors, every sensor fails somewhere. Dust and glare blind cameras, canopy blocks GPS, and dense foliage kills radio signal. So we fuse redundant sensing modalities and treat degraded conditions as the normal operating case, not the exception. We apply the same rule to compute and connectivity: everything the machine needs to perceive, decide, and act runs onboard, so a weak signal at the far end of a field never stops the work.
A recurring theme in field robotics is that customers don’t want “science projects”; they want systems that integrate into existing workflows with minimal operational friction. How do you approach building human-machine interfaces that augment existing operations rather than forcing users to redesign labor, infrastructure, or processes around the robot?
We start from the workflow and work backward to the technology. A farm already plans its work in blocks, tasks, and schedules: mow this block Tuesday, spray that one when the wind drops. Our interface mirrors that structure: a manager plans, dispatches, and monitors work from a mobile app on the phone already in their pocket, and the machines execute the plan.
The same principle shapes everything around the interface. There is no infrastructure to install, no beacons, wires, or towers; the machines work in the fields as they are. The equipment is a trusted platform from a brand the customer already owns and a dealer they already know, and anyone can still climb in and drive it manually when the job calls for it. We hold ourselves to a simple bar: an operator should be able to put the fleet of machines to work on day one without changing how the farm operates.
That's also why this model travels so well beyond the farm. A construction crew or a land maintenance company already plans in blocks and trusted equipment too — the interface just has to fit into that world, not force them to rebuild it.
Many industrial environments aren't homogeneous, yet OEMs often design as if customers will operate within a single vendor ecosystem. As physical AI adoption accelerates, how critical is interoperability across equipment, data systems, and vendors?
In industrial markets, like agriculture, interoperability is essential across all AI-enabled equipment. Walk any commercial farm operation and you'll find a mixed fleet: tractors from different brands, implements from other manufacturers, machines that vary in age by decades. No single color dominates, and customers have no appetite for being forced into one. An equipment technology provider that designs autonomy as a closed ecosystem is designing against how customers actually work.
I've seen this same pattern in every domain I've worked in before agriculture. Underwater vehicles have to talk to sonar and navigation gear from a dozen manufacturers, delivery robots have to work with whatever sidewalk and city infrastructure already exists, and none of it waits for a single vendor's roadmap. Sea, air, or land, the platforms that actually scale are never the ones that assume a clean, single-vendor world.
We built for that reality with a multi-partner approach. The autonomy, the fleet or implement interface, and the data stay consistent across brands and machine types. A manager plans work once and machines of any color execute it. Operational data flows into the customer's own systems instead of sitting in a vendor portal. As physical AI adoption accelerates, customers will standardize on the layer that works across their whole fleet, not the one that works for a single logo.
What do you envision as the future of the workforce in these types of rugged industrial environments?
The autonomous machines become the workforce, and people move up to managing it. Instead of one person on one machine for ten hours in dust and heat, a newly skilled industry tech worker runs a fleet of machines of all shapes and sizes getting real work done, planning the jobs and handling the exceptions. That role is safer, pays better, and looks a lot more like the career a younger generation actually wants.
We should also be honest about what gets displaced: mostly work that already goes undone. The labor shortage isn't a forecast — in agriculture, 56% of farmers already report they can't find enough workers, and in 2025 just 182 domestic applicants came forward for the more than 415,000 farm jobs posted through the H-2A program. Construction faces the same math, needing an estimated 349,000 net new workers in 2026 alone. In both industries, the people who hold these skills are aging out faster than they're being replaced. Autonomy fills that gap and creates roles that did not exist before: fleet supervisors and autonomy-trained technicians in dealer networks, what we're already calling "AgTech operators". That's how skilled work stays in rural communities instead of leaving with the workforce that used to do it.
That labor math is also why we didn't have to chase our way into construction and land management — those industries came to us. They're running the same aging, understaffed operations, and once they see this workforce model solving it on the farm, they want the same fix on their jobsites.
As we consider rapid adoption of physical AI across different outdoor industries, what training and skills are needed to attract and retain key talent?
The talent this field needs is drawn by hard problems, and physical AI in off-road environments is one of the hardest in robotics. The work spans firmware, controls, perception, machine learning, and cloud infrastructure, and it ships into conditions that expose every weakness. A demo video proves nothing here; the machine either did the season's work or it didn't.
The skills that matter most are systems thinking across those disciplines, classical robotics fundamentals combined with modern machine learning, and a willingness to get your boots muddy. Our best engineers spend time in the field watching customers use what they built, and that feedback loop teaches faster than any course.
I'm proof of how fast that loop can move a career. I joined Agtonomy in 2021 to lead autonomous software, took on all of engineering three years later, and became CTO within four. Not because I planned it that way, but because the work kept expanding as fast as I could learn it. That's the trajectory this field rewards: show up, solve the hard problem in front of you, and the scope follows.
Retention works the same way. Engineers stay where their work ships to real customers and shows up in measurable results at the end of a season. Helping close a real labor gap in food production, and now construction, is the kind of impact that keeps me and my engineering team inspired — and the kind of mission that pulls new talent in.
Nic Fischer, CTO at Agtonomy, is a multi-disciplinary technical and product leader specializing in robotic autonomy, systems architecture, and software for complex ground, airborne, and maritime platforms. At Agtonomy, Nic heads technology strategy and product innovation for physical AI in off-road environments, building deterministically safe, commercial-scale autonomy that allows a single operator to supervise fleets of intelligent machines in real-world conditions.
He previously led systems and robotics at Postmates by Uber (developing the award-winning Serve last-mile delivery robot) and has held senior engineering roles at Bluefin Robotics, PRENAV, and other autonomy ventures, with multiple patents in autonomous systems and control. He holds a Ph.D. in mechanical engineering with a focus on dynamics, systems and control, and a B.S. in mechanical engineering from the University of Florida, where his research centered on Lyapunov-based nonlinear control theory, computer vision, and mobile robotics.
The content & opinions in this article are the author’s and do not necessarily represent the views of RoboticsTomorrow
Featured Product
