Back

Andy Lonsberry of Path Robotics on Why Physical AI Is Ready for the Factory Floor

Andy Lonsberry co-founded Path Robotics with his brother eight years ago with a mission to help rebuild U.S. manufacturing as a source of national strength. As CTO, he leads development of Obsidian, the company's core neural network, and the custom sensors that let robots handle skilled labor tasks. Path's robots are now deployed across the U.S., Canada, and Mexico, spanning shipbuilding, energy infrastructure, AI infrastructure, and heavy industry. In this episode, Andy shares why he believes this moment in physical AI is fundamentally different from a decade of prior automation attempts, and what it will take to keep production growing onshore as the skilled labor shortage deepens.

Listen Stations:

Apple podcast Spotify Google podcast Stitcher

Stay connected
https://www.linkedin.com/showcase/forging-the-future-with-chris-howard/

Stay inspired and ahead of the curve by subscribing to Forging the Future. Share your thoughts
on this episode with the hashtag #ForgingTheFuture or tag us online!

Click to read the Transcript

🎧 Episode Highlights

[01:38] What Path Robotics actually builds, and why they combined a large neural network with a custom sensor to make robots capable of skilled labor tasks.
[02:52] Why welding, specifically — the result of interviewing 100 U.S. manufacturers about their single biggest bottleneck to scaling.
[03:55] "Robots are coming to keep your jobs." Why 100% of Path's customers are trying to grow production, not cut headcount.
[08:16] Why no single robot form factor wins. Path's traditional six-degree-of-freedom arms handle indoor manufacturing, while its new quadruped, Rove, is built for outdoor work like shipyards and construction sites.
[13:54] Teaching a robot to see and learn: the custom sensor built to see through welding smoke and arc glare, and the "Weld World" model that predicts welding physics before a second model learns to become a great welder through reinforcement learning.
[24:28] The surprising discovery that production-based pay incentives turned skeptical floor workers into the loudest advocates for getting more robots on the line.
[31:00] Why complex, judgment-heavy work will likely always stay human, and how "Mission Control" telemetry lets Path monitor every deployed robot in real time.
[36:51] The customer case study: a part that used to take 150 manual welding hours saw a 91% reduction in labor time, unlocking production volume the customer couldn't hit with people alone.
[38:48] The labor math that makes this urgent: the American Welding Society estimates 320,000 additional welders are needed over the next four years — a number Andy calls "not even close to being possible" to hire.

🔑 Key Takeaways:

Takeaway 1: Physical AI in manufacturing isn't displacing workers, it's filling a gap that traditional hiring can't close. Every one of Path's customers adopted the technology to grow production, not to cut headcount, and the skilled labor shortage (320,000 additional welders needed over four years, by one industry estimate) makes the case starker than a typical automation pitch.
Takeaway 2: No single robot form factor will dominate. Traditional arms, quadrupeds, and humanoids each have distinct advantages depending on the environment. The real differentiator is the intelligence layer that can control any of them, not the hardware itself.
Takeaway 3: Getting buy-in from the shop floor may matter as much as the technology itself. When one customer tied weekly production bonuses to output, the workers most likely to resist automation became the ones requesting more robots, turning a feared job-loss narrative into aligned incentives for everyone.