Podcast: What does AI mean for your robot?

RB: That's very helpful, actually. Thank you for that. So let's bring it back to robotics. Are robotics automation systems well exposed to agentic AI at this time?

DS: I would say not. I am aware of two pilot cases. One was off the record, very large company. They built out a full agentic system, covered supply chain, covered logistics, and the company didn't deploy it for an undisclosed reason. I don't know why. 

And then we have Siemens, which at Hannover Messe a few weeks ago premiered Eigen, which is used for reprogramming robots. So Eigen is able to look at code and you can tell Eigen what it is you want to do, and then Eigen can make the changes for you. So is that agentic AI? I don't know. These things are just buzzwords, Robert. I first heard agentic two years ago in a conversation at Fluke Reliability’s Xcelerate conference, and since then, I just feel like agentic is another buzzword that's been thrown in front of AI just to keep us talking about it, because generative turned out to be not so useful. Agentic AI on the other hand…

RB: Well, this is fascinating for a couple of points. First, your exposure to it at Hannover Messe is interesting because that's a destination for a lot of companies to exhibit their really somewhat emerging technologies, not necessarily at the point of practical application. But my audiences will be tuning in very soon to IMTS, where the applications are a lot more practical and physically presented there. So we're sort of in the gray zone between these two possibilities or outputs. It makes me ask you specifically, is Eigen or any such model pointing us toward robots as a service and robotic programming as a service? Is that where you see this emerging?

DS: I think Eigen is more about self-service. It's being able to do that yourself, not having to depend on a vendor or integrator to reprogram your robots every time you have a different task for them. There's actually software that uses cameras to look at the motion of the human arm and then can translate that motion directly out to a cobot. So there are some tools out there to make things easier.

RB: But are we pointing users toward robotics as a service? That is to say, they invest in the arm or they invest in the installation, but all of the programming is done by some outside subscription-based model that they can access when needed to get the programming adjusted or retooled for them. Is that the direction?

DS: I mean, by evidence, I wrote an article about Behrens, the steel manufacturer. I know them mostly for making steel buckets. So they have two kinds of robots. They have the large industrial caged robots, and then they have the cobot. The industrial robots, people are familiar with, they're probably more familiar with programming. Cobot is something different. So Behrens went with the robots as a service model. That said, I think the program is getting easier such that I think the hardware, the robots themselves, are likely to remain strictly within the robots as the service model for SMBs. 

The programming, however, I usually find that at these companies, small companies deploying, there's usually a champion. There's somebody there who's really into robotics, really wants to learn, and we're getting into no-code software. It's becoming easier to reprogram. There's the camera-based system I described earlier. So I think programming might become less a robots-as-a-service model, and that robots-as-a-service would be more of the actual robots themselves, the physical hardware maintenance. But the program, I think even small businesses might be able to figure that out. I mean, the tools are there or they're right on the horizon.

RB: A lot of those walls were broken down by cobots already, so I think that the development would not be that much of a hurdle. So then let me ask, will agentic AI reduce implementation complexity for these smaller manufacturers or will they still need to access consultants and integrators to get the job done?

DS: I'd say for agentic, you're probably still talking about consultants and integrators only because I think agentic is more complicated. We talked before, we have the conductor, we have the orchestra. To have the conductor, you have to have the orchestra. Has everybody hired all the musicians they need to put together an orchestra, especially in small businesses? Probably not. Agentic again is very new. I think this is another example of technology companies throwing something advanced out at us because they need some talking points. When I think most small manufacturers, if they're stepping into this at all, they're probably still at the agent level. So the agentic is above that. It's more complicated, more useful, more complicated.

RB: Since you wrote fairly specifically about the Siemens development, which is called Eigen, this is described as a brand agnostic AI agent. Is it likely that smaller and mid-sized manufacturers can avoid dependency on proprietary AI ecosystems? Or are they simply turning themselves over to a new master?

DS: While it might sound a little bit like a snap answer, I think that anybody creating these systems had better make them brand agnostic if they want to be competitive, because especially, again, talking about small to medium-sized businesses, they don't have to worry about hooking in a proprietary ecosystem for each robot they have. So I think Eigen is paving the way that companies need to follow when they want to deploy these systems. They have to be brand agnostic. I think that's going to be one of the points of competition between different offerings. But yeah, absolutely.

RB: Are there provisions available to prevent the disasters and catastrophes for small and mid-sized manufacturers? You know, the nightmare scenario of the program taking over or the program doing exactly the opposite of what you want? How can they insulate themselves from that possibility?

DS: Data has to be validated by humans. Garbage in, garbage out. It's the same problem for any size manufacturer. Large manufacturers might have more people to do the task, but that's, I think, the most important thing, is humans are validating everything. And so in essence, you could think of not allowing the AI to take action, you know, as kind of a kill switch, a software kill switch.

And I just want to throw this in there because it's interesting. When you talk about kill switches and possible operational issues, humanoid robotics, I didn't see any easily accessible kill switches on those things. And I wonder where it's, you know, I mean, people are already naming little ones like Spot. They start naming humanoids when they really want to just turn them off. But yeah, I think that's my answer to that one. Humanoid robots.

RB: Yeah, it's something that we have to keep ourselves focused on, not lose the plot on this development here, especially if you're a manufacturer. Well, I want to thank you, Dennis. Your time today has been very well received and very illuminating to me. I want to encourage all our readers and listeners to go to industryweek.com. I'll provide the link, read his article titled Agentic AI Could Make Robots Affordable for Small Businesses. I love the premise, and I hope it's true. Thanks to the audience for listening, and please stay tuned to the next installment of the Great Question Podcast presented by Endeavor Business Media Manufacturing Group.

About the Podcast
Great Question: A Manufacturing Podcast offers news and information for the people who make, store and move things and those who manage and maintain the facilities where that work gets done. Manufacturers from chemical producers to automakers to machine shops can listen for critical insights into the technologies, economic conditions and best practices that can influence how to best run facilities to reach operational excellence.

Listen to another episode and subscribe on your favorite podcast app