Robots & Physical AI:
EXPLAINED

Introduction: What is Physical AI?

Michael Krigsman: An AI chatbot that hallucinates is annoying, but robots that screw up can cause death and injury. Burkhard Boeckem is CTO of Hexagon, which builds physical AI.

Burkhard Boeckem: Physical AI is actually AI that shows up in the real world. It is embedded in machines that move, operate, and affect people, assets, and operations. Physical AI lets autonomous systems, like drones or robots, perceive, understand, reason, and perform actions in the real world.

Michael Krigsman: The core aspect here is operating in the real world, and when we talk about the AI aspect, where does that get overlaid into this?

Burkhard Boeckem: It's important for physical AI, and specifically you just mentioned, if a chatbot is hallucinating, it's annoying. But if a humanoid, for example, is hallucinating or getting out of control, it can be dangerous. So basically, the bar is much higher for physical AI because it affects the real world.

That means you have to ground your AI in reality. That means you use calibrated sensors, verified maps, do accurate reality capture in order to create an exact replica of the real world into a digital twin.

And you constantly have to update this, so it's evergreen, and the system needs to be anchored in the truth, so that a robot knows, perceives things right and can handle things accordingly or perform actions and tasks accordingly, but into the real world, and therefore you need the strong connection into the reality or the strong link of the digital twin to the reality.

Digital Twins as the Foundation

Michael Krigsman: You've used the term digital twin a couple of times. How does the digital twin aspect relate to physical AI?

Burkhard Boeckem: I would say it's the base. Because robots need to learn, and they do it in various techniques and technologies, imitation learning, reinforcement learning, and so forth.

But I would say the most straightforward and the one that is laying the bar highest for functional safety, for instance, is if you take the reality, the real world, and map it one-to-one into a digital reality with a precision and accuracy that guarantees you the ground truth that is needed for the actions.

And then ideally, your robot has on board a lot of sensors that first of all perceive the world, and then basically can also control the manipulating or the handling or the dexterity of the robot, or the locomotion, where the robot is going.

Understanding Ground Truth

Michael Krigsman: You have used the term ground truth.

Burkhard Boeckem: Yeah.

Michael Krigsman: What do you mean by that? We're really covering the basics of physical AI, and then, of course, we'll get more into the details, but this is super helpful to give us an understanding of a CTO of Hexagon, which develops these products, to give us a basic context.

Burkhard Boeckem: What is important, and also where you base afterwards your simulations on, is to have an understanding of the real world, of the real environment.

And you can do beautiful things in terms that if you accurately create a digital twin by reality capture technologies, by vision technologies, for example, and this is dimensionally accurate, this is accurate in scale, and so forth, then you can actually train the robot in this digital twin and try things out, make errors and failures in the digital twin.

And then when you deploy it into the real world, where you trained it on the digital twin, it's far more easy to do this. The robot directly knows the environment, perceives the right things.

And what I also find, it has to have a certain sensor stack on board because when you created the digital twin, and we say for autonomous systems, the digital twin should be autonomously updated, but that only works to a certain extent.

So for the things that have changed, for example, a table or a chair that have been moved, or basically a workplace that looks slightly different at the time when you generated the digital twin, the robot needs to have enough perception sensors and reality capture sensors on board in order to perceive the difference and act accordingly.

But I think the dimensional ground truth, and I think that's where the interaction between the matter, the workspaces, things you do, and also collaborating with actual humans starts, is the ability to base it all off of reality.

And I think that's also a hard problem that we need to solve furthermore in the future, is to close this gap between the reality, the digital twin, and then the digital twin is the sandbox where you simulate and train on in order to close this real to sim gap again when deploying into reality, because robots in essence then live in reality, in the real world.

Digital AI vs. Physical AI: Safety, Reliability, and Metrics

Michael Krigsman: Folks, if you're watching on LinkedIn, pop your questions into the LinkedIn chat. If you're watching on Twitter, X, use the hashtag CXOTalk.

Burkhard, can you talk about some of the differences between AI, digital AI on a computer versus what happens when AI operates in the physical world, in areas like safety, cost, reliability?

Burkhard Boeckem: It's also two very different things, because we all know the digital world, and we know the agents and the chatbots and the copilots, and so forth.

But let's focus for a second on the physical world. Where robots and physical AI make a difference or need to make a difference is basically if they solve and help humans to collaborate and do meaningful things.

And here we're talking about very different metrics. For example, we are talking about predictability, we talk about uptime, we talk about safety, and that's something that needs to be reliable, modeled into the AI.

And as I said before, we're talking about maintaining security, workflows that have to be fulfilled in order. So it's not about showing the next coolest trick. Also, I like this very much. If I see a robot doing something stunning or having a very liquid motion and these kind of things. No, it's probably first and foremost about the uptime, the reliability, and performing the workflows we intended it to do.

Real-World Business Applications

Michael Krigsman: So it's really all about the business applications and being very clearly focused on the business problems that are trying to be solved.

Burkhard Boeckem: Absolutely. You see, of course, a lot of YouTube videos where humanoids or robots are launched that do super cool things, and it's so liquid these days.

But I think the really hard tasks look actually a little bit more boring, to be honest, is working in, for example, material handling, checking things, quality inspection.

What, for example, we have shown when we launched our humanoid was our humanoid scanning with a laser scanner the door of a car for defects. It's probably far less cool to watch, like a robot dancing or doing somersaults and these kind of things, but these are the things that we expect specifically humanoids to do in the future to solve the tasks that are tedious, maybe dangerous, maybe in scenarios where you have a high degree of repeatability, but the workplace is changing, and these kind of things.

And probably less about general purpose. It's doing jobs, but being very flexible in doing jobs. Otherwise, you could install a stationary robot for just doing one or few things, and it would probably outperform what we currently see in humanoids.

Security and Functional Safety

Michael Krigsman: Anthony Scriffignano, who is a data scientist and a prominent data scientist, and he has been a guest on CXOTalk a number of times, he actually asks about physical AI exploits by malefactors, and you mentioned security earlier. So can you talk a little bit about that aspect?

Burkhard Boeckem: It starts already with the design of the robot, and it's also basically in how you guarantee functional safety along the way in all the aspects. It starts also, I lately looked at a report where we have shown why are the current humanoids not giants and these kind of things.

So it starts in building in things, in building, separating also the control loop from the AI, the physical AI parts of it.

But you have to be in a position, and probably the best design goes in a direction, is that the robot does know when it does not know anymore, when it's lost, and slowing down, stopping, instead of doing something, going on, tripping over, these kind of things.

And also, it's important by design that you design quite fail-safe. You have independency, redundancy in what you do, and also have a sensor stack that allows you to perceive if things go basically out of the range where you trained it for, whether it's by reinforcement or imitation learning or other techniques in physical AI.

But I think what is important is that you control all the stack. It comes from the locomotion, the dexterity, the sensor stack, and then basically the physical AI. That this is really fail-safe because it could end up quite badly in a disaster if you trip over on something and people might get hurt, and that is basically to avoid at all costs in terms of, I would say the big theme that is coming in 2026, which is functional safety.

Michael Krigsman: Unlike digital AI, where in most instances an exploit can have an impact on a network or have financial consequences, with physical AI, there can be physical damage that's caused. So you have this additional layer of complexity.

Burkhard Boeckem: Yes, absolutely. Manufacturers or development teams of specifically humanoids or quadrupeds or drones, or any autonomous system that interacts and is inherently dangerous in this regard. You have to be able to control the whole system, not just the model.

So it's about the sensors, the compute, the controls, the integration, the maintenance even. It's nice if it's working out of the box, but what if there's wear and tear, and over time, systems maybe not perform to the standards you had them at the beginning, and then security.

So all the workflows, these workflows determine whether it actually works in the real world. So I would say the bar is far higher than in, as you call it, digital AI.

Industry Announcements and the State of the Market

Michael Krigsman: Greg Walters asks the question that I think lots of us have, which is: at CES, Computer Electronics Show, there have been lots of AI announcements from Atlas and Gemini to NVIDIA. Do you have any comments and thoughts about all the announcements that took place at CES?

Burkhard Boeckem: It's hard to comment on every one on its own, because there are so many. What we currently see, and when we started, for instance, our humanoid project, and it was basically based off of our experience we made with autonomous drones and what we made of quadruped products, or the robot dogs, as we like to call them.

That probably, and these announcements show the way, and probably with an ever-increasing pace, that when we started this, we thought, "How in the world are we solving all this control loop tasks that we have to do?" And with all the announcements, all the technology that came, it brought more and more tools to basically build robots.

And I would say currently, we don't see the end there, and what we saw for NVIDIA, the physical AI announcement around physical AI and, as NVIDIA is a good partner for us. We have been on the trajectory with them for a longer time now.

But it's amazing to see what is the capability today of physical AI and humanoids.

Yet, I would say, 26 and 27 is probably the years where we maybe step back a bit and look for deploying robots in actual manufacturing scenarios, in workspace scenarios, in order to solve real-world problems.

Because we have seen now that locomotion is great. The liquidity we see in the motion, the fluidity of robots is stunning. It's so stunning that you probably need to take them apart on the stage, that you see that there's not a human in a costume.

But yet to be seen are tasks, I have a box full of gearbox, of gears, for example, and there are maybe 20 in there that don't belong in the box. Can a humanoid sort them out by looking at them? Can they measure it? Can they see whether it's in tolerance or yes? Can it be taken further?

I think that's yet to be proven, these use cases, that these use cases can be fulfilled. I'm an absolute optimist in terms of robotics and humanoids, and I think it's absolutely possible.

But I think this year and the next year is probably the ones where we see a lot of robots, humanoids specifically, deployed in these real-world scenarios. Maybe still isolated from the actual workforce, and then doing this task.

And you probably need then a more refined physical AI to do specific things, starting with various interactions. Already sorting and picking things is one that we should see that is quite doable in the future.

Cloud vs. Edge Computing in Robotics

Michael Krigsman: Let's jump to another question, this time back to Twitter, to X. You can see I love taking the audience questions. You guys in the audience are so smart.

And by the way, now would be an excellent time to subscribe to the CXOTalk newsletter. Go to cxotalk.com right now, subscribe to our newsletter, and we'll keep you updated, because we have really incredible shows all the time coming up, and you should join us.

Okay, here is a question, a technical question from Chris Peterson on Twitter, who says, "Do you see embodied AI and robotic systems using more centralized cloud, distributed edge, or both?" So is it cloud, edge computing, or both?

Burkhard Boeckem: I would say both. The cloud part is, as we said in the very beginning of the discussion, very important when we talk very big data in order to create digital twins, and then the sandboxes for the robots to learn, and then also in order to orchestrate, maybe in the future.

Currently, we have a limited amount of robots, but when there's a large amount of robots, you need also to orchestrate the fleets. And that's probably what you ought to do within a digital twin environment.

However, to be fully autonomous, to fulfill tasks, and basically also learn on the job, I think the edge AI is decisive. That's why I think it's also a success story to have sufficient compute in your robot.

But I think together with something probably that is in the direction of federated learning, federated AI, where one robot learns everything for the fleet, you probably need to find a good balance between the cloud and the edge when there's then a deployment of a whole fleet.

But I think in humanoids specifically, or in robots in general, I think edge AI is really decisive for them, because the more compute, the more things you do. It's going more into the direction of a multipurpose robot, which I think is still a little bit out, but I think edge AI is really important, have to have enough compute on the edge.

Regulations and Responsible AI for Physical Systems

Michael Krigsman: This is from Jotaramm Ladd, who says, "Most responsible AI regulations are written with digital AI in mind. Do you think today's regulatory frameworks are truly fit for physical AI, where AI decisions directly affect the real, the physical world?"

And Anthony Scriffignano asks the same question, essentially. He says, "Are there emerging regulations beyond general AI in the context of physical AI?" So both of them want to know about regulations and responsible AI, and how is it different when it comes to physical AI?

Burkhard Boeckem: What we see now is, and I recently gave an interview about this, functional safety and the upcoming regulations, and I dug deeper into the topic.

Maybe to keep it on this level, what we currently see is probably what we've seen with the UAVs and the drones a couple of years back, that while the technology is emerging and the capabilities and the skills are coming up, then also the regulatory body will come further.

We have probably a very strict body already in the EU in order there, but probably for safety regulation in terms of physical AI, we see probably in this year and next year more regulatory bodies coming up or basically manifesting.

And I think that functional safety is still something that currently everybody's happy that the locomotion looks good and these kind of things, but needs to address that, and it needs to address if you have a larger amount of, for example, humanoids working together, collaborating with humans.

And so in short, yes, I anticipate that we see a rise in the regulatory bodies for specifically physical AI. Because what we have currently is machine guidelines, which probably are not that sufficient for the autonomous systems, although they already cover a fair part.

Addressing Bias in Physical AI

Michael Krigsman: And we have a related question from Arsalan Khan on Twitter, who says, "AI algorithms use data to make decisions. In non-physical AI, bias can potentially be corrected by updating the algorithm and the data. How do you tackle bias in physical AI?" And he's asking also whether we need regulations. He's also interested in the regulation question. So the issue of bias in physical AI.

Burkhard Boeckem: It starts already with the accurate representation of the real world in a digital twin.

If you basically solely base your deployments and your trainings on synthetic data without dimensional accuracy and without the right scale factors and so forth, I think this is where you introduce the bias.

If you train in scenarios that are based on digital twins, which are dimensionally accurate, and an exact replica of the physical world, I think you can limit this bias.

And then it's also one thing, that the robot itself is capable of sensing the bias and sensing whether, for example, whether the perceived step height is X, Y, Z, with enough sensing power on board, and that also influences the compute then.

I think foremost, these are the things, and then basically, once you've found a bias, you have to close the loop and bring it back. But that's something, to be honest, that we did in other industries and in other technologies since quite some time.

Talking about, for example, the quality inspection of a car. There, basically, you close the loop from the CAD drawing to the machine part. You inspect, and if it's out of tolerance, you alter your machine tool.

And I think maybe this is a very mechanical explanation of this, but the same needs to hold true here as well.

Timeline to Autonomy: How Close Are We?

Michael Krigsman: Let's go to another question. This is from SAI Penumuru, and Sai says, "How close are we really to autonomous operations, and what is the one thing still preventing us from getting there?"

And Greg Walters asks a very simple, very related question. He says, "This is the next stage. Embodied AI is general once it has all the 6 senses of a human. How close are we to this?" So basically, how close to autonomy and to this idea of robots as really helper humanoids?

Burkhard Boeckem: Very interesting topic, and good questions in this. And I think that what we currently see is probably still, of the humanoids in demos and so forth, are teleoperated and these kind of things.

However, I think probably we are, in certain tasks, not that far away from autonomy if we basically constrain the environment to some extent.

For me, it depends a little bit on what is the capability of the sensor stack of the robot. That said, we have currently also a lot of imaging devices, a lot of time-of-flight cameras, microphones, and so forth, in the sensor stack.

The sensor stack is not yet fully there where basically humans are, to my knowledge, in what we currently see on the market or on the prototype stages.

But I would say strong vision systems, strong lidar systems, combined with some specialty systems that do, for example, touch sensitivity in the end effectors, enable us already now to basically autonomously perform certain tasks.

For example, picking things and placing things. I would say there we are already quite advanced, probably in the longer term, because I think the whole humanoid business was quite fascinated with locomotion and basic dexterity right now.

I think we see in the future much more sensors being built in, and that enables then autonomy.

Probably drones are a little further advanced, what we've seen here, in this, but also the interaction with matter is, of course, less here.

But I would carefully estimate there that we are, yeah, maybe one to max three years away for autonomy in certain constrained areas. I would be very careful in saying general purpose everywhere, or a robot in a household can do everything. I would be careful with this, but in constrained environments, it's foreseeable. Yeah.

Creating Economic Value: Beyond Demos

Michael Krigsman: Can you talk about the timeline for robots to create actual economic value, not just demos? And of course, that's related to what you were just discussing.

Burkhard Boeckem: There are various stages and phases where we're in. Probably what we've seen in the last couple of months and maybe year is a lot of announcements with robot manufacturers and I would say early adopter companies, whether it's automotive and so forth.

And I think in this stage, certain use cases were tested in a lab environment.

And I would see currently that this goes out now onto a true manufacturing site, but not in the quantities of dozens of robots. Maybe a small amount in a constrained part of the factories. I think that's what we can expect fully this year, yeah, in this.

And then a more rolled out deployment probably then the year after. But still to some extent, constrained, and to some extent limited to certain use cases.

Where Robotics Deployments Fail

Michael Krigsman: Can you describe where robotics deployments today most often fail? Is it technology limitations, integrations with existing systems, organizational readiness? Where are the obstacles today?

Burkhard Boeckem: Probably the technology is already quite advanced in this, but it does not have a track record yet in terms of deploying hundreds and thousands.

So basically, all the teething errors, the teething problems that you would get with a bigger fleet, and all the maintenance, and all the serviceability topics, are probably not addressed in something if you compare it, for example, to drones or even in higher scale, cars, at this point in time.

They're basically, we yet need to see how maintenance intensive such humanoids are.

But probably, the failure level is higher, not on the technology and the capability part, but probably for, I would say enterprise is not yet ready for basically welcoming a humanoid fleet, and welcoming it to work alongside with the workflows.

We just talked about the regulations. I think there we need to do a couple of steps more in order to have this correct.

And then basically on the other side is, for me, the question still open then, how do enterprises, corporates, basically incorporate them, the humanoids alongside with workers and operators in an actual factory or manufacturing, discrete manufacturing environment?

Probably the latter part is the more decisive going forward, because technology will evolve and we will see it quite quickly.

The Future of Humanoid Form Factors

Michael Krigsman: And Greg Walters on LinkedIn asks a question directly relating to this. He said, "Aren't humanoid robots in a human-like form soon to be obsolete in the manufacturing realm, as new production lines will be designed to be run without humans?"

Burkhard Boeckem: We have a division that is called Manufacturing Intelligence, and they're working towards what is called lights-off manufacturing, where basically manufacturing is, or the manufacturing environment is designed for having no humans at all in the loop. And even taking the lights off in order to, it runs autonomously, the whole manufacture.

I think for these environments, you probably would not develop a humanoid for.

However, there's still a world that is made for humans. There's still a large legacy all over the world, where basically sometimes whole shifts break down because the machine to adjust has to be reset.

And these environments are made for the last decades for humans. So I think the proportion, the height, and everything, is still very beneficial for a humanoid.

But I give you 100% right, if you design a fully autonomous lights-off manufacturing plant, then probably humanoid would be not the desired form factor.

But for everything else, and there's a large legacy still existing, I would say the humanoid is a very nice form factor for this.

Because we have seen it in the, there was an early hype for these robot dogs. And we developed scanners that went on robot dogs. It sometimes did not work because there was an elevator button to push, or it was not the right point of view for the system.

These are the things humanoids, of course, overcome, but I give you right, for an autonomously designed manufacturing place, you would probably work differently.

The Humanoid as User Interface

Michael Krigsman: So the humanoid form, could we say that it's essentially the user interface?

Burkhard Boeckem: I would say, right. Well, basically, it's the embodiment that resembles the user interface that we are used to in this regard. Yeah.

Michael Krigsman: So we have the robotic system that is independent of the particular form in which it's embodied, right? You could have a robotic system that's inside a drone, something flying, or a dog, or a humanoid, or a machine in a factory, essentially, right? So it's the packaging, it's the external packaging.

Burkhard Boeckem: No, absolutely. And I think what we still see is that in the humanoid there, they're not trying to exaggerate certain qualities of a human, not making it much faster, not longer legs, nor slightly longer arms, these kind of things.

Now, we see quite a nice representation of a human as far as you want to go, without going into the uncanny valley situation.

Digital Twins for Robotics: What Makes Them Different

Michael Krigsman: Talk more about digital twins. You described earlier how the digital twin concept is foundational, but how are robotic and humanoid digital twins different or more complex from those used in industry? For example, in jet engine maintenance operations, which is an example that we hear about all the time.

Burkhard Boeckem: What is very important for these autonomous systems and the digital twin, and for example, certain, a digital twin is always fit for purpose.

For example, if you want a digital twin of a jet engine, then certain qualities need to be really in this digital twin in order to drive a simulation, in order to manufacture it, in order to lay out your production of the blades and blisks and whatever you need for this.

What is important for the digital twin that you have in the space of robotics, is that the autonomy of these systems requires that these digital twins have to be alive.

Of course, they need to have the geometry, physics, and the right scale.

However, you have to acknowledge the sensor behavior, the dynamic obstacles that there are, and basically, they have to be constantly updated because, yeah, a digital twin, and yeah, famously, I think the probably the most famous digital twin years back was basically the digital twin that allowed the reconstruction of Notre Dame in Paris.

That was basically by accident or by luck, we had a digital twin of this. This is an absolute static one, and you want to rebuild a building from this.

Digital twins for robots need to be evergreen, always updated, and sometimes even in real time, and that's what differentiates them from, for example, the digital twins for a part that you want to manufacture.

Fleet Collaboration and Swarm Intelligence

Michael Krigsman: So Anthony Scriffignano is asking about collaboration, thinking among the robots. So he's thinking, for example, drone swarms. Can you talk about that aspect of it?

Burkhard Boeckem: That's something that is essential, working in fleets, collaborating robots to robots. Drone systems, in swarms, as I said, coming back to the, for example, autonomous drones, when there's communication or basically autonomous vehicles in mines, yeah, the communication between the robots, the autonomous systems is essential.

And it's essential for fulfilling the task sometimes. It's also essential for learning together in terms of federated learning. If one robot is training something new, then basically you want the whole fleet to have the skills as well.

What Boards Get Wrong About Robotics Investments

Michael Krigsman: Let's talk about some of the business implications here. What do boards of directors typically misunderstand about investments in robotics and physical AI?

Burkhard Boeckem: I think many boards overestimate the speed and underestimate the system work. It's not a software rollout, it's a complex engineering system.

That's what also we saw from when we talked about the regulation from digital AI to physical AI. It's safety engineering, it's operation change, it's integration and service and maintenance, and you have to think the whole cycle through in order to execute on a robot investment.

Michael Krigsman: If you were advising a board considering a major robotics investment, what is the one thing you would insist that they understand before approving it?

Burkhard Boeckem: I would insist on clarity about the job to be done and the risk envelope. I think you clearly need to know the task, the environment, safety constraints, and what happens when confidence drops, and I would also insist they treat data and digital twin readiness as a prerequisite.

If you do not have a reliable picture of the reality you're in, you're probably funding a perpetual pilot.

The Future of Work: Humans and Robots Together

Michael Krigsman: Robotic efficiency or robot efficiency can mean job losses and displacements. Can you talk about the future of work in an increasingly robotic and humanoid robot world?

Burkhard Boeckem: In an ideal scenario where we currently work towards is the collaboration between humans and robots.

And we would like the humanoids to do things that are quite mundane, that are tedious, that are dangerous, that are risky, going into situations, whether it's mining or in the nuclear, where humans should not go.

So basically, the working together, and in this scenario, I would say we upskill the humans to a different level.

And I just read an article about a famous roboticist and said, "I'm optimistic about robots, but they should always work together with humans," and I believe in this as well.

Responsible Deployment of Robots

Michael Krigsman: What does responsible deployment of robots therefore look like?

Burkhard Boeckem: For me, it means safety by design. You need the validation, the traceability, the fail-safe. You need to have the audits done, clear handover rules, and this is especially crucial in what I said before, especially in the mixed human-robot environments.

And it also means, for me, human-centered integration, training, workflow redesign, and the accountability to determine whether robots create value or maybe chaos in here.

Do We Need Manager AIs to Oversee Worker AIs?

Michael Krigsman: Arslan Khan asks the question: Do you see a time where robots become, physical AI becomes more and more autonomous, and will we need a manager, a manager AI that checks if worker AIs are supposed to be doing what they are tasked to do?

Burkhard Boeckem: Probably at some levels, yes. But I would see that ultimately you should have somewhere a human in the loop. That is unavoidable, whether it's then a management AI or basically it's a management AI copilot that helps a human to watch over a fleet.

Yet to be seen, but I think the layering stacks of AI probably makes a lot of sense.

Looking Ahead: The Next 2-3 Years

Michael Krigsman: Where do you see all of this going over the next 2 or 3 years?

Burkhard Boeckem: That enterprises deploy more and more autonomy in constrained domains. Inspection into logistics, picking, placing, handling, site monitoring, and assisted manipulation will scale, in brackets, when the environments are mapped, or you have digital twins, and the success metrics are really clear.

What I would say is a little bit far away and probably still out of reach, is general purpose solving everything, complete human-level dexterity, reliable operation in chaotic spaces, and so forth.

Because probably, even if the technology would be there, it would probably be hard to get the safety level and the safety certification right in this environment. So probably the latter one is the hardest engineering problem.

So, personally, I feel the so-called butler humanoid is probably still a way away.

Core Technical Challenges

Michael Krigsman: Are there one or two core technical challenges that you are very much interested in and focused on right now?

Burkhard Boeckem: If you have an accurate digital twin in order to do the training right, in order to close the simulation to reality gap as narrow as possible, that's one of this.

Then basically, the clarity on the use case and the clarity on what is out of scope currently, that's one other thing.

And then for me, it's also what are the right sensors that I put on a humanoid in order to fulfill all these tasks, and yet to have it at a cost level that is creating an economic value and a sufficient ROI in order to see the investment go through?

Conclusion

Michael Krigsman: Well, this has been such a fascinating conversation. Burkhard Boeckem is CTO of Hexagon. Burkhard, thank you so much for being with us. It's been a really interesting conversation.

Burkhard Boeckem: Thank you so much, Michael, for having me. Thank you.

Michael Krigsman: And everybody, thank you for watching. Now, before you go, now is the time to go to cxotalk.com and subscribe to our newsletter so we can keep you up to date.

And soon, we're going to be starting to send out summaries of our episodes on our newsletter, so check it out.

We have awesome shows coming up, and everybody, we'll see you again next time. You guys are so smart. I always say that because it's really, really true. Take care, everyone.