Best of LinkedIn: Smart Manufacturing CW 03/ 04

00:00:00: This episode is provided by Thomas Allguyer and Frenas, based on the most relevant LinkedIn posts about smart manufacturing in calendar weeks O three and O four.

00:00:09: Frenas is a B to B market research company that supports enterprises across the smart manufacturing industry.

00:00:15: with the market, customer and competitive insights they need to navigate dynamic markets and drive customer centric product development.

00:00:23: Good to be back.

00:00:24: And yeah, we're looking at a very specific window here, calendar weeks three and four of twenty twenty

00:00:31: six.

00:00:31: And you know, looking at the sources for these two weeks, it really feels like the vibe has shifted.

00:00:36: Oh,

00:00:36: definitely.

00:00:36: For the

00:00:37: last couple of years, everything was about future tech, right?

00:00:40: The shiny objects, the what ifs.

00:00:42: But digging into this feels like the industry has, I don't know, collectively sobered up.

00:00:47: The conversation isn't what is this tech?

00:00:49: anymore.

00:00:50: It's how do we actually operationalize this?

00:00:52: so it pays the bills?

00:00:53: It's a really distinct move from the pilot phase to let's call it the integration phase.

00:00:58: You see it everywhere we're going to look today.

00:01:00: Industrial AI, digital twins, robotic.

00:01:02: People aren't really talking about experiments in the lab anymore.

00:01:05: They're talking about infrastructure and talent and global strategy.

00:01:10: Well, let's start with the heavyweight in the room then.

00:01:12: Industrial AI.

00:01:15: But again, the angle feels different.

00:01:16: We're not seeing posts just hyping up some new model.

00:01:19: No.

00:01:19: It's a much more pragmatic debate about infrastructure and what's being called a agentic AI.

00:01:27: And this is where Dr.

00:01:28: Monica McIterian really set the tone.

00:01:30: She put forward this, well, provocative argument that the AI models themselves, she referenced Deep Seek and Laume, are basically becoming commodities.

00:01:39: So if the brain is a commodity, where's the edge?

00:01:41: Infrastructure efficiency, that's her whole point.

00:01:43: Yeah.

00:01:44: Mictarian shared a data point that honestly should worry a lot of CTOs.

00:01:48: Okay.

00:01:48: Most enterprise AI stacks are only delivering about fifty to sixty percent of their theoretical performance.

00:01:54: Fifty to sixty, that's a staggering inefficiency.

00:01:56: Yeah.

00:01:56: I mean, if a production line were running at fifty percent capacity, heads would roll.

00:02:00: Exactly.

00:02:01: Why is it acceptable here?

00:02:03: And she says it's because of a lack of hardware software co-design.

00:02:06: You've got companies running these super advanced models on totally unoptimized architecture.

00:02:10: Wow.

00:02:11: But the real kicker in her analysis was about latency.

00:02:13: She stated it so clearly.

00:02:15: Latency defines UX in enterprise AI.

00:02:19: Meaning, if I'm a technician on the floor and I ask the AI a question.

00:02:23: and I have to wait five seconds.

00:02:25: You

00:02:25: stop using it, you go ask Bob at the next station.

00:02:27: If the latency is high, adoption just crashes.

00:02:30: Which brings us right to the shop floor.

00:02:32: Yeah.

00:02:33: Mammal Costa shared this great example that tackles this exact UX problem with Joule.

00:02:39: from SAP.

00:02:40: Yeah, this looks like that digital co-pilot concept finally hitting the ground.

00:02:44: It really does.

00:02:44: The use case he describes is all about removing the interface barrier.

00:02:48: So instead of a worker having to navigate through, you know, all these complex ERP menus to check in order, they just use natural language.

00:02:55: It turns a database query into a simple conversation.

00:02:57: And that's

00:02:58: so critical.

00:02:59: The people on the floor are experts in manufacturing, not in navigating SAP.

00:03:03: Hmm, but it's not just for the operators.

00:03:05: We saw a really similar thread for the engineers.

00:03:08: Flavio Arsani was talking about the process simulate AI co-pilot.

00:03:12: Okay, so this is the engineering equivalent.

00:03:14: Exactly.

00:03:15: Arsani calls it a force multiplier for Canadian manufacturing engineering.

00:03:20: Just think about the time an engineer spends menu diving, trying to find a specific tool to inventory some robots or generate collision pairs.

00:03:28: It's that death by a thousand clicks.

00:03:29: That's it.

00:03:30: This co-pilot just removed that friction.

00:03:33: You just ask for the collision pairs and the system generates them.

00:03:36: It keeps the engineer in that flow state instead of fighting the software.

00:03:40: I have to say though I really appreciated that amidst all this AI can do everything talk.

00:03:45: We got this.

00:03:46: Strong reality check from Thomas Ripplinger.

00:03:49: Oh, he's checklist.

00:03:50: Yeah.

00:03:50: He dropped a checklist for agentic AI readiness that I think every manager needs to like tape to their wall.

00:03:55: It was so grounded, wasn't it?

00:03:57: I mean that term agentic AI gets thrown around a lot implying AI that takes action not just generates text.

00:04:03: Right.

00:04:04: But Ripplinger's core point was you can't just deploy agents into a broken process.

00:04:09: His rule about manual expert solutions really stood out.

00:04:12: He basically said not every problem needs AI.

00:04:16: And that's the core of it.

00:04:17: Before you try to automate a decision with an AI agent, you need a really well defined problem and a manual solution that actually works.

00:04:25: If you don't know how to solve it manually, the AI is not going to figure it out for you.

00:04:29: You'll

00:04:29: just automate the chaos.

00:04:30: And

00:04:30: Sebastian Schmitz backed that up from a different angle.

00:04:33: He pointed out that when these projects fail, it's rarely because the code was bad.

00:04:38: No, it's organizational, it's culture, it's training.

00:04:41: It's a lack of openness.

00:04:43: Which is a perfect segue into our second theme, the digital twin.

00:04:47: Because nowhere is that cultural and organizational gap more visible.

00:04:51: You're talking about what Muserat Hussein calls pilot purgatory.

00:04:55: Exactly.

00:04:55: Pilot purgatory.

00:04:56: We hear that term all the time.

00:04:58: But he was actually diagnosing the cause, not just the symptom.

00:05:01: He pointed specifically to the ITOT divide.

00:05:04: This is the classic struggle, right?

00:05:05: You have operations technology.

00:05:07: your OT machines speaking these old languages like Modbus, and they're trying to talk to the information technology, the IT that lives in modern cloud APIs.

00:05:17: Hussein argues the skills gap there is just massive.

00:05:20: We need what he calls industrial data scientists who can speak both languages.

00:05:25: You had that biting quote.

00:05:26: A digital twin without connected data is just an expensive three-D drawing.

00:05:31: It's harsh, but it's accurate.

00:05:33: So many companies pay for the cool visualization software before they've even solved the data pipeline problem.

00:05:39: When you do get it right, when the data is flowing, the potential is just undeniable.

00:05:44: We saw Bettina Rotemund and Matthias Heinecke talking about the Siemens digital twin composer.

00:05:50: And this is where we move from just monitoring to validating, right?

00:05:53: Correct.

00:05:53: The value proposition they discussed is being able to validate entire factory layouts or robotic behaviors before anything physical is built.

00:06:01: It's all about risk mitigation.

00:06:03: And Chevrolet framed this so well, talking about Swiss manufacturers, she said they fail fast digitally, succeed once physically.

00:06:09: That's the economic argument for the twin right there.

00:06:11: Yeah.

00:06:12: It is infinitely cheaper to crash a digital robot than a real one.

00:06:15: And speaking of the tech that enables this, Johannes Kallchaffer highlighted Cinematic One.

00:06:20: This is that digital native CNC system.

00:06:23: Yeah.

00:06:24: And the distinction is really important.

00:06:26: It's not just a controller on a machine.

00:06:28: The system itself actually powers the digital twin.

00:06:31: So it knows what's going to happen before it happens.

00:06:34: Precisely.

00:06:35: It allows for better surfaces, faster machining, because the machine knows its own dynamics before it even starts cutting metal.

00:06:42: The digital and physical are completely fused.

00:06:45: And then at the ecosystem level... Ryan Suerte flagged the Siemens and NVIDIA partnership.

00:06:51: They're basically building an industrial AI operating system.

00:06:54: And he dropped a stat from PepsiCo that I found hard to believe.

00:06:57: The twenty percent throughput, Jane.

00:06:59: Yeah,

00:06:59: twenty percent.

00:07:00: In the world of high volume manufacturing, FMCG, a twenty percent gain is astronomical.

00:07:05: You usually fight for one or two percent.

00:07:07: So that

00:07:07: really proves that when you move past that expensive, three-D drawing phase and actually integrate the simulation with AI, the ROI is there.

00:07:15: It's absolutely there.

00:07:16: Okay, so we've got the brain with AI, we've got the simulation with the twin, let's talk about the hands, robotics and automation.

00:07:23: Jeff Winter had a fantastic analogy for this to explain the evolution.

00:07:27: he compared it to snow removal.

00:07:29: I like this one.

00:07:30: It strips away all the jargon.

00:07:31: It does.

00:07:32: So industry.

00:07:33: one point oh is the shovel.

00:07:34: Pure manual effort.

00:07:36: Industry.

00:07:36: two point oh.

00:07:37: and three point oh.

00:07:37: That's the snow blower.

00:07:39: It's mechanized.

00:07:40: It's faster.

00:07:41: But you're still walking behind it guiding it.

00:07:43: So what's industry?

00:07:44: four point oh?

00:07:44: That's the autonomous machine clearing your driveway while you sit inside with a coffee.

00:07:48: So the task moving snow, it hasn't changed.

00:07:51: The physics haven't changed.

00:07:52: But our relationship to the task has completely flipped.

00:07:56: Exactly.

00:07:57: But even the smartest autonomous robot needs to interact with the world.

00:08:02: And Mark Dutonstetter wrote this sort of, I don't know, a love letter to the unsung hero of robotics.

00:08:08: The

00:08:08: gripper?

00:08:08: The gripper, the end effector.

00:08:10: It's funny, we obsess over the robotic arm, but Dutonstetter points out that the gripper is where the actual work

00:08:16: happens.

00:08:17: And the market reflects that, right?

00:08:18: He notes, it's expected to exceed three billion dollars by twenty thirty.

00:08:22: And the complexity is just in the variety.

00:08:25: A gripper for, say, a soft bakery item needs completely different physics than one handling a heavy automotive door panel.

00:08:34: If the gripper isn't smart, the whole multi-million dollar robot is useless.

00:08:38: Completely useless.

00:08:39: Speaking of smart robots, though, we have to talk about the humanoids.

00:08:42: Roland Locker had some insights that went beyond the usual viral videos of robots doing backflips.

00:08:47: He focused on the deployment reality.

00:08:49: He did.

00:08:50: These humanoids are moving from the lab to the shop floor, but you can't just draw.

00:08:54: them in.

00:08:55: He emphasizes using simulations specifically, process simulate to validate safety and ergonomics first.

00:09:03: Because unlike a standard industrial robot that's bolted down in a cage, these things are walking around next to people.

00:09:08: Exactly.

00:09:09: If a humanoid trips or drops a payload, that's a serious safety incident.

00:09:13: You have to simulate all those failure modes digitally first.

00:09:16: It connects right back to that fail fast digitally idea.

00:09:19: And on the topic of mobile robots, Alexander Zalhoffer highlighted the IW.

00:09:24: This seemed less about replacing humans and more about augmenting a workforce that just isn't there.

00:09:30: That is the key driver, especially in the German market right now.

00:09:34: He points out these mobile robots are addressing acute labor shortages.

00:09:38: It's not about efficiency in the abstract.

00:09:40: It's about keeping the lines running when you literally can't hire enough staff.

00:09:44: We also saw this convergence of robotics with other technologies.

00:09:49: Daniel Cooper was talking about using six-axis robots for

00:09:53: Large format additive manufacturing.

00:09:55: A standard, three-D printer is limited by its gantry.

00:09:58: But if you put a printhead on a six-axis robot, suddenly you have incredible reach, dexterity, and scale.

00:10:05: But Michael Finocchiaro added a necessary layer of software intelligence on top of that.

00:10:10: He talked about SelectAM.

00:10:11: Which is so important, it basically answers the question, should I even print this?

00:10:15: Just because you can print a part on a robot arm doesn't mean it makes any economic sense.

00:10:19: So

00:10:19: the software provides the decision support.

00:10:22: It identifies which parts in your inventory are actually viable for additive manufacturing.

00:10:27: It adds that business logic.

00:10:28: So we've covered the tech stack.

00:10:29: I want to zoom out now to the factory level.

00:10:32: How does all this come together in an actual building?

00:10:35: We saw some heavy hitters talking about lighthouse factories this week.

00:10:38: We did.

00:10:39: And there was a fascinating contrast between the two main examples.

00:10:42: First, you had Cedric Nike giving an update on the Siemens Nanjing factory.

00:10:47: The digital native factory.

00:10:49: That's

00:10:49: the one.

00:10:49: The stats are just incredible.

00:10:51: Lead time dropped seventy-eight percent.

00:10:53: Productivity up fourteen percent.

00:10:55: It's a shrine to efficiency.

00:10:57: But then you compare that to Gwennale... have as cue its update on the Schneider Electric factory in Wuhan.

00:11:03: They were also recognized as a lighthouse, but specifically for talent.

00:11:07: And that distinction is so vital.

00:11:09: They used AI not just to speed up the line, but to support their technicians.

00:11:13: And the result, they cut technician turnover from down to six

00:11:19: percent.

00:11:20: That's the metric that stopped me in my tracks.

00:11:21: Yeah.

00:11:22: Forty-eight percent turnover is a crisis.

00:11:24: Six percent is world-class.

00:11:26: It proves

00:11:26: that smart manufacturing is actually a retention strategy.

00:11:30: If you use AI to remove the frustration, the menu diving, the waiting, the confusion people actually want to stay in the job.

00:11:38: Now, moving geographically, we saw some really interesting strategic moves in the U.S Stefan Mayer and Litz LeBisch both posted about Trump beef in Connecticut.

00:11:48: This is the local for local strategy and action.

00:11:52: Trump PF is a German company, but they're building their two machines in the US specifically for the US market.

00:11:58: So

00:11:58: why do that?

00:11:58: Is it just about avoiding tariffs and things like that?

00:12:01: That's part of it for sure.

00:12:02: But it's also about lead time and logistics.

00:12:04: I mean, shipping massive industrial machinery across the Atlantic is slow and expensive.

00:12:09: But here's the catch.

00:12:11: To manufacture in a place like Connecticut where labor costs are high, they have to use lights out automation.

00:12:17: So the automation.

00:12:18: what makes the localization economically viable in the first place.

00:12:21: Exactly.

00:12:21: They need to be able to produce all the way down to a batch size of one flexibly automatically.

00:12:27: If they tried to run that factory with traditional manual workflows, the cost structure just wouldn't work.

00:12:32: But to run a factory that flexible, you need software that can keep up.

00:12:37: Good day, Amanda.

00:12:37: Teresa argues that our old MES platforms are really failing us here.

00:12:41: He's so right.

00:12:42: The legacy MES is just rigid.

00:12:45: If marketing decides to switch from a one point five liter bottle to a one liter bottle, a traditional MES could take weeks of reprogramming.

00:12:54: Wow.

00:12:54: He's advocating for AI native platforms and video SO keys that allow for that kind of flexibility overnight.

00:13:00: Bring Carol the second added to this emphasizing the link between MES and PLM product lifecycle management.

00:13:06: He calls them the heartbeat and the brain.

00:13:09: It's a great analogy.

00:13:10: If the brain, which is the design doesn't talk to the heartbeat, the execution, the whole organism fails.

00:13:16: Eva Sanchez Gil's visit to the SeatSmart factory was a great benchmark for this.

00:13:20: She highlighted how they have successfully integrated that heavy industrial automation with that MES layer.

00:13:26: Before we close that, I just want to touch on the regional insights we saw.

00:13:29: We mentioned China in the US, but Robert Little had a really specific take on India.

00:13:33: Yeah.

00:13:33: Little observed that India is trying a different path.

00:13:36: It's deregulation.

00:13:38: Instead of just subsidies, they're trying to remove the bureaucratic friction to unleash their manufacturing potential.

00:13:44: And Bilal Ahmad gave us a view of Europe, specifically the car manufacturing network across Spain, the Czech Republic, and others.

00:13:52: He describes it as a single distributed factory.

00:13:56: It's not just a collection of isolated plants.

00:13:58: It's a network that acts as one entity.

00:14:01: It really shows that the definition of a factory is expanding way beyond just four walls.

00:14:07: So if we wrap this up, if I look at these two weeks as a whole, what's the big takeaway?

00:14:11: We're not talking about buying tech for tech's sake anymore.

00:14:14: No, not at all.

00:14:15: We're talking about maturity.

00:14:16: The lighthouse examples prove that this works.

00:14:18: The pilot purgatory warnings prove that it's hard.

00:14:21: But the winning companies are the ones treating AI, twins, and robots not as magic wands, but as infrastructure.

00:14:27: They're fixing the plumbing, the data, the culture, the training, so the tech can actually flow.

00:14:32: So the question for you, the listener, really is.

00:14:35: Are you just buying tools to fix a broken process?

00:14:39: Or are you building the infrastructure to support a whole new way of working?

00:14:44: That's the question to ask in the next strategy meeting.

00:14:46: If you enjoyed this episode, new episodes drop every two weeks.

00:14:50: Also, check out our other editions on digital construction and digital power tools.

00:14:53: Thanks for listening.

00:14:55: Don't forget to subscribe.

00:14:56: Until next time.