Best of LinkedIn: Smart Manufacturing CW 09/ 10

00:00:00: This episode is provided by Thomas Allgaier and Frennus based on the most relevant LinkedIn posts about smart manufacturing in calendar weeks nine and ten.

00:00:08: Frenness, a B-to-B market research company that supports enterprises across the smart manufacturing industry with the market customer and competitive insights they need to navigate dynamic markets and drive customer centric product development

00:00:22: right?

00:00:23: And our mission today really just cut through noise.

00:00:26: we're surfacing absolute top smart manufacturing trends that are dominating the conversation across LinkedIn right now.

00:00:34: Yeah, we're looking at a really highly curated stack of insights from industry leaders over just the past two weeks mapping out exactly where the factory floor is actually heading.

00:00:43: Exactly so if you're managing a shop floor or because the current landscape of manufacturing tech I mean it's moving incredibly fast.

00:00:53: but what really stands out in this week?

00:00:54: data

00:00:58: theoretical hype to rigorous practical application, we are finally seeing that maturation.

00:01:05: Especially when major focal point, which is industrial AI and data.

00:01:11: Right?

00:01:11: The dialogue is noticeably pivoted.

00:01:13: it's moving away from running these flashy isolated pilot programs just to say you know hey our facility has AI.

00:01:19: now the focus is entirely on scalable shop floor orchestration.

00:01:23: in a perfect illustration of that maturity actually came from Jerry Brown and Nick Lieder over on the trend detection podcast.

00:01:30: they highlighted this recurring failure mode in the industry.

00:01:32: Oh!

00:01:32: The FOMO issue.

00:01:33: Yes exactly Industrial AI initiatives consistently stall out when they're driven purely by the fear of missing out.

00:01:40: You see these engineering teams adopting this, uh, AI first mindset where

00:01:45: they procure a shiny new model and then basically wander around the shop floor hunting for a problem to solve with it right?

00:01:51: And Jerry in next point is that Ai only acts as an operational accelerator.

00:01:56: When you reverse that equation start With a tangible well defined business or process bottleneck.

00:02:02: You see that constantly, don't you?

00:02:04: A facility will try to slap some advanced predictive maintenance algorithm onto a machine center where the underlying maintenance processes are already completely

00:02:15: broken.

00:02:16: And

00:02:21: speaking of broken processes, Allison brought up a massive architectural bottleneck this week regarding how we actually handled data across all these systems.

00:02:29: Yeah he warned against the industry obsession with integrating CAD PLM and MES systems by physically moving data files around

00:02:37: The classic ETL trap extract transform load You design a product in your CAD software, engineering pushes it to Product Lifecycle Management and then manufacturing needs those specs on the Manufacturing Execution System.

00:02:50: And the traditional instinct there is to build a pipeline that just copies data from one box to next

00:02:55: Which sounds completely logical until you realize what absolute nightmare it creates on the floor.

00:03:00: Allison calls this Data Drift

00:03:02: Data drift.

00:03:03: Yeah, because you end up extracting parameters maybe pasting them into a spreadsheet to bridge the gap and suddenly

00:03:08: Suddenly you don't have a single source of truth anymore.

00:03:11: You have four conflicting versions of the exact same product spec floating around.

00:03:15: exactly true digitalization means your systems intelligently interact with one central data set without ever copying it.

00:03:23: right when the MES needs Tolerance spec, it shouldn't look at a copied file.

00:03:27: It should query the PLM in real time pulling exactly what it needs for that specific operational second

00:03:34: and The latency of those system interactions brings up another really critical debate right now.

00:03:39: Which is where this AI computation actually lives physically.

00:03:42: Dan Rodriguez made a fantastic argument about edge versus cloud computing In the context of what he calls physical AI.

00:03:49: He used the example of a factory robot needing to execute his safety stop which perfectly frames the problem.

00:03:56: It really does because if that robot has to ping a cloud server, wait for the data payload to travel to a centralized datacenter and then wait for break command to return

00:04:06: The collision is already happened.

00:04:07: it's too late

00:04:08: Right.

00:04:09: Real-time actions must live locally at edge.

00:04:12: A millisecond delay in physical manufacturing environment means shattered spindle or worse severe safety incident.

00:04:20: But Edge hardware can only handle localized immediate logic.

00:04:24: It doesn't have the compute for macro level analysis, which

00:04:27: is where the hybrid architecture comes into play.

00:04:30: once every machine cell is running intelligent edge logic you need macro-level coordination across the entire fleet

00:04:37: right?

00:04:37: The cloud handles the heavy lifting of long term machine learning analyzing years of predictive maintenance data and scaling those insights across multiple global facilities.

00:04:46: so Edge Is For Real Time Survival And Cloud Is For Long Term Evolution

00:04:51: Perfect way to put it.

00:04:52: But even with the perfect edge-to-cloud infrastructure, The AI itself can still fail if it doesn't understand what its actually looking at.

00:05:01: Kudzu Amanda Therese shared an insight from Zach Eater that challenges a huge industry assumption here.

00:05:06: Yeah!

00:05:07: Eater argues that the real bottleneck for deploying reliable AI agents on the shop floor isn't computing power and It isn't even the volume of data.

00:05:15: Its context architecture

00:05:17: Right Because so many facilities assume that simply dumping millions of rows of raw sensor telemetry into a data lake will automatically yield a smarter AI.

00:05:26: But without context, it's just digital noise!

00:05:29: I have to play devil's advocate for a second here.

00:05:31: Isn't context engineering just a fancy new buzzword?

00:05:34: For basic data cleaning, we've been doing ETL for decades!

00:05:38: No it actually goes much deeper than just cleaning formatting errors.

00:05:41: Context Engineering is about structuring highly relevant domain specific knowledge so that AI can actually reason.

00:05:48: Ah okay... So like.. A temperature spike on a milling machine means something entirely different if the machine is roughing block of titanium versus taking finishing pass aluminum

00:05:56: Exactly.

00:05:57: If you want an AI agent to actually optimize throughput it needs the precise operational state, material properties and machine history at that exact moment.

00:06:07: It requires curated reasoning not just a plain data dump.

00:06:11: And that level of sophisticated reasoning really sets the stage for the massive industry shakeup highlighted by Colin Masson.

00:06:18: He broke down the implications of the fifty billion dollar pact between AWS and open AI,

00:06:23: which is huge!

00:06:24: This isn't just a standard corporate partnership.

00:06:27: it's a structural shift from manufacturing giving industrial clients access to stateful AI on platforms like Amazon Bedrock.

00:06:34: Stateful AI Because traditional algorithms treat every query as an isolated event.

00:06:40: But stateful AI actually maintains memory and context across long-running, highly complex operations.

00:06:46: Think

00:06:46: about it in the context of a severe supply chain disruption.

00:06:49: A Stateful AI agent remembers critical parts shortage that you navigated three weeks ago

00:06:54: Right!

00:06:54: And correlates to past events with say...a port strike happening today Proactively reroutes your billed materials

00:07:00: Without requiring human planner.

00:07:02: sit there and feed the entire historical context all over again.

00:07:06: It bridges the gap between these massive language models and real-world physical supply chain execution,

00:07:13: but All of this software level reasoning is effectively useless if The physical factory floor is too rigid to execute the pivots

00:07:20: which leads us directly into our next major focus Software defined factories in digital twins.

00:07:26: because modern supply chains operate In such a VUCA environment

00:07:30: volatile uncertain, complex and ambiguous.

00:07:34: Right because of that the physical factory itself has to become just as adaptable is the code that's running it.

00:07:40: Gwennale Avis Huey laid out the reality that the era of closed hard-coded factory systems is officially ending.

00:07:46: Historically production lines move slowly enough And product life cycles were long enough That you could afford to bolt down machinery and lock yourself into rigid proprietary hardware configurations.

00:07:56: But that static approach Is a death sentence today.

00:07:59: Software-defined automation abstracts the control logic away from physical hardware.

00:08:04: It allows manufacturers to adapt production runs, plug in new diagnostic technologies and scale up capabilities through software updates

00:08:12: Rather than ripping out and replacing physical infrastructure that still mechanically works perfectly fine.

00:08:17: And

00:08:17: before you push those software updates into a physical floor You need know they aren't going cause massive bottleneck.

00:08:23: David Morley shared a stunning real-world application of this from PepsiCo.

00:08:27: Yes, PepsiCo is heavily utilizing Siemens Digital Twin Composer and NVIDIA Omniverse to completely virtualize their US manufacturing facilities.

00:08:36: They aren't just optimizing isolated machine cells either.

00:08:39: they are building high fidelity three D digital twins of entire plants.

00:08:44: The traditional alternative to that is incredibly costly.

00:08:47: if you want change in material flow You physically move conveyors run a test batch measure the downtime into hope.

00:08:53: the new configuration actually works.

00:08:54: But PepsiCo is simulating the entire material flow virtually by running AI-driven scenarios on a three D digital twin before a single wrenches turn, they're catching up to ninety percent of potential operational issues

00:09:07: Simulating an entire Material Flow Before Any Physical Work Starts.

00:09:12: That Is Incredible And The Financial Impact Of That Is Staggering.

00:09:16: They Are Reporting A Twenty Percent Increase In Throughput

00:09:19: and Up To A Fifteen Percent Reduction in Capital Expenditure.

00:09:22: They validate line designs and automation routing with near-perfect confidence before authorizing any real capital deployment.

00:09:29: And this predictive capability totally changes the role of our foundational software tools.

00:09:34: Jakob Alorek Kassal addressed his growing anxiety in industry regarding MES manufacturing execution systems.

00:09:41: Yeah, there's a narrative going around that AI is going to render traditional MES obsolete

00:09:46: But he argues the exact opposite.

00:09:48: MES combined with AI is future standard.

00:09:51: For decades, an MES has essentially functioned as a rear-view mirror.

00:09:55: It tracks work orders logs quality defects and basically tells you exactly what went wrong on the shift yesterday.

00:10:00: but injecting AI into that system turns it in to GPS stops logging hindsight starts generating predictive foresight

00:10:07: alerting operators that specific spindle vibration pattern indicates equality defect will likely occur next.

00:10:14: forty five minutes.

00:10:15: Jan Berner, from eDags expanded on this execution-focused mindset when discussing the industrial metaverse.

00:10:22: He's heavily involved in a platform called Metis and he makes it really crucial distinction here

00:10:26: which is that the Industrial Metaverse is heavily misunderstood as just being pretty three D visualization

00:10:31: right?

00:10:32: It's not just an expensive video game for process engineers to fly around.

00:10:35: Platforms

00:10:36: like MEDIS are designed to be rigorous execution frameworks.

00:10:40: They orchestrate these digital twins, the simulation environments and AI decision engines into one cohesive system.

00:10:47: The goal is dramatically reduce investment risk and validate every physical action in virtual space before any steel cut on a shop floor.

00:10:56: So if software logic is mapped, the virtual twin has validated material flow?

00:11:02: Now we have to look at the physical muscle executing these commands.

00:11:06: Which brings us to robotics, automation and its concept of operator five point.

00:11:10: oh

00:11:10: The sheer scale.

00:11:12: a physical hardware deployment happening right now is difficult even comprehend.

00:11:16: Jake Hall shared some data highlighting that over four point six million industrial robots Have been deployed globally.

00:11:23: Four point six Million.

00:11:25: But truly disruptive metric there Is shift in global manufacturing dominance.

00:11:31: Chinese automation manufacturers like Eston Automation and Buronte are now producing industrial robots at a volume that rivals legacy global giants, like FANUC.

00:11:41: That represents a massive deployment of physical automation capacity though it's primarily being integrated directly back into chinese domestic facilities right now.

00:11:48: but the rush to scale automation brings a massive risk.

00:11:51: no matter where you Christchurch, you offered a really sobering reality check about the white elephant on the shop floor.

00:11:57: Oh I love this point!

00:11:58: He frequently sees high-budget automation projects fail miserably because the underlying manufacturing process wasn't fully defined or brought under control before robotic arms were purchased.

00:12:09: You cannot automate chaos If a facility has severe variance in its casting dimensions Or highly unpredictable assembly sequence.

00:12:17: handing that process over to robot won't solve the issue.

00:12:21: If you automate a broken process, congratulations!

00:12:24: You have just spent millions of dollars in capex to manufacture scrap parts at record speeds.

00:12:29: Exactly management often views robotics as a magic bullet for labor shortages or quality issues.

00:12:35: but automation requires a foundation of stringent process control To actually generate an ROI.

00:12:41: however when the processes controlled The efficiency gains are undeniable.

00:12:45: Tobias Claus highlighted a highly specific successful use case involving Karkin's Autonomous Material Transport Robots.

00:12:52: Right, ADEC Inc.

00:12:53: deployed these mobile robots specifically to handle internal logistics automating the repetitive low value task of building-to-building material transport.

00:13:01: and The metric that really stands out there is the human time recovered.

00:13:04: They saved roughly three hours per employee per day while simultaneously smoothing outline interruptions.

00:13:10: giving an employee Three hours back in their shift changes the entire nature of their job

00:13:14: which connects directly to the evolving role of The Human Worker.

00:13:18: Moosarot Hussein introduced a vital concept that frames this transition perfectly, operator five point oh

00:13:25: because for The dream of the dark lights out factory run entirely by algorithms and machines, but

00:13:36: the pendulum is finally swinging back.

00:13:38: Industry five point.

00:13:39: Oh it's fundamentally about human centric collaboration.

00:13:43: We aren't building terminators to replace the workforce we're building iron man suits to augment them.

00:13:49: operator Five point oh work safely.

00:13:50: alongside collaborative robots co-bots that handle a heavy repetitive lifting in dangerous material handling.

00:13:57: The human worker is augmented with wearable technology, augmented reality displays for complex assembly and AI-driven decision support.

00:14:05: Because the machines are handling the massive data processing in physical fatigue.

00:14:09: it frees up human cognitive value from what algorithms still struggle

00:14:12: With anomalous problem solving critical Decision making an adapting to completely unforeseen production challenges.

00:14:20: you integrate Human ingenuity with machine efficiency And that is the actual definition of a smart

00:14:26: factory.

00:14:26: But of course, the moment you connect all these software-defined twins, cloud AI agents and mobile co-bots.

00:14:32: You massively expand your operational attack surface

00:14:35: which brings us to our final area focus cyber security and massive regional investments.

00:14:41: as factories evolve into highly connected software environments The stakes for security have escalated from data loss to physical catastrophe.

00:14:49: Erdem Ozturk provided a critical update on this shift featuring insights from Siemens Mark Koons regarding the EU Cyber Resilience Act.

00:14:56: The era of treating cybersecurity as an IT afterthought, you know something you patch over after the machine is installed...is over

00:15:03: Under these new frameworks?

00:15:05: Cybersecurity's becoming a core engineering architecture requirement for Machine Tools.

00:15:10: It's mandated with the exact same regulatory rigor as physical safety interlocks.

00:15:15: Tobias Walkman illustrated exactly why this regulatory shift is so necessary, with some very practical advice aimed at glass manufacturing plants.

00:15:25: Because historically the IT networks that handled emails and operational technology were air-gapped completely separate

00:15:34: but today they are deeply linked.

00:15:36: Washman warned that a cyber incident on the IT side can now rapidly cascade into the OT environment, affecting machine availability and physical safety.

00:15:45: And in a glass plant... ...a network going down isn't just an inconvenience.

00:15:49: If control systems freeze and molten glass is allowed to cool & solidify inside production equipment

00:15:55: You aren't just facing downtime you are facing catastrophic physical destruction of facilities entire infrastructure.

00:16:02: Watman stressed reliable backups Rigorously separated critical networks and trusted software supply chains are now non-negotiable survival requirements.

00:16:11: But securing and building these resilient, highly automated environments requires an unprecedented volume of capital.

00:16:17: Erdal Aroglou pointed out that major players are aggressively funding this transformation.

00:16:22: Siemens for example is investing over two hundred million euros into their amber site to build a highly automated AI driven smart factory.

00:16:31: They are essentially building the global blueprint for what a secure, digitalized manufacturing facility looks like at scale.

00:16:38: Two hundred million euros is a massive commitment for single site.

00:16:42: but if we zoom out to the macro level Robert Little broke down a regional transformation that dwarfs those numbers.

00:16:49: The Great

00:16:49: Texas Industrial Rebuild!

00:16:51: The Scale of Capital Deployment Happening in Texas alone Is Staggering Totalling A One Hundred and Eighty Nine Billion Dollar Manufacturing Transformation.

00:16:59: It's a complete re-architecting of the domestic supply chain.

00:17:02: You have Samsung building a forty five billion dollar leading edge logic cluster in Taylor.

00:17:07: Texas Instruments is constructing a forty billion dollar high volume analog and power chip fabrication plant in Sherman,

00:17:14: And you have Tesla's Gigatexus facility in Austin scaling with ten billion dollars investment.

00:17:19: And Gigatexus isn't just bending sheet metal.

00:17:22: it houses their Cortex AI supercluster alongside primary assembly lines for their Optimus humanoid robots.

00:17:28: They are co-locating the massive data center brains needed to train physical AI directly next to manufacturing lines, building robotic muscle.

00:17:39: The geopolitical and supply chain implications of having that density of advanced manufacturing concentrated in one state will resonate for decades.

00:17:51: Which really leaves us with a fascinating, highly complex puzzle as we look at the global landscape.

00:17:56: Yeah As Johannes Radle pointed out his analysis... machine tool integration, and rigorous compliance frameworks.

00:18:10: Texas and the U.S are rapidly building the massive AI brains —the hyperscale data centers—and the Silicon

00:18:16: Foundations.".

00:18:17: Meanwhile China is churning out physical industrial robots in automation hardware at an unprecedented mass scale.

00:18:24: so... The big question for you to mull over as you analyze your own operational roadmap.

00:18:29: In an era of increasing geopolitical tension and deliberate supply chain decoupling, how will the Smart Factory Of The Future successfully assemble these critical pieces?

00:18:59: Thank you so much for joining us on this deep dive.

00:19:02: Make sure to hit subscribe on your podcast feed,