Best of LinkedIn: Smart Manufacturing CW 25/ 26

00:00:00: This episode is provided by Thomas Allgeier and Freeness based on the most relevant LinkedIn posts about smart manufacturing in calendar weeks, twenty-five and twenty six.

00:00:10: Freenes is a B to B market research company that supports smart manufacturing providers with building feature by feature competitive intelligence that shows exactly how their product stacks up against the competition.

00:00:22: you can find more info.

00:00:25: Yeah, so welcome to The Deep Dive.

00:00:27: We are looking at some seriously fascinating stuff today.

00:00:31: Oh

00:00:31: absolutely

00:00:32: I mean we're talking about the top smart manufacturing trends that have been surfacing all over LinkedIn lately.

00:00:39: and if you're in the construction or manufacturing industry You really need to hear this

00:00:43: because we're basically looking at how the industry is finally moving past.

00:00:47: You know, pilot purgatory.

00:00:48: Right the era of just running these isolated highly controlled experiments Just so you can put them in a slide deck.

00:00:54: that's

00:00:54: over.

00:00:55: it is entirely over.

00:00:56: now It's all about.

00:00:57: well ruthless industrial great execution

00:01:00: exactly.

00:01:02: But um to get to that execution stage you need a totally different mindset.

00:01:06: I mean we're so used to looking at the shiny stuff right?

00:01:08: The AI co-pilots the autonomous robots Yeah the

00:01:12: things that look great on video.

00:01:14: But what the top voices are actually focusing on right now is the plumbing.

00:01:17: Deploying, yes because you simply cannot deploy advanced AI.

00:01:23: if

00:01:23: your underlying data architecture is a total mess...

00:01:26: It just won't work!

00:01:27: No and Watter van Hedigem shared this really clear six-level roadmap about exactly that.

00:01:35: Manufacturers basically have to establish connected systems and really heavily govern data first.

00:01:40: Before they even think about scaling AI?

00:01:42: Exactly, he made the point that AI isn't the starting point of a digital transformation.

00:01:46: it's actually the final layer that just sits on top of perfectly orchestrated data foundation

00:01:51: Which makes sense.

00:01:52: but uh... The challenge there is that manufacturing data is famously difficult to orchestrate because it lives in two completely different worlds.

00:01:58: you know

00:01:59: Oh!

00:01:59: The classic OT versus IT battle.

00:02:02: Yes You've got operational technology, the OT down on the factory floor and then information technology IT up in the corporate office.

00:02:11: And historically these two sides kind of hated each other...

00:02:14: They did!

00:02:14: I mean they just have completely different goals.

00:02:16: O.T is deterministic

00:02:18: Right meaning if a robotic arm is told to stop it has to stop exactly ten milliseconds

00:02:24: Not fifteen not fifty

00:02:25: Yeah because any leg there means broken machine or worse serious injury.

00:02:31: So OT networks were historically air-gapped, like literally unflugged from the outside world for safety.

00:02:38: Right

00:02:38: while IT on the other hand is built entirely around accessibility and cloud storage.

00:02:43: a two second delay in sending an email doesn't break anything.

00:02:46: Exactly

00:02:46: so bringing those to opposing philosophies together as basically the main bottleneck right now.

00:02:52: And Pavan Pusiluri recently broke down this four layer architecture that actually bridges that gap.

00:02:57: Yeah The unified namespace concept

00:03:00: The UNS, because usually factories just use point-to-point connections.

00:03:04: An inventory app talks to a specific conveyor belt A quality control app talks with camera

00:03:08: Which is a nightmare.

00:03:09: to manage

00:03:10: Total Nightmare It's like having fifty different translators shouting across the crowded room.

00:03:16: But...a unified namespace replaces all of that.

00:03:19: it acts as this single central bulletin board.

00:03:23: So every device and sensor just publishes its data to that one hub?

00:03:27: Exactly.

00:03:27: And any application that needs the data, it's a massive structural shift.

00:03:33: Because without this single source of truth your machine learning models are essentially starving.

00:03:38: I mean if you're data scientists they still begging for manual CSV exports like it was in two thousand

00:03:44: nine.

00:03:44: Oh man!

00:03:44: The CSV export

00:03:45: Right.

00:03:46: Your AI is dead on arrival at that point.

00:03:48: But with a UNS running, you can finally put intelligent models where they actually belong.

00:03:52: Like EDGE ML?

00:03:53: Yeah!

00:03:54: Poo Saluri really emphasized EDGEML-running machine learning right next to the physical asset... ...to make those critical millisecond decisions while Cloud ML can kind of sit back digest months of historical data and find the big optimization patterns.

00:04:07: And the really crucial part about all that architecture is feedback loop.

00:04:11: Oh totally

00:04:11: If that EDGE model detects an anomaly That insight cant just die on a dashboard, it has to flow back up the IT system and actually trigger maintenance work

00:04:21: order.

00:04:22: And flowback to the model to retrain it?

00:04:25: Otherwise its not.

00:04:25: an intelligent system is just really expensive.

00:04:27: alarm clock

00:04:28: Just an Alarm Clock.

00:04:29: I love that!

00:04:30: An alarm clock doesn't give you context which ties perfectly into what Maria Karolina-Ralho said at The Connected Manufacturing Forum.

00:04:39: Right...the data versus context problem.

00:04:41: She pointed out that manufacturing doesn't have a data problem anymore.

00:04:45: We are swimming in data, what we have is a context problem

00:04:49: Because knowing the motor's going to vibrate of tolerance and fail Is just raw data

00:04:54: Exactly!

00:04:55: The actual business insight is knowing this specific motor Drives the conveyor belt for critical batch of aerospace parts.

00:05:02: Right if it fails on Thursday you ruin fifty thousand dollars worth Of materials

00:05:06: And delaying massive customer shipment.

00:05:08: That...is context.

00:05:10: But getting that context requires really deep integration between the shop floor and top floor.

00:05:15: I might be so careful about how we build it!

00:05:17: Yeah, Mohamed Kamal had a stark warning about this.

00:05:20: He did!

00:05:21: he pointed out that trying to scale industrial IoT on flat unsegmented OT networks is just massive security risk.

00:05:29: Oh it's total disaster waiting for you to happen.

00:05:31: You can't take a twenty year old machine designed to be isolated plug into the cloud and cross your fingers.

00:05:37: It doesn't work?

00:05:39: I think of taking modern Formula One engine and dropping it in a rusted thirty-year-old gokart chassis.

00:05:45: That is a great analogy.

00:05:47: Right!

00:05:48: The engine won't make the go-kart win, the sheer torque will just shake the entire frame apart... ...the structural integrity has to come first.

00:05:55: So true

00:05:56: It really begs your question How do we bridge this gap without ripping out decades old multi million dollar factory equipment?

00:06:06: Well you build new operating layer.

00:06:08: right on top of it You virtualize environment

00:06:11: Which brings us digital twins

00:06:13: Exactly.

00:06:14: Once you have that governed data flowing securely through your unified namespace, You can build a real-time virtual representation of your physical reality.

00:06:22: But we need to be clear about what kind of twin we mean here?

00:06:25: Right because Kevin Sullivan from Dissalt Systems was very blunt about this.

00:06:30: He basically said static digital twins are dead

00:06:33: Because the static twin is just a three d blueprint.

00:06:35: right.

00:06:35: yeah exactly.

00:06:37: It only shows what should happen in a perfect world and the shop floor is never a perfect World, it's messy.

00:06:42: And chaotic

00:06:43: very chaotic.

00:06:44: So Sullivan argues we need dynamic scientifically accurate virtual twins that reflect the real operational reality In real time

00:06:54: And that real-time reflection changes everything.

00:06:56: Let me give you a really tangible example of this from Paul Rogers.

00:06:59: Oh, the Heisegun facility?

00:07:00: Yes

00:07:01: their Wetzlar factory in Germany.

00:07:04: instead of relying on old floor plans they actually installed a continuous laser scanner On The

00:07:09: Roof A laser scanner on the roof... That is wild!

00:07:12: Right.

00:07:12: and it constantly maps the shop floor In Real Time.

00:07:15: It feeds live spatial data into a dynamic digital twin.

00:07:19: So the Twin knows exactly where Everything Is down to the millimeter at any second

00:07:23: Exactly.

00:07:24: And because of that, they can deploy autonomous coordinate measuring machines or CMMs.

00:07:30: these inspection robots just navigate the assembly line completely on their own guided by the twin

00:07:35: Wow

00:07:36: and even integrated thousands of solar panels to power it independently.

00:07:40: By relying on this live spatial twin They achieved a thirty percent faster assembly rate.

00:07:45: Thirty percent I mean discrete manufacturing.

00:07:48: That is an astronomical gain It really is.

00:07:50: And we're seeing similar breakthroughs in process manufacturing too, like with liquids and chemicals or not only a tell us share this great example about Unilever site in Vietnam.

00:08:00: Oh what are they doing?

00:08:01: They built an AI enabled digital twin specifically for an intelligent detergent mixer.

00:08:06: Okay interesting

00:08:07: yeah because in Process Manufacturing raw materials were always varying in moisture Or chemical composition.

00:08:13: so instead of human guessing how to adjust the recipe The Twin simulates the exact raw material dosing thousands of times a second.

00:08:21: In real time?

00:08:23: It adjusts instantly to prevent overuse chemicals and that precise automation saves one-to two percent in premium ingredients per batch,

00:08:31: which sounds small.

00:08:33: but when you scale across Unilever's global footprint the return

00:08:36: on investment is absolutely staggering.

00:08:38: yeah

00:08:39: no kidding!

00:08:39: And applications for this kind dynamic virtualization are going way beyond just factories.

00:08:45: Oh, for sure.

00:08:46: Costiantine Pulasuken brought up this crazy crossover fact the FIFA World Cup is actually adopting digital twin technology.

00:08:55: Wait

00:08:55: really?

00:08:56: The world

00:08:56: cup?!

00:08:56: Yeah!

00:08:57: They are actively creating dynamic digital twins for all twelve thousand two hundred and forty eight players in all sixteen stadiums.

00:09:04: That is incredible.

00:09:05: using the exact same underlying logic from manufacturing to support match analysis real-time officiating And crowd management.

00:09:11: it just proves that any complex physical system can be optimized this way.

00:09:16: It does.

00:09:16: but you know, I think we do have to address a major pitfall here.

00:09:20: Yeah Jim Barrett raised a really valid question about this whole explosion of digital twin tech.

00:09:25: He asked if a lot.

00:09:26: these so-called twins aren't just, well, glorious viewing things?

00:09:29: Oh that's good pushback.

00:09:30: like are we building expensive interactive dashboards?

00:09:34: Exactly!

00:09:35: Dashboards look super impressive to executives but don't actually control the single thing.

00:09:40: Right because if manager is staring at three D model on screen That isn't an operating system?

00:09:46: No it not.

00:09:47: But Barrett actually answers his own question by pointing right back to the data architecture.

00:09:51: A digital twin is only a glorified dashboard if the underlying data is trapped in silos.

00:09:57: Ah, so how do you break that?

00:09:59: He advocates for You don't necessarily dump all your data into one massive central lake.

00:10:07: You leave the data where it lives, like in a quality control system but you expose as a formalized product that the twin can instantly

00:10:14: consume.".

00:10:15: Okay so when the data flows is a product The Twin actually transition from just monitoring things to making autonomous closed-loop adjustments

00:10:24: Precisely.

00:10:25: And that transition to autonomous action is where the digital world finally meets the physical

00:10:29: world.

00:10:30: Right, because The Digital Twin can plan and simulate all day but software can't lift a five hundred pound engine block.

00:10:37: Nope!

00:10:37: Software can turn a wrench.

00:10:39: Things still have to physically move which brings us to physical AI in advanced robotics on the shop

00:10:46: floor.

00:10:47: Yes...and Will Healy had massive hot take about this from the Automate twenty-twenty six show.

00:10:53: What did he say?

00:10:54: He argued that the industry is obsessing over the completely wrong things.

00:10:59: The media, you know totally mesmerized by humanoid robots doing backflips or making coffee.

00:11:04: Right!

00:11:05: The viral videos?

00:11:06: Yeah but he says thats a distraction.

00:11:08: Physical AI Is already quietly running in heavy production right now.

00:11:12: Mobile manipulators Automated guided vehicles Cobots They are solving real complex problems today.

00:11:19: It's not science fiction anymore.

00:11:20: No

00:11:21: Peeley makes the point that technology isn't the bottleneck anymore.

00:11:23: The bottleneck is just our own hesitation to adopt what already works

00:11:27: And, To really understand scale of adoption when a company finally does commit.

00:11:30: you have look at the data Gershon Selnaker highlighted.

00:11:33: Oh!

00:11:33: General Motors numbers.

00:11:34: Yes!

00:11:34: General motors replaced over one thousand factory workers with exactly fifty advanced robots A twenty-one

00:11:41: replacement ratio?

00:11:42: That's staggering.

00:11:43: It's

00:11:43: unbelievable and Selnaker makes this very provocative argument.

00:11:48: He believes that collaborative robots, the co-bots designed to work safely alongside humans are really just a temporary bridge.

00:11:55: A bridge?

00:11:56: To what?

00:11:56: Too true!

00:11:57: Lights out manufacturing.

00:11:59: The ultimate destination isn't collaboration it is removing human entirely.

00:12:03: Which brings us those dark factories in China Exactly.

00:12:07: Sunnaker pointed specifically to Chinese automakers like BYD and Geely.

00:12:12: Geely's Xi'an facility is building luxury vehicles with literally zero human presence on the main production floors.

00:12:19: Zero humans?

00:12:20: Zero!

00:12:21: And let us just unpack physics of that for a second, because there are no humans they turn overhead lights completely off which saves massive amounts electricity

00:12:29: Because machine vision uses infrared.

00:12:31: Right.

00:12:31: They don't need visible light.

00:12:33: You do not need expensive HVAC systems.

00:12:37: Steel and silicon don't care if it's ninety degrees inside.

00:12:40: That is so true!

00:12:41: You strip out the human requirements, you just have this highly efficient swarm of physical AI running twenty-four seven.

00:12:47: It's why... It's crucial to understand that physical AI isn't just taking over assembly.

00:12:54: it's actually Taking Over the upkeep of the machines themselves like

00:12:57: predictive maintenance.

00:12:58: exactly Brent Roberts was exploring this.

00:13:01: Traditionally factory maintenance is purely calendar-based, right?

00:13:04: You swap out a bearing every six months Just to be safe regardless of whether it actually needs for a place.

00:13:09: That's

00:13:10: incredibly wasteful

00:13:11: very but industrial AI completely changes that paradigm basically signal their own failures well before they happen by continuously analyzing, you know micro vibrations tiny spikes in energy microscopic temperature fluctuations.

00:13:28: So the machine essentially tells the team it's getting sick before it dies.

00:13:31: exactly that and

00:13:32: we are also removing humans from the diagnostic process.

00:13:35: entirely right especially in dangerous environments.

00:13:38: yes Tobias Claus made a great point about drone based NDT or non-destructive testing.

00:13:43: think You have to build massive scaffolding or send a worker into hazardous chemical tanks just to check a weld.

00:13:52: Terrifying, honestly!

00:13:54: Yeah.

00:13:54: But now they're deploying autonomous drones with ultrasonic sensors right into those tanks.

00:14:01: It completely removes the human risk and cuts down time from days to just

00:14:05: hours.".

00:14:06: I mean, all of this sounds like a silver bullet but we really need to inject some operational reality in this conversation.

00:14:11: Okay let's hear it!

00:14:12: James A. Gamble brought up his framework.

00:14:14: he calls Lean-Fourpointo And is massive Reality Check.

00:14:19: His core argument is simple...but devastating.

00:14:22: Yeah

00:14:23: If you digitize a broken process, You do not get a smart factory.

00:14:27: You just get faster broken processes.

00:14:29: Exactly

00:14:30: That is such brilliant point!

00:14:31: if your floor is disorganized or supply chain routing makes no sense.

00:14:35: dropping the fleet of AI robots in there just automates chaos.

00:14:39: It's like

00:14:40: it's equivalent to buying thousand dollar AI powered Smartwatch To get fit but never bother going to gym.

00:14:46: I love that announcement.

00:14:47: You need operational discipline.

00:14:48: first

00:14:49: Clean up the process manually eliminate the waste, and then apply AI to scale that discipline exponentially.

00:14:56: Spot on because AI is an amplifier!

00:15:00: And when you amplify a lean foundation with governed data and dynamic twins in physical AI it culminates in massive strategic shift for the whole industrial base.

00:15:09: A complete

00:15:10: redesign from software up

00:15:12: Right.

00:15:13: An Axel van Rath introduced this fascinating theory to describe it.

00:15:16: You know the concept of software-defined vehicles right?

00:15:19: Tesla's idea that a car continuously improves via over-the-air updates.

00:15:24: Yeah, it just hardware waiting for new software.

00:15:26: exactly.

00:15:26: well Van Raff argues that the next ten years are entirely about software-defined automation.

00:15:32: Software defined automation, That is going to be monumental especially in regulated fields like life sciences and pharma.

00:15:38: Oh absolutely

00:15:39: Because right now The cost of physically changing a production line To make new drug Is astronomical.

00:15:45: And it's not even hardware thats expensive.

00:15:46: Thats the compliance

00:15:47: right?

00:15:47: Yes!

00:15:48: The massive validation and testing protocols.

00:15:50: Every time you change A physical programmable logic controller or PLC.

00:15:55: Every wire you move has to be

00:15:57: documented.".

00:15:58: But VanRath points out that software-defined automation platforms handle all of the versioning and traceability intrinsically within a software layer,

00:16:06: which means decoupled control logic from physical hardware makes changing factory fast and fully compliant—the holy grail of agility!

00:16:16: It is a total paradigm shift…and if we zoom into broader macroeconomic landscape, Michelle Vassishith offered a perspective that ties all this tech directly to global economics.

00:16:27: And we are reporting this impartially regardless of where anyone stands on Global Trade Politics?

00:16:32: Of course, Vest's sysheth approaches it through purely economic lens He argues while tariffs and trade policies totally dominate the political conversation around protecting domestic manufacturing.

00:16:44: It is actually advanced AI in robotics That will rebuild specialty manufacturing competitiveness In western markets.

00:16:51: By deploying autonomous systems in precision hardware and clean energy, you fundamentally collapse the cost of labor math that sent production overseas.

00:16:59: Exactly!

00:17:00: And the economic urgency behind this shift is huge.

00:17:04: Romain Pady Barad echoed it with a very stark warning aimed at Europe.

00:17:08: What did he say?

00:17:10: He cited BCG research indicating industrial AI could unlock up to a sixty percent gain.

00:17:18: But if Western Europe and the Nordics hesitate while other regions adopt this tech, a staggering one trillion dollars of existing manufacturing value is at risk of relocating.

00:17:28: One trillion dollar?

00:17:30: Yeah It's high stakes global race And Joseph Mason tied these macro threads together perfectly.

00:17:36: He pointed out that the explosion of data centers the race for critical minerals and the fight against manufacturing decline, they aren't separate issues.

00:17:43: They're all part of one giant quiet battlefield of industrial rebuild

00:17:47: right?

00:17:48: And to win that battle Mason says we have to fundamentally rethink our technical workforce.

00:17:53: For twenty years The most brilliant software talent has been building consumer apps and social media algorithms.

00:17:58: We need to redeploy them into physical economy

00:18:00: Exactly Programming the PLCs in the SCADA systems That actually make the physical world move.

00:18:05: So bringing all of this back to you, the professional actually navigating this shift on the ground.

00:18:11: Alex, Allison mentioned the process of replacing legacy systems like Oracle Agile in a recent post

00:18:16: which is huge undertaking

00:18:17: it.

00:18:18: so when you look at replacing your legacy bottlenecks are you treating as just one time IT migration?

00:18:27: Or are you treating it as a multi-year foundational transformation to build true software defined agility?

00:18:35: Because

00:18:36: the platform choices, You make right now will permanently decide if your factories longed at the stagnation or built for continuous evolution.

00:18:42: Couldn't agree more?

00:18:43: and As you map out that path I want to leave you with a powerful framework from Andreas Braun to mull over.

00:18:49: He points out that industrial software has gone through distinct phases.

00:18:53: Historically It was a system of record just reporting what happened after The fact.

00:18:57: Then it evolved into a system of reasoning, analyzing data to help humans understand why machine broke.

00:19:04: But the leap we're making right now is toward A System Of Action

00:19:07: Driven by agentic AI Right?

00:19:09: Exactly!

00:19:09: AI that actively makes and executes decisions without human ever pressing an approve button.

00:19:16: So The provocative question every manufacturing leader must face Is this As your factory becomes capable of making thousands autonomous split.

00:19:26: second decision today Where exactly do you draw the governance boundary?

00:19:30: At what specific point, stop trusting the human operator and start fully

00:19:46: trusting.