Best of LinkedIn: Digital Construction CW 10/ 11

00:00:00: This episode is provided by Thomas Allgeier and Frennus, based on the most relevant LinkedIn posts about digital construction in calendar weeks ten-and eleven.

00:00:08: Frenness is a B to B market research company that supports enterprises across the construction industry with the market customer and competitive insights they need.

00:00:22: Welcome to today's deep dive.

00:00:24: if you're listening.

00:00:25: You probably already know the real pain of trying to bridge that gap between a pristine digital plan, and well a highly unpredictable physical job site.

00:00:35: Oh the muddy reality of the jobsite?

00:00:36: Exactly!

00:00:37: The muds...the delays all of it.

00:00:39: so today we're extracting the absolute top digital construction trends buzzing across the industry.

00:00:45: We are getting these directly from people actually building tech in pouring concrete curated from calendar weeks ten to eleven.

00:00:51: Right no fluff cutting straight through noise.

00:00:53: whole mission here is take those signals unpack them figure out how they you know apply your daily operations construction and manufacturing.

00:01:01: Because if you think about it, there's this profound sense of comfort when looking at a perfectly rendered three D construction model right?

00:01:09: Everything is just pristine.

00:01:11: Oh yeah like the piping doesn't clash with HVAC The structural steel aligns to

00:01:17: Exactly, and the timelines look like this beautiful cascading waterfall of efficiency.

00:01:22: And then you step out of a trailer...

00:01:24: Right!

00:01:24: You step it's raining The drywall deliveries three hours late.

00:01:28: Two trades are arguing over sequencing.

00:01:31: Your superintendent is screaming about a spec change buried in an email from last Tuesday.

00:01:38: Yeah that collision of bits & atoms Is exactly the battleground we're exploring today.

00:01:43: And to kick things off, I mean we really have to talk about the massive elephant in the room which is AI.

00:01:48: Oh for sure it's dominating everything right now?

00:01:49: It is.

00:01:50: but i want to start with a serious pushback question.

00:01:53: this is based on some data shared by Monseit Aran.

00:01:56: apparently ninety one percent of AEC firms architecture engineering and construction they plan to boost their AI investments by twenty-twenty six.

00:02:05: I mean ninety one per cent.

00:02:06: that is a staggering majority market.

00:02:09: It's massive.

00:02:10: But then, Widomochiuchi pointed out a really glaring issue with those numbers.

00:02:14: like only twenty-seven percent of these firms are actually using it right now.

00:02:18: Wait!

00:02:18: Really?

00:02:19: Just twenty seven percent.

00:02:20: Yeah.

00:02:21: So if AI is so revolutionary for the job site why that adoption gap is huge?

00:02:25: Are we just stuck in innovation theater?

00:02:28: Innovation

00:02:28: Theater Wow That was great way to frame

00:02:31: it Running a few shiny pilots, but not fundamentally changing how building actually comes out of the ground.

00:02:37: Well in my COG hits on something fundamental there.

00:02:40: The gap isn't actually technology problem.

00:02:42: It's a governance problem.

00:02:44: What do you mean by that?

00:02:45: well A lot of firms are merely playing with the outputs so they They use language model to generate a quick schedule template or maybe summarize a chaotic meeting and they pat themselves On the back and call it digital transformation.

00:02:59: Oh sure like the easy stuff

00:03:01: Exactly, but the firms seeing real actual bottom line returns.

00:03:07: They're doing the incredibly unglamorous work of completely redesigning specific workflows around AI capabilities.

00:03:15: You know you can't just slap a sophisticated chatbot on top for broken legacy process and expect magic.

00:03:20: Right because a broken process is still broken And that actually brings in the regulatory side too like the EU AI

00:03:26: Act.

00:03:27: Oh absolutely

00:03:27: if you aren't actively governing these tools and restructuring your data protocols, you're exposing your firm to massive compliance risks down the line which is precisely why we are seeing heavy weights from outside of traditional context space aggressively moving in.

00:03:43: Yeah

00:03:43: that's a huge shift.

00:03:44: It really is.

00:03:44: Like Li Xiaolong made this fascinating observation about Palantir entering the hundred and fifty billion dollar construction tech market.

00:03:52: Palantire,

00:03:53: that's interesting!

00:03:54: Right... And Palantiri isn't jumping in to build like another basic system of record or a simple document repository.

00:04:00: they are building systems decisions

00:04:02: because if you look at average contractor margins They have around what?

00:04:05: A knife thin.

00:04:06: four percent

00:04:07: Exactly, four percent.

00:04:08: At that margin the goal isn't just to store your data better.

00:04:11: you need to instantly simulate cost and schedule scenarios across highly siloed ERP software.

00:04:18: Right those massive enterprise resource planning databases tracking all the money in materials.

00:04:23: Yep.

00:04:24: so Palantir is looking to optimize operational efficiency by connecting those silos mathematically.

00:04:30: but let's um Let's break that down for the folks in pre-construction, because AcarSephani laid out a highly practical on-the-ground roadmap.

00:04:38: Oh

00:04:39: I saw.

00:04:39: he argues you don't need massive multi million dollar plant here level overhaul to see value

00:04:45: right?

00:04:45: Exactly!

00:04:45: You can use AI right now for heavy lifting.

00:04:48: He points at things like auto takeoffs where the AI drafts initial quantity takes off.

00:04:52: so it is counting every door and window Every linear foot of drywall

00:04:55: And then human estimator just does QA quality assurance on the Deltas.

00:05:01: He also highlighted using AI for spec package comparisons to flag scope gaps before bid day.

00:05:07: Which is huge.

00:05:08: finding those gaps Before buyout, Is where the real risk mitigation lives?

00:05:12: I mean discovering a missing lighting package when you're already pouring foundations.

00:05:16: Yeah that destroys your four percent margin instantly.

00:05:18: oh

00:05:19: it vaporizes It.

00:05:20: and i actually saw demo illustrating That exact principle from fernando metareno with his startup construct ai.

00:05:26: Oh what are they doing?

00:05:28: They're taking standard TD PDFs and using AI to turn them directly into three D Revit models.

00:05:33: Like people were losing their minds in the comments asking for beta access,

00:05:37: I bet that's historically been such a manual slog.

00:05:40: it really is so.

00:05:41: To me Ai In its current state is like having uh A highly caffeinated incredibly literal intern.

00:05:47: a caffeinated intern.

00:05:48: That's perfect

00:05:49: right?

00:05:49: It handles all the mind-numbing manual extraction reading thousands of pages of specs converting pds counting fixtures all so that your experienced senior humans can focus entirely on analyzing the actual risk profile of The Build.

00:06:02: And, you know... That human element gets lost in the hype.

00:06:05: So often there's been this pervasive anxiety across the industry that AI is going to replace the trades or automate estimators out-of-a job.

00:06:13: Yeah people are definitely worried about that.

00:06:15: But Sarah Buckner from Trunk Tools and researcher Chris Brady They both highlighted that This fear completely misreads what the technology actually does.

00:06:25: Oh, so?

00:06:25: Well AI's exposure is firmly behind-the-desk.

00:06:28: it's not out on the site swinging a hammer.

00:06:31: its highest and best use is sifting through like two dozen conflicting architectural and structural documents up-to-date drawing the first time.

00:06:43: Oh, I see.

00:06:43: so it's amplifying the human builder by removing the administrative chaos?

00:06:48: Exactly!

00:06:48: It prevents the crew from executing the wrong plan flawlessly.

00:06:52: Executing the wrong planned flawlessly yeah that happens way too often.

00:06:55: but here is the catch with all of this.

00:06:58: AI cannot function in a vacuum

00:07:00: oh hundred percent.

00:07:01: it requires a massive incredibly structured data foundation or you just get highly confident hallucinations, and Alistair Lewis actually brought up a brilliant historical perspective on this.

00:07:11: Right because the AEC industry didn't just suddenly discover AI six months ago?

00:07:15: No we've actually spent the last twenty years inadvertently preparing for it.

00:07:20: every time affirmed digitized to paper drawing into building information model Everytime they structured an Information Delivery specification They were laying the groundwork.

00:07:31: They're building this structured data foundation that modern AI requires to run complex spatial calculations

00:07:37: Exactly, and now we are finally seeing tools that automate the cleaning of that twenty year old Foundation.

00:07:44: Conrad Fugus was talking about The Nightmare Of Validating thousands of properties across different industry foundation classes or IFC model.

00:07:52: Oh man, manually checking if every single steel column has the correct load bearing metadata?

00:07:58: That is soul-crushing work!

00:08:00: Soul crushing is the right word.

00:08:02: but now with tools like information delivery specification or IDS people are using Reggie East to automate that validation.

00:08:09: For anyone who isn't a software engineer, Regie X is basically super powered search command right?

00:08:14: Yeah exactly.

00:08:15: It scans thousands of lines code or data to instantly find and correct formatting errors forcing the data be perfectly clean before AI even touches it.

00:08:23: But

00:08:23: we need address massive limitation here.

00:08:26: The entire concept of perfectly structured BIM data assumes that all project knowledge actually lives in a database.

00:08:33: Which it absolutely doesn't, right?

00:08:35: So much of actual construction knowledge the real boots on-the-ground.

00:08:39: reality isn't sitting neatly in a BIM model or an ERP drop down menu.

00:08:44: It

00:08:44: lives and gut feelings.

00:08:46: Yes gut feelings, angry emails from a subcontractor group text messages hurried phone calls at six AM when a concrete truck gets flat tire.

00:08:56: How does an algorithm make sense of that chaos?

00:08:59: Well that is the multi-million dollar question.

00:09:01: and it brings us to concept shared by Robert Dong context graphs.

00:09:05: Okay, walk us through how a context graph actually works in practice because this sounds fascinating.

00:09:09: So

00:09:09: historically AI could only read clean tables of numbers.

00:09:13: It couldn't reliably extract meaning from unstructured offline data.

00:09:17: Right it needed neat little rows and columns

00:09:19: Exactly.

00:09:21: But with modern large language models these contest graphs act almost like a detective's string board

00:09:25: Like the red string connecting photos on a corkboard.

00:09:28: Yes exactly like that.

00:09:30: They ingest the unstructured chaos, emails, daily logs and meeting notes.

00:09:36: Then AI tokenizes that text and maps mathematical relationships between entities.

00:09:42: So

00:09:42: it connects dots?

00:09:43: Right!

00:09:44: It draws a string connecting an angry text message from sub to weather report Tuesday To delayed invoice in ERP.

00:09:53: So instead of a project executive relying purely on gut instinct during a gonna-go bidding decision, the context graph surfaces historical pattern.

00:10:02: Can you give an example?

00:10:02: what that looks like?

00:10:03: Sure!

00:10:04: You can say looking at our unstructured communications and financial data whenever we work with this specific general contractor on a data center project over fifty million dollars Our change order rate is forty percent higher than average due to sequencing disputes.

00:10:17: That is insane, it mathematically weighs how three completely unrelated things caused a schedule slip.

00:10:23: It's essentially quantifying the superintendent's intuition and turning institutional memory into a queryable

00:10:29: database."

00:10:30: It's a massive leap forward.

00:10:32: however we have to navigate what is known as The

00:10:35: Reality Gap.

00:10:37: Well even with perfectly clean BIM data in comprehensive unstructured context graphs Digital models can fundamentally misrepresent physical reality.

00:10:47: Because

00:10:47: the model only shows what should be happening according to the math, not what is actually happening in the mud?

00:10:52: Precisely!

00:10:54: Prené Egerthy shared a fascinating thought experiment that perfectly illustrates this reality gap.

00:10:59: Okay let's hear it.

00:11:00: Imagine a digital model of complex chemical plants distillation column The steady state math the software solvers, the advanced process controls.

00:11:10: They all calculate that column is running perfectly at maximum feed rate.

00:11:14: The dashboard is entirely green.

00:11:16: Sounds good so far

00:11:17: Right but out in physical plant over head purity drops instantly and millions of dollars a product goes off spec.

00:11:22: Wait why?

00:11:23: If it's green?

00:11:24: Because the steady state digital model failed to capture fundamental thermodynamics happening inside the pipe.

00:11:30: Oh wow So math was flawless.

00:11:32: But physics were wrong.

00:11:34: A standard, three-D model is basically useless for operations if it assumes a frictionless environment.

00:11:40: And that gap... ...is exactly why the industry's being forced to evolve from static three D models into living digital twins

00:11:48: Which brings up a stark warning from Basim Di.

00:11:51: He stated A digital twin without real data is just a beautiful illusion.

00:11:56: A beautiful illusion.

00:11:57: that's incredibly accurate,

00:11:58: right?

00:11:59: If your gorgeous three D Digital replica doesn't have continuous Real-time Data feeding it from the actual physical site It drifts form reality immediately and basically becomes a very expensive video game.

00:12:11: Yeah

00:12:12: You need material flows equipment usage logs environmental sensors, like

00:12:17: the ones manufactured by Wittra for example.

00:12:19: You need those sensors feeding that model constantly so it reflects the physical world in real time.

00:12:25: but building a continuous data feed is notoriously difficult.

00:12:29: Florian Humer pointed out integration where these twin projects usually fail

00:12:33: because they just can't get systems to talk with each other.

00:12:35: Exactly The ultimate goal of digital twins isn't have cool three-D visuals spinning on monitor and sight trailer.

00:12:43: The goal is a unified data environment.

00:12:45: Okay, so what does that actually look like when it works?

00:12:48: What you actually want?

00:12:50: Is this an area where a physical internet of things or IOT sensor to text the moisture leak on the site and That physical event automatically triggers a work order in your ERP system Which is perfectly aligned with the materials specified In your BIM

00:13:05: man?

00:13:05: that level of integration is the holy grail And the stakes for getting this right are just skyrocketing.

00:13:12: Ahmad Jandali brought up high-rise construction in this context.

00:13:15: Oh

00:13:15: yeah, data volume there is wild.

00:13:17: It's staggering!

00:13:18: If you're building a tower with three hundred and fifty plus units The sheer volume of mechanical electrical and plumbing MEP data Is insane For a fifty year maintenance life cycle Especially when heating and cooling efficiency is heavily regulated.

00:13:34: Having a living digital replica isn't illuxury for the asset owner anymore.

00:13:38: No, it's fundamental necessity to manage energy costs over half-a century.

00:13:41: But okay let me play devil's advocate here because this sounds like an absolute avalanche of noise.

00:13:46: It does.

00:13:47: If we are tracking every single delivery Every workers movement Every environmental spec And every IOT sensor trip across a fifty story building.

00:13:56: Yeah

00:13:57: Aren't project managers just going drown in data?

00:14:00: That is very real fear

00:14:01: Like how does a human being process all of that without missing the forest for trees?

00:14:06: Well, Santosh Kumar Boda provided an absolutely brilliant framework to solve this problem.

00:14:12: He argues that the future of digital twins isn't actually about collecting more data.

00:14:18: it's learning what to ignore.

00:14:21: Learning what to ignore, we just spent the last ten minutes talking about capturing every single detail.

00:14:25: Break that down for me!

00:14:26: He compares it to biological intelligence and neuromorphic computing.

00:14:30: Okay Think of how a human brain works.

00:14:33: You aren't constantly consciously processing the feeling of cotton short on your back or hum from air conditioner in room

00:14:40: Right And tuning out

00:14:41: Exactly.

00:14:42: Your brain filters static noise out to conserve energy.

00:14:46: It only reacts meaningful events like a sudden drop in temperature or someone tapping your shoulder.

00:14:51: Oh, so instead of the dashboard constantly yelling that everything is fine it only speaks up when something's actually wrong.

00:14:58: Event driven architecture.

00:15:00: right now digital twins use standard API polling.

00:15:04: they basically ask the sensor are you okay?

00:15:06: Are You Okay?

00:15:08: Every millisecond which creates massive data bloat

00:15:11: Yeah!

00:15:12: That sounds incredibly inefficient.

00:15:13: It

00:15:13: Is Future.

00:15:14: Digital Twins will act more likely.

00:15:16: central nervous system.

00:15:17: They will filter out the static noise of a functioning site, and only send a signal when threshold is breached or an anomaly occurs.

00:15:24: That shift from a static dashboard that requires human to constantly monitor it... ...to living nervous system alerts you when feels pain….

00:15:33: …that's next great leap in spatial intelligence!

00:15:37: It treats this structure as organism with a nervous system.

00:15:40: That is a wild way to think about building.

00:15:41: But let's bring this all the back down into dirt, we've talked AI context graphs, Muramorphic event driven architectures, Automated Regie X validation All

00:15:51: very high level stuff

00:15:52: Right but all of this brilliant engineering means absolutely nothing if humans on job site refuse open app.

00:15:59: Oh one hundred percent.

00:16:00: The ultimate bottleneck and always will be human adoption.

00:16:04: Elliot Christensen said it perfectly.

00:16:06: he says technology doesn't create productivity.

00:16:09: Adoption does, the job site is The Ultimate Product Manager.

00:16:13: That's so true!

00:16:13: If a software vendor from Silicon Valley hands a superintendent A new tablet tool and that tool makes their already stressful day even five percent harder it Is dead on arrival.

00:16:24: And Steph Morgan had this painfully True critique of why So many contact companies fail at This exact hurdle.

00:16:32: What did she say?

00:16:33: She pointed out that Software vendors fail when Their pitch relies On venture capital jargon.

00:16:38: You know, using phrases like synergistically AI-optimized workflows instead of addressing what actually happens when a pipe breaks at three AM.

00:16:46: Yes

00:16:46: the reality of a job set UI is completely different from a sleek office.

00:16:50: an interface designed for a mouse and a mechanical keyboard Completely fails When it's being poked by a muddy thumb on a cracked iPad in The Blinding Sun.

00:16:58: Oh yeah If your software doesn't solve that Three A.m Panic...the guys who have been running sites For twenty years will just stop returning Your calls.

00:17:06: And James Walker actually connected this adoption friction directly to lean construction principles.

00:17:11: How so?

00:17:12: Well, projects rarely lose their profit margin in one massive dramatic mistake like a multi-million dollar structural failure.

00:17:19: they lose their margin in hundreds of small unknown delays

00:17:23: death by thousand cuts.

00:17:25: exactly it's a crane sitting idle for forty five minutes waiting for a lift sign off its materials dropped on the wrong side of the compound because The coordination model wasn't updated.

00:17:34: So where does Lean come?

00:17:36: Lean construction is a relentless focus on removing that micro-friction.

00:17:40: If new technology requires the superintendent to enter the same data twice, it's adding friction and actively working against lean

00:17:48: principles.".

00:17:48: That makes total sense!

00:17:50: So how do firms bridge this massive gap between Silicon Valley software engineer and job site reality?

00:17:57: Well, Wraggon Paramanatham shared what he calls A Secret Truth about Construction Innovation Teams... He says they often fail because they lack operational authority.

00:18:09: Meaning, that don't actually run the job?

00:18:10: Right!

00:18:11: They sit outside of operations usually in an innovation lab and they do not own the profit-and-loss statement for their actual build which

00:18:18: means they can't mandate how a project team actually completes work.

00:18:23: Regan argues that the industry desperately needs hybrid operators.

00:18:26: Hybrid operators?

00:18:27: Yeah,

00:18:28: you don't need a pure software engineer trying to guess what a superintendent wants.

00:18:32: You need the project estimator who got so frustrated copying and pasting cells in Excel That they spent their weekends learning Python to automate their bid tabs.

00:18:40: Ah I love that.

00:18:41: Right!

00:18:42: You need The Project Engineer Who understands data infrastructure but also knows exactly why the site scheduling logic differs from the textbook version.

00:18:49: Because those hybrid operators have credibility.

00:18:52: They've solved real painful internal problems with their own two hands.

00:18:57: So the field teams actually trust them.

00:18:59: exactly when a hybrid operator brings A new tool to the site, The site knows it's built To solve a three AM problem not to secure a series B funding round for a tech

00:19:08: startup.

00:19:09: It all comes back to the human element.

00:19:12: You can have the most advanced AI and the most perfectly synced digital twin But construction is still a human endeavor.

00:19:20: It is, it's people organizing massive amounts of heavy materials in the physical world.

00:19:25: Which is really the perfect lens to view where the entire industry is heading next.

00:19:29: because while we are figuring out these adoption challenges and refining our context graphs The market itself is shifting beneath our feet.

00:19:37: Oh

00:19:37: definitely

00:19:38: Drawing on some deep insights from Ishae Smith and Terry Olanek We can see this macro level restructuring happening.

00:19:45: What does that restructoring look like?

00:19:47: On-the-ground for you listening.

00:19:48: well

00:19:49: If you look at the underlying signals, fabrication backlogs, crane order books power availability.

00:19:54: The construction market isn't shrinking.

00:19:56: it is transitioning away from commodity commercial buildings and rapidly moving toward incredibly dense heavy and complex industrial infrastructure.

00:20:05: We are talking about hyperscale AI data centers semiconductor fabrication plants in advanced energy grids.

00:20:12: The stakes are getting higher vastly more complex.

00:20:20: And because of that escalating complexity, the AEC firms The

00:20:30: ones who use connected tools to execute with absolute clarity and remove the friction from their project teams.

00:20:41: Yeah, Terry Onick put it bluntly He said firms don't lose projects because they lack capability.

00:20:46: They loose because of our legacy.

00:20:47: processes simply cannot keep up with the speed market now demands.

00:20:51: It really makes you think about that blueprint versus the mud.

00:20:54: Our tools are finally getting smart enough to handle the reality at a mud But only if we have the operational discipline To actually redesign or workflows around them.

00:21:02: That is great take away.

00:21:04: And I want to leave you with one final thought to mull over on your next project, building this idea of AI acting as a nervous system.

00:21:12: Lay it on us!

00:21:13: We've talked extensively about these systems filtering noise and identifying risk.

00:21:18: but as we hand more operational trust over to event-driven digital twins... ...we really have to ask a difficult question about liability.

00:21:25: Oh

00:21:26: high ability?

00:21:27: That's

00:21:28: tricky.

00:21:28: Very.

00:21:30: If a neuromorphic AI predicts critical sequencing failure or structural risk, and human superintendent ignores the alert because they rely on their twenty years of gut instinct who is legally liable when project halts.

00:21:44: Wow.

00:21:45: As our sites get smarter and our algorithms predict reality with greater accuracy, the legal reality of who is ultimately responsible for a digital decision... ...is the next great frontier we will have to

00:21:55: navigate.".

00:21:56: That's heavy but so true!

00:21:58: If you enjoyed this episode new episodes drop every two weeks.

00:22:01: also check out other editions on smart manufacturing in Digital Power Tools.

00:22:04: Thank You So Much For Joining Us On This Deep Dive.

00:22:06: Don't Forget To Subscribe.

00:22:07: We'll See Ya Next Time.