Best of LinkedIn: Digital Construction CW 24/ 25

00:00:00: This episode is provided by Thomas Allgaier and Frenus, based on the most relevant LinkedIn posts about digital construction in calendar weeks twenty-four and twenty five.

00:00:09: Frenuse is a B to B market research company that supports industrial automation and ICT companies with market intelligence across the construction industry to prioritize segments identify high value accounts and validate use cases.

00:00:21: you can find more info

00:00:23: Right, so we are jumping straight into the deep dive today.

00:00:26: Yeah and if you're joining us our mission here is really simple.

00:00:29: We are exploring the absolute top digital construction trends that we've seen across LinkedIn specifically from calendar weeks twenty four and twenty

00:00:38: five And we're keeping this completely fluff free, no over-explaining the basics.

00:00:42: You already know the industry.

00:00:43: so we are getting right into the smart focused operational realities

00:00:47: Exactly.

00:00:47: and the first major theme you need to tackle is well it's AI but not The fun experimental chatbot stuff.

00:00:54: Oh No!

00:00:54: We are way past that phase.

00:00:56: Yeah...we

00:00:56: are officially in what can call the platform data war.

00:00:59: It a highly contested battle for proprietary workflows & data control.

00:01:03: I mean..the focus used be on the AI models themselves Right?

00:01:06: But now it's entirely about the data feeding those models.

00:01:11: Li Xiaolong and Rahul Vaishnav actually highlighted this perfectly, they're calling it The Procore Data War.

00:01:20: Right so look at the timeline.

00:01:21: Procore recently banned a rival AI tool trunk tools from its API.

00:01:26: Now trunk tools have been doing really well deployed by massive general contractors like Gilbane and Suffolk.

00:01:32: Wow, so they were already deeply integrated

00:01:34: completely.

00:01:35: They we're running a gentek AI across their projects.

00:01:38: And just to be clear agentic AI isn't just answering questions.

00:01:41: it autonomously triggers actions routing a submittal to an engineer based on some specific project trigger.

00:01:48: That's a huge operational upgrade!

00:01:50: It is, but within months of cutting off trunk tools from the API Procore acquired DataGrid which its own agentic AI provider and just started releasing those native capabilities.

00:02:00: You know Rohu Vaishnav pointed out really funny historical irony there.

00:02:04: Oh about

00:02:04: Autodesk?

00:02:05: Yeah Brocore basically built it's own ground break conference years ago specifically because Autodesk kicked them out of the Autodesc University ecosystem.

00:02:15: Yep,

00:02:15: they got locked up.

00:02:17: and now Procore is doing the exact same thing to trunk tools.

00:02:19: it's just a wildly aggressive play for dominance.

00:02:24: I have to push back here with an analogy Is this just the wall garden era of BIM repeating itself?

00:02:31: Like a decade ago, BIM was supposed to be open and interoperable.

00:02:35: Right!

00:02:36: Seamless collaboration

00:02:37: Exactly.

00:02:38: But instead it devolved into these closed ecosystems.

00:02:41: If every platform locks its training data.

00:02:43: today AI might deliver impressive results Inside isolated silos instead of actually improving multi-party coordination.

00:02:51: That is exactly the risk and that platform risk Is actively driving massive general contractors to look at sauce And just completely

00:02:58: skip.

00:02:58: We really skipped software as a service entirely.

00:03:01: Yeah, Meadle Sharon and Vragan Paraman Anthem Observe this huge shift.

00:03:05: You have a hundred and sixty year old firm like McCarthy Partnering with Palantir To build their own internal AI Which they call Pulse.

00:03:12: So They're building it from scratch.

00:03:14: Basically yeah They're using Palantir's forward-deployed engineers, so these software engineers are embedding directly into the project offices to write code based on actual gritty daily workflows.

00:03:25: That is fascinating!

00:03:27: And Turner Construction is doing it too — they built a proprietary tool called SafeCoach with open AI...

00:03:33: But trained their own data I assume?

00:03:34: Exactly – purely internal safety standards and historical incident reports not the open internet…

00:03:41: Which makes total sense.

00:03:42: As Guido Masiuchi warns, if you build your AI inside a vendor's platform... ...you give them all the leverage

00:03:50: Right!

00:03:50: They can change pricing and access

00:03:52: And product roadmap is theirs.

00:03:54: The real winners are going to be those who own their intelligence outright Definitely.

00:03:58: But I will say trunk tools still proving massive value even with API drama.

00:04:03: Ben Cohen noted they cut submittal review times from fifty-five days down to thirteen for Cleveland construction.

00:04:09: From

00:04:09: fifty five to thirteen, that's incredible!

00:04:11: Yeah the AI just instantly cross references dense specs against product data.

00:04:16: it's undeniably valuable

00:04:18: It is but naturally transitions us into next big issue because AI only as good at ground truth data feeds on.

00:04:25: Right garbage in and out

00:04:26: Exactly which brings us out of the digital files and onto the actual job site.

00:04:31: We're looking at reality capture now, moving from just basic photo documentation to true visual

00:04:38: intelligence.".

00:04:38: And this is where things get really tangible?

00:04:41: Yeah!

00:04:41: Robert Zimmerman shared a powerful insight.

00:04:44: quoting open spaces Jeevan Kalanithi.

00:04:47: He said that percent complete is your biggest lie.

00:04:50: Your biggest lie?

00:04:51: Yeah, I love that phrasing but let me ask you a pointed question here.

00:04:54: sure if traditional progress reporting is just full of subjective noise how does capturing a bunch of visual data actually solve the hardcore commercial pain points?

00:05:04: well Zimmerman addresses that directly.

00:05:06: it comes down to trust in cash flow.

00:05:08: Right now, specialty subcontractors wait an average of fifty-six days to get paid.

00:05:12: Fifty

00:05:12: six days is brutal for cash flow!

00:05:14: It's terrible.

00:05:14: and it happens because the general contractor or owner just doesn't trust that work was done to spec.

00:05:20: They withhold money until they can verify in person.

00:05:23: Ah, so visual data removes that friction.

00:05:26: Exactly!

00:05:27: When a subcontractor can just staple a spatially accurate time-stamped photo of the field reality right to their invoice it takes the fear completely out of the payment process.

00:05:37: You're proving

00:05:37: the ductwork is right before drywall hides it?

00:05:40: Right and ScytheHunt made adds this ease of use creates massive flywheel effect because open space is easy for field teams.

00:05:49: They use it all the

00:05:49: time.

00:05:50: More used means more data

00:05:51: Exactly, and it shifts from just being a camera capture tool into a foundational AI platform.

00:05:57: That is a huge shift And The ecosystem around this Is opening up so fast.

00:06:02: Antonio Gonzales Viegas Just announced This massive reality-capture pipeline On that open platform.

00:06:08: Oh I saw that they are handling Huge files right?

00:06:10: Yeah like fifty gigabyte files Directly in a web browser.

00:06:13: They're using Gaussian splats Instead of traditional point cloud.

00:06:16: So it renders instantly.

00:06:17: That's

00:06:17: a game changer for processing speed.

00:06:20: And it fully integrates with BIM using zero proprietary lock-in.

00:06:24: The market is finally admitting that holding visual data hostage, Is A Losing Strategy.

00:06:29: It really is!

00:06:30: But capturing the reality of this is great... It brings us to our third theme.

00:06:36: You have to actually align that physical capture with the original digital plans.

00:06:41: Oh, The BIM and Digital Twin Data Pipeline Nightmare!

00:06:44: It is a nightmare.

00:06:45: Florian Humer put out this stark warning about building digital twins.

00:06:48: He says you aren't just parsing JSON files... ...you are basically fighting a five-front war

00:06:53: A five-front war.

00:06:55: Okay, it's like.

00:06:55: well if they try to assemble a puzzle where the pieces were cut by five different factories Yes using five totally different measurement systems.

00:07:02: That's the perfect analogy.

00:07:03: humor says you're fighting divergent IDs coordinate chaos geometric mismatches

00:07:08: missing metadata

00:07:08: Yeah and linkage gaps.

00:07:10: so your left with this massive data swamp instead of a functioning digital twin.

00:07:14: Which is exactly why Santosh Komarboda argues for a strict, ninety-day ROI framework.

00:07:20: Right because you can't spend three years cleaning data.

00:07:23: Exactly!

00:07:24: You need credible evidence of value fast.

00:07:26: If it doesn't show ROI in one quarter...you kill it.

00:07:29: Cut your losses.

00:07:31: But you know grassroots innovation is actually solving some of these BIM headaches while the big platforms catch up.

00:07:38: Really?

00:07:38: Like

00:07:39: who?

00:07:39: Suzuki Kwan Zhang built this open source Revit plug-in that batch links Autodesk construction cloud models.

00:07:46: Oh, I hate linking those manually.

00:07:48: how does it work?

00:07:49: It's brilliant.

00:07:50: Normally, an IT admin has to register and app with OOOTH protocols which takes forever but his plugin just pulls the active session token.

00:07:58: that's already sitting in Revit.

00:08:00: Wait it uses the authentication you have?

00:08:02: Exactly!

00:08:03: It bypasses need for an Admin completely.

00:08:05: You run it and it links models.

00:08:08: That is so smart.

00:08:09: I saw Benjamin Friedman doing something wild too Experimenting what he calls Buildings as code.

00:08:15: Oh yeah using AI Agents.

00:08:16: Yeah

00:08:17: specifically Claude Code He's building BIM workflows straight from natural language prompts.

00:08:21: You just type what you want and the AI agent codes the building parameters.

00:08:26: It's mind-blowing where this is heading, but...and it a big software doesn't pour concrete.

00:08:31: Which brings us to our final theme The human element And lean execution.

00:08:36: To close the gap between a design office and field, we really have to look at people.

00:08:41: Right because Catherine Greville made this incredible point.

00:08:43: She said that every single construction challenge whether it's productivity or digital adoption Or whatever ultimately traces back-to-people problems.

00:08:51: Always

00:08:51: always comes down into culture in behavior.

00:08:53: Christopher Lange is highlighted as massive irony right now.

00:08:57: We had these huge AI tech boom But that boom desperately needs electricians iron workers, and concrete crews to physically build the data centers that house the AI.

00:09:10: AI cannot build its own

00:09:11: buildings.".

00:09:12: It's The Ultimate Paradox but let me challenge that AI jobs narrative for a second.

00:09:17: John Morris actually asked an AI what a job site will look like in the year twenty one hundred?

00:09:22: And it generated a picture with zero humans just autonomous robots everywhere...

00:09:27: Just robots is where we're heading.

00:09:29: Well, Morris completely refutes that and Lindsay Renkel backed this up perfectly.

00:09:34: She notes that AI's real value isn't replacing the builder.

00:09:38: it's handling their repetitive tedious

00:09:40: tasks like what?

00:09:42: Like quantity takeoffs?

00:09:44: if they can instantly count architectural symbols estimators get hours of time back.

00:09:50: so is about augmentation not replacement.

00:09:52: construction will always need human adaptability Absolutely.

00:09:56: Navigating a dynamic, muddy job site requires human judgment

00:10:00: and to help those human builders we need to communicate better.

00:10:04: I loved what Adam Hoots did.

00:10:05: he co-author the old dog Lean Thesaurus.

00:10:07: yes

00:10:08: stripping away the academic jargon

00:10:09: exactly taking lean concepts and translate them into gritty, plain language for the trades.

00:10:15: Because a pipe fitter doesn't need an academic lecture on efficiency!

00:10:17: Right they need practical tools.

00:10:18: And Dan Desio brought this home by pointing out that modern superintendents really NEED A HYBRID SKILLSET TODAY.

00:10:24: Yeah it's not just about raw field experience anymore.

00:10:26: No

00:10:26: you need THAT GRITTY EXPERIENCE but also needs sharp engineering AND TECH EDUCATION to manage these complex digital models in the field.

00:10:34: The cognitive load is just incredible

00:10:37: It REALLY IS.

00:10:38: Well that brings us to the end.

00:10:40: I want to leave you with one final really provocative thought from Craig Park's upcoming book, Building Beyond.

00:10:46: Oh this is a great insight!

00:10:47: Yeah so for forty years we've been totally obsessed with this idea of smart buildings and automation.

00:10:53: Automating the HVAC networking security?

00:10:56: Exactly.

00:10:57: but park says The real question isn't what technology can do...the real questions whether AI will actually make these building better.

00:11:07: That is such a crucial reframing.

00:11:10: Are we building better environments for the human beings living and working in them, or just more complicated machines?

00:11:17: Exactly!

00:11:17: It's definitely something to chew on as all this tech rolls out.

00:11:21: If you enjoyed this episode new episodes drop every two weeks.

00:11:24: Also check our other editions of Smart Manufacturing and Digital Power Tools.

00:11:28: Thanks so much joining us on This Deep Dive.

00:11:30: Keep Building Smart And don't forget to subscribe.

00:11:33: We will catch you next time.