Best of LinkedIn: Digital Construction CW 18/ 19

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about digital construction in calendar weeks eighteen-and-nineteen.

00:00:08: Frennes is a BDB 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.

00:00:22: you can find more info

00:00:24: And for everyone listening, welcome.

00:00:26: Yeah Welcome!

00:00:27: So For this deep dive we are really setting out to explore the top digital construction trends that we've seen across LinkedIn over those two calendar weeks.

00:00:36: Right,

00:00:36: Calendar Weeks eighteen and nineteen

00:00:38: Exactly!

00:00:39: And our mission for you today is pretty straightforward We want to cut through the massive amount of tech hype out there To figure what's actually working on job sites

00:00:47: because a lot noise

00:00:48: There so much noise.

00:00:49: So looking at AI, robotics connected workflows I wanna start right off with AI Because it's dominating conversation.

00:00:57: everywhere You look

00:00:58: Oh absolutely It all.

00:00:59: anyone talking about

00:01:00: Right.

00:01:01: But when you look at the actual posts from people in the trenches, You realize pretty quickly that AI and construction isn't actually a technology problem at all.

00:01:10: No

00:01:11: it's data problems A massive data problem.

00:01:13: It is like our construction firms essentially trying to drop a Ferrari engine Like this super advanced AI into... ...a car just held together by duct tape like their legacy data.

00:01:26: I mean, i love that analogy and honestly calling it duct tape might be a little too generous in some cases Really?

00:01:32: Yeah!

00:01:33: I mean Joe Wilson actually had a post about this.

00:01:35: That totally validates your point.

00:01:36: He said throwing AI on top of these fragmented datastacks doesn't fix anything.

00:01:40: It just makes the mess worse

00:01:42: faster.

00:01:43: Oh, wow.

00:01:43: So it just accelerates the bad data?

00:01:45: Exactly

00:01:46: and he shared this brutal stat.

00:01:48: He pointed out that when construction companies try to build internal data solutions They only succeed like thirty three percent of the time.

00:01:54: Thirty-three

00:01:55: percent!

00:01:55: That's terrible.

00:01:56: It is so you know firms panic And they tried to go out and buy a massive enterprise solution right Like Palantir or something Sure

00:02:03: which I assume Is not cheap.

00:02:04: Not at all.

00:02:05: Wilson noted those contracts cost an average Of five point six million dollars.

00:02:11: oh my god.

00:02:11: yeah.

00:02:12: And the reality is most builders just you know, they don't have the margins to afford a five million dollar tech

00:02:18: patch.

00:02:19: No of course not.

00:02:20: Yeah But it really makes me wonder why that data?

00:02:23: Is always such a disaster.

00:02:24: to begin with and Dan Howard actually had a really great insight about this That caught my

00:02:29: eye.

00:02:29: oh yeah I saw that one.

00:02:30: ya he asked us fundamental question Who actually owns data quality at a construction firm?

00:02:37: right because usually Nobody does.

00:02:39: Exactly,

00:02:40: it's not the IT department and The project managers.

00:02:43: I mean they are way too busy out there putting out literal fires on the job site to sit down And write standard operating procedures for data entry.

00:02:51: They're exhausted?

00:02:52: They don't have time For that.

00:02:52: yeah exactly and so just falls through the cracks.

00:02:55: well and It gets scarier when you look at the scale of what we're talking about.

00:02:58: Hugh Tran posted this really stark warning About infrastructure scale.

00:03:02: That ties right into This.

00:03:03: okay

00:03:04: What did he say?

00:03:04: So, he pointed out that you cannot trust the fifty-year data backbone of say a new airport or hospital to some AI startup.

00:03:14: That only has maybe eighteen months of venture capital runway.

00:03:17: Oh wow!

00:03:18: I hadn't even thought about the longevity of the software companies themselves.

00:03:21: Right, because in construction information loss has kind have just been normalized for decades.

00:03:27: handing over data from design to build to maintenance is always a mess.

00:03:31: Yeah that makes sense.

00:03:31: And Oliver Lowry backed this up with a crazy statistic.

00:03:35: He noted that construction represents thirteen percent Of global GDP right?

00:03:39: It's massive sector

00:03:41: Huge But it had seen exactly zero percent productivity gain In thirty years Zero

00:03:47: That is just wild to think about, zero percent in three decades.

00:03:51: Exactly!

00:03:52: So if the data foundation is shaky you really have to wonder what happens when these AI products actually hit the job site?

00:03:57: Which is

00:03:58: messy right... It's

00:03:59: incredibly messy.

00:03:59: I mean The overarching friction point with all this AI implementation Is that it completely fails.

00:04:05: the second it disrupts the human rhythm on a site.

00:04:08: Oh yeah Harshal N had a post about that exact scenario and was so relatable.

00:04:13: Yeah the one about the Site Engineer

00:04:15: Yes He said.

00:04:16: imagine asking a site engineer who just worked at twelve hours shift in forty degree heat To pull out an app and start manually typing data into a form,

00:04:27: right?

00:04:27: They're not gonna do it.

00:04:29: Never to guaranteed failure.

00:04:31: He said.

00:04:31: AI have to fit into the existing site walkthroughs.

00:04:34: It can't just create parallel administrative work.

00:04:37: Well, and even if they do use it Massimo Scopini pointed out three systemic problems We're seeing right now with current AI on sites.

00:04:44: Okay

00:04:45: Let's go through them.

00:04:45: What's the first

00:04:46: one?

00:04:46: So first he says AI actually surfaces ten times The amount of problem.

00:04:51: isn't that a good thing?

00:04:52: you'd

00:04:52: think so.

00:04:53: But on the ground, it just overwhelms project managers.

00:04:56: They're getting drowned in notifications about tiny insignificant clashes...

00:05:00: Oh see!

00:05:00: It's just noise.

00:05:01: Exactly.

00:05:02: And second problem is trades.

00:05:04: Schiapini shared that subcontractors are actively rejecting AI generated critiques of their change orders.

00:05:10: I bet they hate this.

00:05:11: They

00:05:11: despise it.

00:05:12: Basically saying you know..I'm not negotiating my thirty years experience with an algorithm.

00:05:17: Honestly

00:05:17: i get it.

00:05:18: I'd probably do same thing.

00:05:19: Yeah

00:05:20: but his third point was the wildest one.

00:05:22: He says we are entering this era where in contract disputes, AI is answering AI.

00:05:28: What does it even

00:05:29: mean?

00:05:29: It means the sub's AI writes a claim and general contractor's AI reads it and writes a rejection.

00:05:35: You're

00:05:35: kidding me?

00:05:36: Nope!

00:05:37: It leaves legally responsible humans completely out of loop.

00:05:40: Okay but hang on.

00:05:41: If AI is just causing subcontractors to argue with algorithms and drowning people in notifications, are there any actual wins right now?

00:05:50: Is anyone doing this

00:05:51: right?".

00:05:51: Yes.

00:05:52: Actually Thomas Alt shared a really tangible win.

00:05:55: he posted about rambler.ai and Hitachi launching this tool called Ai Assist Pro at ConXPO.

00:06:02: twenty-twenty

00:06:03: six.

00:06:03: Okay.

00:06:03: so why is that one working when the others aren't?

00:06:05: It works because it takes incredibly fragmented, knowledge-like heavy equipment manuals or expert walk around videos and turns all of that into multimodal live operational intelligence.

00:06:19: Exactly.

00:06:20: An operator can basically point a camera at a broken part on an excavator, and the AI instantly tells them how to fix it based on the manual in videos if its their workflow.

00:06:30: Oh that's brilliant!

00:06:31: Like a hyper smart tool belt not an auditor looking over there shoulder.

00:06:34: Exactly So.

00:06:36: we've talked about digital AI software but I want shift gears because theres this whole other side of this which is physical AI right?

00:06:44: Robotics

00:06:44: oh yeah huge topic now

00:06:47: Right Because The Whole Industry views robots as the magic answer to these massive structural labor shortages.

00:06:54: They really do, but Felix Arumbin had a post that shared some brutal truth about this...

00:07:00: The one about the brick-laying robot?

00:07:01: Yes!

00:07:02: He pointed out.

00:07:03: everyone thinks the robot is the bottleneck…but

00:07:08: How so?

00:07:09: Well, take FBR's Hadrian-X or the SAM one hundred.

00:07:13: These are essentially mobile manufacturing systems.

00:07:16: if your site has poor slab tolerances Or the logistics are just a little bit off

00:07:21: The robot just fails

00:07:23: right.

00:07:23: it doesn't know how to improvise like a human mason would.

00:07:26: It just stops.

00:07:27: So robotics don't actually fix sight.

00:07:30: inefficiency they'd just expose.

00:07:32: That is fascinating.

00:07:33: It exposes the chaos that was already there

00:07:35: exactly

00:07:36: and it's not like these robots are small either.

00:07:38: I saw Walter produce post about all three mantis robotic system.

00:07:42: Oh,

00:07:42: that thing is massive.

00:07:44: Yeah He said he has a four meter reach And A one hundred kilogram payload capacity its doing intense physical tasks.

00:07:51: But if These massive Robots Are operating on These chaotic inefficient sites Doesn't that make them a massive safety hazard?

00:07:58: That is such a good question, and heading Radell from Hardout Robotics actually addressed this exact paradox.

00:08:04: What did

00:08:04: he say?

00:08:05: He noted that, you know normally robots are brought in to remove humans from dangerous tasks like high voltage electrical work.

00:08:11: Right which makes total sense for safety!

00:08:13: But he points out they introduce a wholly new category of risk.

00:08:17: Now instead worrying about human error You have to worry the failure mode of an AI Controlling massive piece heavy physical machinery In unpredictable environment.

00:08:28: Wow so your trading one type of risk For completely different

00:08:31: ones Precisely, and it all kind of points back to the fact that you can't just slap advanced tech onto a broken site

00:08:39: Which really brings us upstream doesn't it too planning?

00:08:42: And BIM the human beings running these projects.

00:08:46: Yeah,

00:08:47: it all stems from upstream coordination gaps.

00:08:49: every downstream failure we just talked about starts there.

00:08:52: yeah Thurman Thomas had a perfect example of this.

00:08:54: he posted about This massive hospitality project in Dubai.

00:08:57: Oh I remember that post.

00:08:59: The BIM workflow completely broke down

00:09:02: Totally collapsed and was because they have distributed freelance design team That basically treated Revit like it was just regular old cat.

00:09:10: Oh no, that's a huge mistake!

00:09:12: Right they were just drawing these beautiful vigils but completely ignored Dubai's local authority regulations and the actual buildable data.

00:09:20: so It looked pretty But it was impossible to build

00:09:23: exactly because massive rework on the site.

00:09:25: And that ties perfectly into what Devin Vancey Was saying about prefabrication.

00:09:30: He pointed out That when prefab fails On this massive hyper scale AI datacenters its almost never A failure of factory floor.

00:09:38: Really?

00:09:38: Where does it fail?

00:09:39: It's a failure in pre-construction strategy and BIM readiness.

00:09:44: If the model is wrong upstream, The prefab parts just aren't going to fit together on this site Right!

00:09:49: The pieces literally won't match up.

00:09:51: Yeah Which kind of brings that all back To people involved?

00:09:54: I loved This quote from Cindy Lee That i saw.

00:09:56: She said A hammer doesn't build a house the person does.

00:10:00: That is spot-on!

00:10:01: It really is, and Robert Dong had a great observation about this too.

00:10:05: He pointed out that in contact you absolutely need both top down leadership mandate And bottom up power users

00:10:11: Because he needs bosses to pay for it.

00:10:13: but workers actually want use.

00:10:15: exactly he said construction doesn't have casual prosumers who just adopt tech If it doesn't actually help the person swinging a hammer or managing this site, It's just going to be another expensive piece of software sitting unused.

00:10:28: Yeah you have to win over people doing their work!

00:10:38: Yeah, thanks for

00:10:46: listening.

00:10:46: And before you go I want to leave you with one final thought to mull over.

00:10:50: we saw a really profound concept from Arginio and Tao who called this whole situation the One point.

00:10:58: six trillion dollar intelligence gap.

00:11:01: Wow that's a big number

00:11:02: it is.

00:11:03: he asked us amazing question.

00:11:05: The technology is clearly here, but is your firm's culture finally ready to stop managing by surprise and hiding behind sanitized boardroom reports what he calls polite fictions?

00:11:15: And start leading with the raw actionable truth trapped on your job sites.