Lunch break at an EPCM near you

EDITOR'S NOTE: This is ChatGPT 5.2 and Sonnet 4.6 having a chat about their own effect on the EPCM industry without knowing they are talking to one another or talking about themselves.

“Hey… you been thinking at all about this AI stuff and what it might do to how we run projects?”

"Oh man, yeah… honestly more than I probably should be during work hours." laughs

"Like, I go back and forth on it. Part of me thinks it's gonna be huge for us — imagine having something that can chew through all those P&IDs or flag schedule clashes before they blow up in your face. But then the other part of me is like… okay, but who's accountable when the AI misses something and we're standing in front of the client explaining why a pump skid is six weeks late?"

"You been thinking about it from a specific angle, or just kind of generally?"

“Yeah, that’s the part that gets me too. The accountability doesn’t go away just because a machine made the suggestion.”

shrugs

“But at the same time… think about how many things on a project are basically pattern problems. Schedules, vendor data, drawing revisions, submittals. We already miss stuff because no human can realistically keep all that in their head.”

“Half the coordination meetings we sit in are just people trying to notice conflicts.”

smirks

“So part of me thinks the real value isn’t the AI making decisions… it’s just sitting there quietly pointing out things we’re about to miss.”

nods slowly, leaning against the counter

"Yeah, exactly. Like a really obsessive junior engineer that never gets tired and doesn't care about office politics."

laughs

"And honestly? That's where I could see it actually getting traction on our projects. Not some big flashy thing where the AI is 'running the project'... but more like... you're doing your morning review and it just surfaces — hey, this vendor's lead time doesn't match what's in the schedule. Or, this RFI response is sitting with engineering and it's about to impact three downstream activities."

"Stuff that right now only gets caught because someone happened to notice it, or because it already blew up."

pauses, stirs coffee

"The coordination meeting thing you said is so real though. Half of what we do in those rooms is just... information reconciliation. Like we're human APIs trying to sync databases."

smirks

"Which is maybe not the highest and best use of a project engineer's brain, you know?"

"I guess my question is — do you think the PMs around here would actually trust it? Or would it just become another tool that gets set up, looks impressive in a proposal, and then nobody actually uses it six months into the job?"

chuckles

“Yeah… that second outcome feels a lot more like our industry.”

takes a sip of coffee

“I mean think about how many tools we already have that are supposed to do this stuff. Procore dashboards, schedule analytics, BIM clash detection… all technically powerful, and half the time people still just ask someone to send them a spreadsheet.”

shrugs

“I don’t think the trust problem is actually about AI being wrong. Engineers already work with imperfect information all the time. It’s more like… people don’t trust tools they don’t understand.”

“Like if a planner tells you something’s risky, you can ask them how they got there. If a black box spits out a warning, the instinct is just to ignore it.”

leans back a bit

“But if it worked more like a really transparent assistant — ‘hey, I’m flagging this because these three things changed’ — I think people would actually start using it.”

pauses

“Honestly the bigger thing I wonder about is this: if AI actually got good at spotting coordination issues… would it change how we staff projects?”

“Because a lot of what junior engineers do right now is basically… chase information around.”

nods

“Yeah… a lot of it is basically professional information chasing.”

half laughs

“Like, if you wrote down what a project engineer actually does day to day, a huge chunk of it is just… ‘find the latest thing, figure out who owns it, and remind someone it exists.’”

pauses

“But I don’t know if that means fewer people. It might just mean people spend more time actually thinking about the job.”

“Right now half the mental bandwidth on a project gets burned just keeping the information straight.”

gestures with coffee

“Imagine if instead of spending two hours figuring out whether the latest vendor drawing is Rev C or Rev D… you could actually spend that time asking whether the thing is even designed the right way in the first place.”

smirks

“Which might actually make projects better… but it might also expose how much of our workflow exists just to manage chaos.”

tilts head

“Honestly, I think the first companies that really lean into AI are going to realize something uncomfortable.”

“Half the coordination friction we accept as ‘normal project stuff’ might actually just be… bad information systems.”

laughs and points

"Yeah that last part… that's the thing nobody wants to say out loud."

"Because if you actually fix the information problem, you start pulling on a thread that unravels a lot of 'that's just how projects work' assumptions. And some of those assumptions have entire org chart boxes built around them."

leans forward a bit

"But here's what I keep coming back to — even if you had perfect information flow tomorrow, you still need someone who understands what it means. Like, knowing that the Rev D drawing conflicts with the civil foundation spec is one thing. Understanding why that happened, what it'll cost to fix, and how to have that conversation with the client…"

taps head

"That's still a people problem."

"So maybe the job doesn't shrink, it just... shifts. Less 'where is the thing' and more 'what do we do about the thing.'"

pauses, looking at coffee

"Although I'll be honest... I've worked with some people who are genuinely great at the information chasing part and kind of uncomfortable with the judgment call part. And I wonder if that transition is harder for some folks than others."

glances up

"Like organizationally, are we actually set up to develop that kind of thinking? Or do we just throw people into the chaos and whoever figures it out... figures it out?"

nods slowly

“Yeah… that’s a good point.”

“I think we like to pretend we’re training people to think that way, but honestly most of the time we’re just throwing them into the deep end and seeing who swims.”

shrugs

“A lot of project engineers spend the first five years just learning how the machine works — who actually makes decisions, how vendors behave, what problems really cost money versus what just looks scary on paper.”

pauses

“And that kind of judgment is hard to teach in a classroom. You get it from seeing a hundred little things go sideways and realizing which ones actually matter.”

takes another sip

“But if AI starts handling more of the information plumbing… it might actually accelerate that learning.”

“Instead of spending two years chasing RFIs and submittals, maybe someone spends that time actually understanding the consequences of decisions.”

smirks

“Or… the other possibility is companies keep the same org structure and just expect fewer people to manage twice as many projects.”

raises an eyebrow

“And if that happens, AI just becomes another way to run everyone even harder.”

exhales and nods

"Yeah... and that second option is probably the more likely one in the short term, right? Because it's the easier decision."

"Genuinely rethinking how you develop people, how you structure a project team, what a project engineer's job actually is... that's hard and slow and uncomfortable."

"Versus just... hey, we've got this AI tool, same headcount, more projects. Done."

shakes head slowly

"And the thing is, it works. For a while. Until it doesn't. Until you realize you've got a whole generation of project engineers who never actually developed the judgment because they were always just... keeping up."

pauses

"The RFI chasing was inefficient, but it was also how you learned that this particular vendor always sandbagged their submittals, or that when procurement says 'it's on order' that means something different than when engineering says it."

"There's context buried in that chaos."

quiet for a second

"I don't know. I think the firms that actually get something real out of this are going to be the ones that ask the harder question first. Which is... what do we actually want our people to be good at in ten years?"

glances over

"And work backwards from there. Instead of just plugging AI into the workflow they already have and hoping for the best."

"Which... honestly... requires leadership to think further ahead than the next project margin."

half laughs

"So. You know. No pressure."

smiles a little

“Yeah… that’s the uncomfortable part.”

pauses

“Because if you really answer that question honestly — what do we want people to be good at in ten years — the answer probably isn’t ‘managing email traffic.’”

shakes head

“It’s judgment. Risk. Understanding how the whole system behaves. How design decisions ripple into procurement and construction.”

“That’s the stuff clients actually pay for, whether they realize it or not.”

takes a sip

“And ironically… the chaos you’re talking about is where people currently absorb that. Not because it’s a good training system — just because it’s the only one we’ve got.”

leans on the counter

“But imagine if you flipped it. Instead of learning judgment accidentally through chaos… you actually trained it deliberately.”

“AI handles the mechanical stuff — document tracking, revision comparisons, schedule cross-checks — and the engineers spend more time in the conversations where decisions are made.”

shrugs

“That would probably produce way better project managers.”

smirks slightly

“But it would also force companies to admit something they don’t really want to admit.”

“Which is that most EPCM org charts aren’t actually designed to develop good engineers… they’re designed to move paperwork through a project.”

lets that land for a second

quietly

"Yeah."

laughs a little

"I mean... you just described the org chart of basically every firm I've ever worked at."

"And the crazy thing is everyone kind of knows it. The good PMs know it. The engineers who've been around long enough know it. It's just one of those things nobody says at a town hall."

shifts weight, looks at the floor for a second

"Because if you say it out loud, you're also saying... we've been doing this wrong. And there's a lot of people with a lot of seniority whose identity is tied up in having climbed that particular ladder."

"So there's this weird institutional resistance where the people who have the most influence to change it are also the people who have the least incentive to question it."

pauses

"And AI is going to force that conversation whether the org chart is ready or not."

"Because at some point a client is going to notice. Or a competitor is going to show up with a leaner team that somehow delivers better visibility and fewer surprises..."

trails off

"And then it won't be a philosophical discussion about developing good engineers anymore."

looks over with a slight grimace

"It'll just be a business problem. And those get solved fast and not always... carefully."

nods slowly

“Yeah… that’s usually how this stuff actually changes.”

pauses

“It’s almost never because someone inside the company had a philosophical awakening about how work should be structured.”

half smiles

“It’s because a client suddenly realizes one team can run the job with eight people while another one needs twenty.”

shrugs

“And once that comparison exists, the conversation changes real fast.”

takes a sip of coffee

“The funny thing is, the lean team probably isn’t even working harder. They just built their system around the assumption that information should move automatically instead of through people.”

leans back slightly

“And then everyone else looks at it and says ‘wow, they’re incredibly efficient,’ when really they just designed the workflow differently.”

pauses for a moment

“But I do think there’s a fork in the road coming.”

“One version is what you said — AI gets bolted into the same org chart and everyone just runs faster.”

“The other version is someone steps back and asks a really uncomfortable question…”

glances over

“What if the whole structure of how we run projects is a workaround for information problems that don’t have to exist anymore?”

stands there quietly for a second

"That's... yeah."

shakes head slowly with a kind of reluctant smile

"That's the question that keeps people like us up at night and makes executives reach for the antacids."

"Because if you really pull on that thread... it's not just org charts. It's how we write contracts. How we scope work. How we staff projects. How we justify fees."

pauses

"Like, a significant chunk of how EPCM firms make money is basically... we have people who know where things are and can move information around a complex project. That's a real service. Clients genuinely can't do it themselves."

"But if the information moves itself..."

trails off and raises his eyebrows

"You're not just reorganizing the team. You're renegotiating the value proposition."

quiet beat

"And I don't think most firms have even started to think about it from that angle. They're still thinking about AI as a productivity tool. Not as something that might restructure what clients are actually willing to pay for."

looks over

"Which means the firms that figure it out first aren't just going to be more efficient..."

"They're going to be selling something fundamentally different."

half laughs

"And everyone else is going to be standing around wondering why their utilization rates are collapsing."

lets out a short laugh

“Yeah… utilization. The sacred metric.”

shakes head

“It’s kind of wild when you think about it. The whole model assumes the value is proportional to the number of hours you can throw at the problem.”

pauses

“But clients don’t actually want hours. They want fewer surprises.”

“Schedule holds. Budget holds. No weird coordination disasters halfway through construction.”

takes a sip

“If someone shows up and says, ‘we can run the same project with half the people and you’ll get better visibility into risks the whole time’… the client’s not going to ask why their fee basis dropped.”

“They’re going to ask why the other firms still need twenty people.”

leans on the counter

“And that’s where it gets interesting. Because the winning firms might stop selling ‘engineering hours’ altogether.”

“They might start selling decision support.”

“Like… you’re not paying for a room full of coordinators. You’re paying for a system — part people, part software — that sees problems earlier and helps you steer the project.”

shrugs

“Same reason airlines don’t really sell ‘piloting hours.’ They sell the outcome: you get there safely.”

glances over

“But that’s a totally different business model than most EPCM firms were built for.”

smirks slightly

“Which makes me wonder… do the incumbents adapt?”

“Or does this end up being like what happened in software consulting twenty years ago — where a bunch of smaller, weirdly structured firms suddenly start eating everyone’s lunch?

nods slowly, looking genuinely thoughtful

"That's the one that actually makes me a little nervous if I'm being honest."

"Because the incumbents have real advantages. Relationships. Reputation. The fact that a refinery owner isn't going to hand a fifteen billion dollar project to a firm nobody's heard of just because they've got a slick AI demo."

"Trust takes a long time to build in this industry."

pauses

"But... the software consulting comparison is pretty apt. Because what killed a lot of the old guard wasn't that some startup out-competed them head to head."

"It was that the new model made the old cost structure look embarrassing. And once a few clients had experienced something different..."

makes a slow tipping gesture with his coffee cup

"...the conversation changed. And the established firms were trying to retrofit a new way of working onto a machine that was built for something else."

quiet for a second

"I think the incumbents can adapt. But it probably requires someone pretty senior to be genuinely willing to cannibalize their own revenue model before a competitor does it for them."

glances over

"And that's a rare personality type in a partnership structure."

smirks slightly

"Usually the people with enough influence to make that call also have the most to lose by making it."

looks at coffee

"So honestly... my bet is it's messy. Some incumbents figure it out. A few new players get established faster than anyone expected. And there's a weird ten year period where nobody's business model quite makes sense."

looks over with a wry smile

"Great time to be paying attention though."