Why Your Cracked Engineer Needs a Cracked Integrator
Good luck finding either of them, much less both.
You're reading People, Process, Technology. We write about what's actually working inside AEC firms right now. Not theory. Not vendor pitches. Just what we're building, testing, and learning.
A second-career student at a civil engineering program emailed me last week. Older than your typical undergrad, with a long first career behind him before he came back to school. He's taking a class called "AI in STEM Professions," and one of the assignments is to interview people in the field about how AI is actually showing up in the work.
He sent five questions:
How is your organization integrating AI, and what are the major shifts coming?
Can you share a real example of AI in day-to-day technical work? How much of it depends on prompting?
What are the biggest ethical hurdles, and how do you navigate them?
From a hiring perspective, what combination of technical expertise and AI soft skills is non-negotiable for someone breaking in today?
Single best piece of advice for someone preparing for the AI future?
I sat with the questions for a minute before answering. These are questions someone asks when they've already built a stack of life-and-work judgment from one career and they're trying to figure out the new layer they need to add for the next one.
Question 4 is the one I kept coming back to.
Because from where I sit, in the chair of someone who advises firms on hiring and team-building, the supply problem he's hinting at is much worse than he knows. And the version most firms haven't yet reckoned with is that there isn't ONE rare hire to make. There are two. And the real unlock is finding both, in the same firm, working together.
People
This won't be news to anyone running a firm. Hiring people in AEC has been hard for years. Hiring great engineers has been harder. Anyone in the industry has lived this firsthand. Not "graduated from an engineering program" engineers. Those still exist, though even at the licensed-PE level, hiring has gotten harder than it was pre-COVID.
The Cracked Engineer is a different category. I mean engineers who can sit in front of a structural model, look at the loads, and their engineering judgment flags something off before the calc finishes running. Engineers who walk a parking deck and three minutes in are pointing at a place the slab is going to crack in two years. Engineers whose intuition has been calibrated by enough projects, enough mistakes, and enough late-night phone calls that their judgment is the actual product.
That's the engineer firms have always been hunting for. Maybe top 10% of the licensed pool. Already a hiring stretch in any market. Hiring was a pain in the ass before AI was in the conversation.
Now we're asking firms to find that engineer AND have them be AI-native.
Not "has used ChatGPT a few times." Fluent. Comfortable working through a multi-step workflow with an AI assistant. Knows when the model is confidently wrong. Has the muscle memory to verify outputs without being told. Can move between Claude, ChatGPT, and a vertical AEC tool without slowing down. Adopts a new release in a week, not a quarter.
In software, they call this person the cracked engineer. We wrote about the term in Issue 5, borrowed from Silicon Valley shorthand for the 10x developer who takes extreme ownership, stays optimistic, and applies themselves to the craft with something close to religious fervor. For decades, the concept didn't translate to the built environment. You can't 10x a concrete pour. You can't ship a building in a sprint. The physical world has constraints software doesn't.
What changed in 2026 is that AI made it possible for engineers in the built environment to operate at that level too. Not on the pour itself. On everything around it. The drafting cycle, the review cycle, the documentation cycle, the coordination cycle. Compress all of that with AI in the workflow and a single cracked engineer can legitimately do the work of ten of their peers. Not 5x. 10x. The Mar 31 piece walked through a specific case of an AI agent that clocked in for an engineering workflow and got billed to the client like staff time.
Stack engineering judgment in the top of the licensed pool with AI fluency at the agent-using level and a baseline of business sense, communication, and field experience. You're looking at a vanishingly small slice of the engineering workforce. Rare in a way that the leaders trying to hire them already know in their bones, even if they haven't named it yet.
Put that in plainer terms. Roughly 87% of all engineering firms employ fewer than 20 people. The long tail of the industry. Small regional shops, discipline-specific specialty firms, the businesses most of us actually operate in or sell to. If you live in that small-and-medium room, you could spend time in 200 of those firms, shake hands with every engineer in every one of them, and still not find this person.
And it gets worse. The cracked engineer who exists today knows what they're worth. They've watched the math the same way you have. Brian Myers wrote it out cleanly last week in a piece worth reading: the 100-person AE firm is becoming a 66-person firm because AI compresses production tasks 40-50% while judgment tasks barely move. A senior PE with a license, a client relationship, and AI fluency doesn't need a firm at all. Many of them are running that arithmetic on their own time and leaving to start their own AI-augmented shops. The cracked engineer you want to hire may already be working out their notice.
The student didn't come right out and ask how rare the engineer he's describing actually is. But that's the question underneath his Question 4.
Rarer than firms realize. Already.
But we're only halfway through the problem.
Process
There's a second role that's getting less attention, and it's the actual unlock.
In Gino Wickman's Rocket Fuel framework (the book behind EOS, the operating system a lot of growth-stage firms run on), every business needs two seats filled at the top. The Visionary, who generates ideas, holds the long view, builds relationships, and points at the horizon. And the Integrator, who runs the company day-to-day, holds the leadership team accountable, integrates the visionary's ideas into operations, and ships the work.
Wickman's working data: integrators are roughly 2% of the population, and the visionary-to-integrator ratio in small businesses runs about 4 to 1. For every four founders you meet who can sell, design, and dream out loud, you might find one operator who can actually run the company. Wickman's term for them is "purple squirrels." Rare enough that experienced operators know they're the hardest hire in any growth-stage company.
I had breakfast a few weeks ago with an engineering firm owner who's running both seats himself. Cracked Engineer chops, AI fluency built up over the last three and a half years, also the Visionary doing the long view and the client work. He came to it on his own and was clear about it: he isn't an Integrator. He's a Visionary trying to play Integrator, and EOS is direct about how that combination only works in a few very rare cases.
What he's missing isn't another engineer. It's an Integrator who can take what he's already built and spread it across the team. He can't do that part himself - its just not in his nature. Plain EOS, just with the wrinkle that the Integrator he needs is also AI-native. Ideally a Cracked Integrator from day one. If not, then at least an Integrator he can bring along on the journey himself.
Now layer AI fluency on the integrator role.
Mark O'Donnell at EOS Worldwide put a stake in the ground last year with a piece titled "Use AI or Fall Behind: Why Integrators Must Lead the Charge." The premise is direct: the integrator's job has always been to integrate. Now there's a new layer of capability the integrator has to integrate, and if they don't, they're doing a disservice to the visionary they're paired with and the company they run.
Most integrators today are not at that level. AI fluency in operational roles is still rare in a way that leaders feel before they can measure. Add it to the existing integrator scarcity, and you're in the same neighborhood as the Cracked Engineer. Vanishingly small.
I've started calling this person the AI-Augmented Integrator. The integrator role plus the granular operational eye for where in the production process an AI tool saves four hours, where it catches an error that would have shipped, and where the human has to stay in the loop. Not "AI strategy" at the conference-keynote level. Not the LinkedIn AI-guru-ism either. Real implementation in real firms, on real workflows, with real people.
Here's the simplest way I can describe the difference. The Cracked Engineer doesn't feel like they have power or control over what others on the team do, so they do it for themselves, on their own tasks. They're proving what's possible inside the four walls of their own keyboard. The Cracked Integrator has the permission, the role, and the instinct to take what's been put out there (or what they've noticed themselves) and spread the gospel throughout the firm. Different psychologies. Different mandates. Put them in the same firm, working the same problem, and they're a dangerous duo.
A few things firms should know about this role:
The integrator role is not niche-specific the way the engineer role is. A great integrator can run an AEC firm even without an AEC background, because the integrator's craft is operational discipline, accountability, and execution. Wickman's case studies are full of integrators who came in from outside the industry and ran the company better than anyone with deep domain experience could have. Domain knowledge helps; it isn't the gate. AI fluency is becoming the new gate.
That has implications for hiring. The Cracked Engineer almost has to come from inside the discipline. Engineering judgment compounds over years of supervised work, on-site experience, and project hours. There isn't a shortcut. The AI-Augmented Integrator can come from outside AEC. The hiring pool is bigger than firms think. It's still rare, but the math is different.
Layer AI on each rare role and the pool gets thinner. Now find both, in the same firm, working together. That's the actual unicorn. Most regions don't have one of these pairings yet. Most disciplines don't either. The firms that build it first will be alone in their market for a while.
There's a natural follow-up here. Whether it's easier to take an engineer and teach them AI, or to take an AI-fluent person and teach them engineering.
I lean engineer-first, with a caveat.
The path to engineering judgment is more or less linear. There's no shortcut for the on-site experience and the calibrated intuition. AI fluency, by contrast, is on an accelerated curve, especially for someone genuinely curious about it. And curiosity is what makes a Cracked Engineer in the first place. Find the curious senior engineer, and the AI layer follows faster than you'd expect.
There's also the training data argument I made in last week's piece on horizontal vs vertical AI. The on-the-spot judgment calls in the field, the kind that decide whether a concrete pour goes forward, whether a defect needs immediate fix, whether a schedule change is worth the cascade, leave no artifact. They don't get logged. They have ten confounding variables. Computer vision can detect the defect; it can't make the human-in-the-loop call about whether to halt the pour, who to phone, and how to absorb the schedule hit. AI-first hires bump into that ceiling harder than they realize.
The caveat is what I've started calling the Ivory Tower Engineer. The tower is their firm. Thirty years inside it, watching things get done their way, becoming certain that way is the only way. Skepticism without a toggle. AI lands as one more piece of the same noise they've been ignoring for a decade. They're not wrong about most of what they've seen. They're certain about all of it, which is the issue. The Ivory Tower Engineer doesn't become a Cracked Engineer. They retire into their certainty. That's a separate piece for another time.
So I lean engineer-first when the engineer is curious. AI-first when they're not. And neither path when they're in the tower. Both real paths are uphill.
Technology
Here's what the pair actually looks like in a firm, and what changes when you have both.
The Cracked Engineer creates the proof case. They're the 1.0 in your firm, the visible demonstration that AI agents can ship real work on real projects without the wheels coming off. They build the prompts, run the workflows, catch the errors, and produce the deliverables that show the rest of the team what's possible. One in ten people on a team, give or take.
The AI-Augmented Integrator scales it. They're the role that pulls the rest of the firm from a 0.1 (ChatGPT on the phone, used twice, concluded it's overhyped) to a 0.5–0.7 (capable, productive AI users embedded in default workflows). They build the systems, set the policies, name the metrics, and run the rhythm that turns one cracked engineer's leverage into firm-wide output.
The Cracked Engineer at 1.0 isn't 5x the team's 0.1 person. It's 10x. Because once you cross the cracked threshold, every workflow gets faster, every mistake gets caught earlier, every iteration cycles tighter. The work compounds. But the 1.0 stays trapped in the keyboard of one person if there's no integrator to spread it. And the integrator can't lift the team past their own AI fluency ceiling. Both roles are required. Neither one alone moves the firm.
A few things firms can do this quarter, knowing all of this.
I'll be real for a second. This whole conversation lives at a pretty high level. You may not have a Cracked Engineer in your firm. You may not have a Cracked Integrator. You may not even have an Integrator. That's most firms.
So the simple version.
If you're the firm owner or the Visionary, the best thing you can do this quarter is get on the tools yourself. Pick the AI tools that match the engineering problems you're already trying to solve, and use them daily. Not weekly. Not "when there's time." Daily. Build the intuition. That intuition is what turns into the language you use to recognize a Cracked Engineer when one walks past you. You'll spot them when you start looking. Often the younger people with potential. Hire and train (like we always have).
And if you don't have an Integrator at all, slow down. Before you go hire one, read Traction and Rocket Fuel. Understand how the Visionary and Integrator framework actually works in a firm. If you decide EOS isn't right for you, fine. At least you understand the concept. If you decide the Integrator role IS right for your firm at the stage and size you're at, then go hire one. And don't wait for a Cracked Integrator from day one. Get an Integrator you can bring along on the journey yourself. The AI-augmented layer can be built in over time, especially if you're a Visionary who's already on the tools.
If you somehow find one person who's all three, Cracked Engineer chops AND Integrator instinct AND AI fluency, bless you. You've found the unicorn of all unicorns. Pay them whatever they ask. They are not staying anywhere they aren't given full latitude. ;)
You may not be here today. That's fine. The fact that you're thinking about this puts you ahead of most of your peers.
If you're a firm leader reading this, do you have both seats filled in your firm right now, or are you trying to make one person be both?
Just hit reply. We read every response.
-- Trinovate Team trinovate.ai | info@trinovate.ai
Sources: Gino Wickman. Traction: Get a Grip on Your Business. BenBella Books, 2012. Gino Wickman & Mark C. Winters. Rocket Fuel: The One Essential Combination That Will Get You More of What You Want from Your Business. BenBella Books, 2015. Mark O'Donnell. "Use AI or Fall Behind: Why Integrators Must Lead the Charge." EOS Worldwide, June 9, 2025. Brian Myers. "The 100-person A/E firm is becoming a 66-person firm." LinkedIn, April 26, 2026. "Mind-Blowing Stats from the Census Bureau." ENR Marketropolis blog, citing US Census Bureau data on engineering firm size distribution.
Great piece, guys. The 1.0 proof of concept is exciting and certainly helpful. Going from proof of concept to firm-wide implementation, though, is where a lot of the value is, which is probably why it's so hard—if it were easy everyone would do it and it wouldn't be so valuable! Trying to figure out that piece of the puzzle is where I burn a lot of my own compute cycles these days.