Garbage In, Sugar Cookies Out
With Orion Matthews at AACE International on March 20, 2026.
According to Forbes, they are the six most expensive words in business. In construction, it's just Tuesday. The phrase, “we’ve always done it that way” is muttered about workflows, about spreadsheets, even about the way we break project data into trees. And all too often it comes right after someone new asks a simple question, eliciting the power to shut down the conversation before it starts.
Orion Matthews chooses to ask those questions for a living. He's the founder and CEO of Queryon and host of the Major Projects Podcast, so most people would say his superpower is analytics: turning complex data into a clear story.
But that's just where his power shows up.
The real superpower sits underneath it. Orion came to construction as an outsider with a deep level of curiosity, and that combination lets him walk onto a project and see what the rest of us stopped seeing a long time ago.
Two Kinds of Exceptional
Orion started programming in 1987 at age seven, writing BASIC on an Apple II. By high school he was winning programming competitions and taking college-level classes. That early start built something hard to describe: a subconscious fluency with systems that most people have to work through consciously.
He's quick to say he wasn't a childhood genius, though. He was what's called a "twice exceptional" kid, testing extremely high in some areas while extremely low in others. Strangely enough, that pattern followed him into construction. He's never shoveled concrete, and the field can smell that on someone. That's the deficiency. The exception is that when he models a data problem in his head, he can see exactly where the blockers are, the spots where construction teams tend to muscle through a configuration and hope it works.
Sometimes that creates sparks. Sometimes it creates something better.
When an outsider and an insider build enough trust to really hear each other, that's where the magic happens. Every project controls team seems to have a story about "the one wizard" who came in and made everything flow. It’s that wizard who is typically the person who just so happened to see the problem differently.
The Construction Lorax
So what does the outsider see that the rest of us don't? Orion's answer is perfectly EngiNerd. Construction is obsessed with its trees.
Work breakdown structures, cost breakdown structures, risk breakdown structures, org breakdown structures. Everything starts with a big concept and decomposes into parts, subparts, and leaves. It's a seductive model because most things in the universe really do look like that.
The problem is that data trees are incredibly difficult to work with.
There's a reason Airtable exists and Airtree doesn't. Merging two tables is straightforward. Merging two trees is a genuinely hard math problem, and every time we try to tie the schedule to the cost structure, that's exactly what we're attempting to do.
But then we add a third tree, a fourth, a fifth even, each one built inside a different platform to make that platform work. Integration forces them to collide, and the silo problem gets a hundred times worse.
But Orion isn't saying throw trees out. Instead, he's saying stop treating trees like they’re simple tables. If you don't approach a tree with respect and a healthy dose of fear, it will trap you in a web of silk and eat you. A lot of engineers spend their careers getting lost in the trees.
The Engineer’s Best Friend
This also explains why we spend millions on platforms and then export everything to Excel anyway. Orion traces it back to the late 1990s, when IT departments shut down the homegrown databases and tools people used to build for themselves. The industry shifted from building solutions to buying them.
That works fine for the finance side of the house, where everyone works with a clean general ledger. But it most certainly doesn't work for construction.
Every construction project spans multiple organizations nobody fully controls. Managing information across those lines is a custom build problem by nature, and for 20 years the only workbench left standing was Excel. Meanwhile, IT has a tendency to hand the project team the same requirements form it hands finance, so even with three years and a few million dollars of work, the projects end up having nothing to show for it. So we all turn to Excel.
Excel isn’t our go-to because it’s so great. It was really the only tool left standing.
The Hidden Variable Problem
Now enter artificial intelligence, the next “quick fix” to our data problems. Except, there’s a catch.
A colleague of Orion's, a PhD in Malaysia, calls it the hidden variable problem. Construction data is missing its context. Take a foreman's budget. There's padding in there, and the number never explains what that padding is for, how that decision got made, or even how much was added. Someone who has been in the trade reads that number with a sixth sense for what's hiding inside it.
AI doesn't have that sense.
And therein lies the real problem. You see, the idea of garbage in, garbage out has died with AI. A bad spreadsheet used to look bad, in obvious ways, and someone up the management chain would catch it. Now AI takes the same pigsty of data and turns it into a polished chart that looks right and speaks with total confidence.
Orion reframes it: garbage in, delicious, poisonous sugar cookies out. They travel through an organization fast, and real decisions get made on them. AI isn't the threat. AI fed from a pigsty is, because it will create beautiful lies and organizations will act on them.
Because those decisions could bring down an entire company.
Two Mirrors
So what comes first, before trusting the next AI tool or dashboard? Orion's first move is to split our thinking in two.
Operational data: everything a system needs to manage the work.
Decision-making data: the important pieces we actually act on.
They don't have to be the same set, and one doesn't have to roll up into the other.
Think about an RFI. A project management platform needs a mountain of fields to manage it (like time stamps, user IDs and IP addresses). A leader making a business decision needs a handful (like, the response). The mistake is decomposing everything to the very bottom, cramming too much into reporting, and launching massive projects to tie it all together. All that operational data won’t change the decisions much.
Decision data lives in periods of time. Orion describes a decision as standing between two mirrors, one showing the past and one showing the forecast. The goal is the least amount of data that makes those mirrors clear enough to act on, which means building a culture that accepts a dirty mirror. The same way nobody looks in the bathroom mirror, spots a smudge, and hires a cleaning crew before deciding they should probably just skip dessert.
The Change is Here
Construction has built for thousands of years without needing to change much. As Orion so plainly states, that era is over. We can't take on the projects society needs next on top of messy spreadsheets and half-implemented systems. Which is why it’s time to start collaborating with curiosity. Construction needs to finally be willing to answer the question "why would you ever do it that way?" without reaching for "we’ve always done it that way."
The next time someone new asks that question, don't end the conversation. Start it.
Construction is cool, tell your friends!