A construction site in 2025 looks almost identical to one from 1975. Same hard hats. Same paper plans rolled out on a folding table. Same guys arguing about whose duct is in whose way.
Software has been reinvented a dozen times since then. Construction? Basically nothing. Which is strange, if you think about it. We build buildings out of the same materials we've used for decades, using processes designed around limitations that no longer exist.
I think that's about to change. And I don't mean incrementally. I mean the whole process of putting up a building is going to be reinvented, and it's going to happen faster than most people expect.
I'm building a company in this space, DrawScale, so I'm biased. But let me explain why I believe this, and you can decide for yourself.
Here's how you build an apartment building today. You hire an architect. Months of design. Then structural engineers, mechanical engineers, electrical engineers, plumbing engineers. Each makes their own drawings. Then an estimator sits down and measures every single line to figure out materials and costs. Then you call suppliers. Then crews spend a year or more actually building it.
The whole thing takes two years and is shockingly manual at every step.
Now here's what I think it will look like. You describe what you want. A generative model produces a complete 3D building in minutes. Zoning codes, height limits, fire egress, ADA requirements—those are just constraints fed to an optimizer. The building runs through finite element analysis automatically, adjusting beam sizes and adding shear walls until the structure works. From there, construction documents are just projections of the 3D model. The geometry already encodes every wall, outlet, and fixture, so material quantities fall out directly. A scheduling system optimizes procurement and deliveries. Robots handle the heavy structural work. Humans focus on finish work that actually requires judgment.
That sounds crazy. But each piece is either already working or obviously coming soon.
The software parts are close. Automated takeoff, extracting quantities from drawings, is happening now. That's what we're building at DrawScale. Generative design is further out but not far. The algorithms work. The bottleneck is training data and edge cases.
The robotics parts are further. A construction site is one of the hardest environments for a robot. It changes every day. It's dusty, wet, uneven. The lumber is warped. The concrete cured wrong. Robots will handle excavation, concrete, steel, bricklaying. But running wire through a wall cavity, soldering a joint in a tight spot, installing trim that actually looks good? That's much harder. Probably one of the last things robots learn to do.
And there are bottlenecks that aren't technical at all. Before you build anything, you submit plans to a building department and wait weeks for a human reviewer. AI can prepare perfect submissions. It can't make the reviewer read faster.
There's another constraint. You can't just pick a plot on Google Maps and ask a model to design a building. You need to know what's underground. Soil type, water table, utilities, contamination. The AI designs in hours. You wait weeks for the geotech report. The software speeds up. The physical world doesn't.
So why is any of this happening?
Construction has always been labor-intensive. That's not because the industry is backwards—it's because labor was cheap enough that automating didn't make economic sense. Why spend millions on a bricklaying robot when you can hire a crew? The math didn't work.
But the math is changing. The average age of tradespeople keeps climbing. Fewer young people enter the trades. Meanwhile, demand keeps increasing—housing shortage, infrastructure bills, reshoring of manufacturing. When labor gets scarce, it gets expensive. When it gets expensive enough, the calculus flips. Suddenly that robot looks like a good investment.
The decision to automate is pretty simple math. If the amortized cost of a machine is less than the wages it replaces, you buy the machine. For decades, that equation didn't favor automation in construction—labor was cheap and available, machines were expensive and limited. But wages are rising, labor is scarce, and the machines are getting better. At some point the inequality flips, and when it does, the transition will be fast.
This is why I think it matters. There's a massive housing shortage. If AI can cut design from months to days, reduce estimation errors, optimize procurement, and let smaller crews do what used to require larger ones—buildings get cheaper. More people can afford homes. Schools get built in communities that need them.
The full pipeline is huge. Generative design, structural simulation, plan production, AI takeoff, procurement, logistics, robotics. No single company will own it. It'll be dozens, connected by data standards and APIs that don't exist yet.
I'm focused on takeoff because it's closest to being solved and immediately valuable. But I'm building with the full pipeline in mind. Every customer correction is training data, and that compounds.
Construction has been around for thousands of years and hasn't changed much. But now you've got AI that can read blueprints, robots that can pour concrete, and an industry that desperately needs to build more with fewer people. Something has to give.
We're going to build things differently. And they're going to be better.