Why Steel Frame Construction is AI-Ready

Everyone's talking about AI disrupting white-collar jobs. Software engineers, lawyers, writers. But there's a $10 trillion industry that's barely paying attention: construction.

Construction is the second-least digitized industry on Earth, according to McKinsey. Only agriculture ranks lower. And yet, AI is about to hit construction harder than most people expect.

But here's what most people miss: AI won't improve all construction methods equally. Some are positioned to benefit enormously. Others will barely change. The difference comes down to one word: verifiability.

The Verifiability Framework

In November 2025, Andrej Karpathy—one of the founding members of OpenAI and former AI Director at Tesla—shared a framework for understanding which tasks AI can automate:

"Software 1.0 easily automates what you can specify.
Software 2.0 easily automates what you can verify."

The old programming paradigm (Software 1.0) automated tasks where you could write explicit rules. Bookkeeping, typing, calculations. If you could specify the algorithm, a computer could do it.

The new paradigm (Software 2.0) is different. We train neural networks by specifying objectives and letting the AI figure out how to achieve them. The AI learns by practicing—trying things, getting feedback, improving.

For this to work, the task needs to be verifiable. The AI needs to be able to:

• Reset — Start new attempts (the environment is resettable)
• Iterate quickly — Make many attempts fast (the process is efficient)
• Get scored — Know if an attempt worked (the outcome is rewardable)

This is why AI excels at chess, code, and math. Every attempt can be evaluated automatically. The AI can practice millions of times and get immediate feedback.

It's also why AI struggles with tasks that require "real-world knowledge, state, context, and common sense." There's no automated way to score whether a novel is good or a business strategy will work.

Applying This to Construction

Now apply this framework to construction. Specifically, to CAD and structural design.

Can AI "practice" building design? Let's check:

Every row is green. Structural engineering and CAD design are highly verifiable. An AI can generate a building design, run it through automated checks, and know immediately whether it works.

This is rare. Most real-world tasks aren't this clean. But construction—specifically, the digital design phase—is one of the most verifiable domains outside of pure software.

Why Steel Framing is Different

But here's where it gets interesting. Not all construction methods are equally verifiable.

Verifiability depends on consistency. The more predictable your inputs, the easier it is to score outputs. And this is where steel framing has a structural advantage.

Steel framing operates in a finite state space. Every stud is identical. Every connection is specifiable. Every load path is calculable. There's no biological variation, no weather dependency, no site-specific mixing.

This is why AI can "practice" steel frame design. The environment is:

• Resettable — Generate a new design instantly
• Efficient — Simulate thousands of variations per hour
• Rewardable — Structural pass/fail, cost minimization, code compliance all automated

Wood has natural variation. Concrete depends on mix, weather, and curing. Masonry is inherently manual. These methods can still benefit from AI in scheduling, logistics, and project management. But the core design-to-fabrication loop? Steel has the advantage.

This Isn't Theory—It's Happening

At AU 2025, Autodesk announced neural CAD foundation models for Forma and Fusion. These models can:

• Generate CAD geometry from text prompts
• Transition from early design concepts to detailed building layouts
• Reason directly about geometry and structural requirements

The market agrees this is real. AI in construction is projected to grow from $3.99 billion in 2024 to $11.85 billion by 2029—a 24% compound annual growth rate.

76% of industry leaders already trust AI. 75% are increasing AI spending over the next 2-3 years.

The transition is happening. The question isn't whether AI will transform construction. It's which construction methods will benefit most.

The Feedback Loop Advantage

Here's the workflow that's emerging:

Natural language in, verified building design out. The architect or engineer becomes a prompter and reviewer, not a draftsperson. The AI does the iteration. The verification layer catches errors.

This only works when the verification is automated and reliable. Physics simulations don't lie. Building codes are explicit. Material costs are calculable.

And it only works well when the inputs are consistent. When every component behaves predictably. When the search space is bounded.

That's steel framing.

What This Means

If Karpathy's framework is correct—and the evidence suggests it is—then we should expect:

• Steel framing costs to drop 20-30% by 2030 — Not because steel gets cheaper, but because AI-optimized design, automated fabrication, and digital coordination eliminate waste and delays.
• The gap between steel and other methods to widen — Methods with more variability can't benefit as much from AI optimization. Their costs stay flat while steel drops.
• 70% of contractors without a tech roadmap to struggle — The industry survey data already shows most contractors aren't preparing. When AI hits, they won't see it coming.

Steel framing isn't just a material choice. It's a bet on which construction method AI can optimize first.

The framework says steel wins. The market is moving that direction. The technology is arriving.

The question is whether you're positioned to benefit.

Sources

• Karpathy, Andrej. "AI's Impact on the Economy." November 2025.
• McKinsey Global Institute. "Digital America: A Tale of the Haves and Have-Mores."
• Mordor Intelligence. "AI in Construction Market Size & Share Analysis." 2024.
• Autodesk. "Upcoming 3D Generative AI Foundation Models." AU 2025.
• Autodesk. "State of Design & Make Report." 2024.
• Stanford University. "AI Frameworks for Structural Steel Optimization."
• World Economic Forum. "Shaping the Future of Construction."