Case Studies

Three immersive scenarios for thinking through AI in the built environment — grounded, fictional, and deliberately extreme. Each ends with a role simulation: step into the decision yourself.

These are teaching scenarios, not reports on specific named projects. They're designed to surface the judgment calls AI adoption really demands.

Scenario type: Grounded

The structural firm that designed out the steel

The setup

A mid-size structural engineering firm is bidding on a mid-rise office building. Margins are thin and the client is cost-sensitive. A junior engineer suggests running the lateral system through a generative-design tool to explore bracing options.

Where AI comes in

The team frames the structural problem as constraints (loads, code limits, column grid) and goals (minimize steel tonnage, maintain constructability). The tool returns dozens of configurations. A few cut material noticeably; some are elegant but impractical to fabricate. The engineers shortlist three and hand-verify each.

What it teaches

AI widened the option space far beyond what the team would have drawn by hand — but the value came from engineers framing the problem well and verifying the output. The tool proposed; licensed professionals disposed. Material savings on repetitive structural systems compound across a project.

Role simulation

You're the lead engineer. The most efficient AI scheme saves the most steel but uses connections your usual fabricator hasn't built before. The bid is due tomorrow. What do you do?

Step into the role and decide before you open the reflection.

There's no single right answer, but a sound path: don't gamble the bid on an unproven detail under time pressure. You might submit the second-best scheme (proven connections, still a real saving) and flag the more aggressive option as a value-engineering opportunity to explore with the fabricator after award. That captures most of the upside while keeping constructability and your professional liability intact.

Scenario type: Fictional

Atelier Verde, the AI-native interior studio

The setup

Atelier Verde is an invented boutique interior-design studio of six people. Its founder decides to rebuild the whole front-end of the practice around AI: concept renders, material sourcing, mood boards, and first-draft client proposals all run through AI tools before any human touches them.

Where AI comes in

Turnaround on initial concepts drops from days to hours. The studio pitches more prospects and wins more work. But two problems emerge: junior designers stop developing their own visual instincts, and a few clients feel the early concepts are 'samey' — the AI gravitates toward fashionable defaults.

What it teaches

AI can transform throughput, but speed without cultivated judgment is fragile. The studios that thrive use AI for velocity while deliberately protecting the human craft — taste, originality, and the designer's developing eye — that clients actually pay for.

Role simulation

You're the founder. Throughput is up and revenue is growing, but your two best juniors say they feel like 'prompt operators' and are losing their design skills. How do you respond?

Step into the role and decide before you open the reflection.

A thoughtful move: separate the work into 'AI-accelerated' tasks (variations, sourcing, admin) and 'craft' tasks (the core concept, the unexpected move) and protect time for the latter. Have juniors critique and redirect AI output rather than just accept it, and review work for originality, not just speed. The goal is AI as an amplifier of a designer's judgment, not a substitute for developing it.

Scenario type: Extreme

The lights-out construction site of 2040

The setup

A deliberately extreme thought experiment: imagine a mid-rise built by an almost fully autonomous site. AI schedules every task, robots place every component, computer vision tracks every tolerance, and a digital twin updates in real time. A skeleton crew of humans supervises from a trailer.

Where AI comes in

In the fantasy, productivity and precision soar and on-site injuries approach zero. But push on it: who is liable when an autonomous system makes a structural error at 3 a.m.? What happens to the trades and the apprenticeship path that produces master builders? What new failure modes appear when the digital twin and reality silently diverge?

What it teaches

Extreme scenarios are useful precisely because they break our assumptions. Pushed to the limit, the hard questions about AI in construction turn out not to be technical — they're about liability, labor, accountability, and what we're willing to stop understanding ourselves.

Role simulation

You're a regulator drafting the first code for autonomous construction sites. What is the one requirement you would not compromise on?

Step into the role and decide before you open the reflection.

Reasonable people will choose differently, but a defensible answer centers on meaningful human accountability: a named, licensed professional who is responsible for the work and has genuine authority to stop it — not a rubber-stamp role. Autonomy can scale, but accountability shouldn't dissolve into the system. Many would pair that with mandatory divergence-detection between the digital twin and the as-built reality.

Learning Program