# Why "AI For Engineers" Is The Wrong Course **By Dan Cumberland** · Published May 7, 2026 · Categories: AI Strategy > Architecture isn't engineering. Your daily work involves the International Building Code, local amendments, HSW (health, safety, welfare) liability, code... ## What Makes Architecture Different From Engineering Architecture isn't engineering\. Your daily work involves the International Building Code, local amendments, HSW \(health, safety, welfare\) liability, code compliance verification, multi\-disciplinary coordination, and design constraints that no "AI for Engineers" syllabus mentions\. NCARB has been clear that architects must retain "responsible control" over design decisions, including any AI\-assisted output\.[2](/blog/blog-ai-for-architects-course#ref-2) The AIA's recent industry research goes further— 94% of architects worry about AI inaccuracy and unintended consequences, and 90% worry about transparency\.[1](/blog/blog-ai-for-architects-course#ref-1) These aren't paranoid concerns\. They're a direct read on what's at stake when you stamp a drawing\. Software engineers can ship a bug, find it in QA, and patch it\. You can't recall a parking garage\. Three things make architecture's training needs structurally different: - **Building codes are non\-negotiable\.** IBC, ADA, energy codes, local amendments— these aren't optional context, they're the framework every design decision sits inside\. - **HSW liability is yours alone\.** Health, safety, and welfare obligations are tied to your license\. An AI hallucination that makes it past your review is your signature, not the model's\. - **Workflows are visual and multimodal\.** Concept sketches, parametric models in Grasshopper or Dynamo, BIM coordination in Revit, renderings, specs— this is not a code\-centric pipeline\. A regulated\-domain analysis from Cara Life Sciences put it bluntly for life sciences and the same logic applies here— generic AI systems create "siloed" capabilities that lack the explainability and auditability regulated professions require\.[3](/blog/blog-ai-for-architects-course#ref-3) Pharma figured this out two years ago\. Architecture is figuring it out now\. One AI suggestion that misses a code requirement doesn't just waste time\. It creates liability\. ## Why Generic AI Training Fails Architects Generic AI courses lead with capabilities\. "Here's what GPT\-5 can do\. Here's what Claude can do\. Here's how diffusion models work\." That's backwards for your firm\. The right starting point is your workflow— concept → design development → code check → construction documents → coordination → delivery— and then a careful question about which AI tools actually fit into that process\. LogRocket's research on engineering teams found that organizations taking the workflow\-first approach saw 30–50% velocity gains, while leadership\-driven, capability\-first rollouts mostly fail to drive adoption at all\.[4](/blog/blog-ai-for-architects-course#ref-4) The data from engineering education tells the same story from the student side\. An ASEE study of engineering students found that 43% never used AI for coursework even after being formally taught it, and another 34% rarely did\.[5](/blog/blog-ai-for-architects-course#ref-5) Capability training without workflow integration produces awareness, not behavior\. There's a deeper issue underneath this\. Just because something is easy doesn't mean it's good\. And we have to find out how to make it good *and* easy\. Generic courses optimize for easy\. They give you a tour of the tools and a feeling of motion\. They don't help you make AI good for the way an architect actually thinks\. A useful frame— capability\-first training is like selling someone a power drill when they need to build a house\. You're teaching the tool, not the craft\. A separate study from the University of Illinois made the design\-versus\-calculation distinction concrete\. ChatGPT scored 100% on closed\-form engineering calculations and dropped to D\-level performance on open\-ended design projects\.[6](/blog/blog-ai-for-architects-course#ref-6) Calculations are exactly the domain where generic AI training does its best work\. Design, judgment, and multi\-constraint synthesis— the substance of architectural practice— are exactly where it underperforms unless training is built for the domain\. This is where a real [AI strategy framework](/services/ai-strategy/) starts to matter more than any specific tool list\. Architects don't need another course on prompts\. They need a way of thinking about which parts of their workflow AI can amplify and which parts it absolutely cannot\. ## How Successful Architecture Firms Are Doing It Architecture firms getting real AI ROI start in a specific order\. They map their workflow first\. Then they look at where AI can amplify a step\. Then— and only then— do they pick a tool\. No magic platform\. No vendor demo\. Just clear thinking about how the firm actually works\. That maps almost exactly to what Autodesk's industry research describes as the design\-centric AI workflow— generative design, real\-time daylighting and energy simulations, visualization, BIM coordination, and project automation as integrated stages, not isolated features\.[7](/blog/blog-ai-for-architects-course#ref-7) 76% of AEC firms plan AI investment, but the ones generating returns are not buying tools\. They're buying integration\. Active AI's case studies from architecture firms using this approach report meaningful operational gains— roughly 60% reduction in design time on iterative tasks, 80% fewer documentation errors, and around 3x project capacity without adding headcount\.[8](/blog/blog-ai-for-architects-course#ref-8) Those numbers are vendor\-reported and they should be read as upper bounds\. But they line up with the more conservative 30–50% LogRocket data\.[4](/blog/blog-ai-for-architects-course#ref-4) The pattern is consistent across studies\. What does workflow\-first training actually look like in practice? ```html-table
Capability-First (Generic)Workflow-First (Architecture)
"Here's what GPT can do""Here's where AI fits in your design development phase"
Tool tutorials in isolationTool integration into Revit, Grasshopper, Forma
Generic prompt engineeringCode-compliance verification prompts and review patterns
Theory + ethics + demosHands-on practice on real firm projects
One-off workshopOngoing practice with peer review and senior mentorship
``` People are the answer, not AI\. AI should amplify your firm's genius— your code judgment, your spatial reasoning, your client trust— not replace it\. The firms doing this well treat AI training as a thinking skill, not a tactical one\. This is closer to how you'd approach a new draft staff member than a new piece of software, and that's the right instinct\. If you want to see how this connects to broader change management, our work on [AI implementation](/services/ai-implementation/) treats workflow mapping as the first deliverable, not the last\. ## What Architecture Firm Leaders Should Do Next If you're evaluating an AI for architects course right now, run it through three questions before you spend any budget\. 1. **Does it start with my workflow, or with the tools?** If the syllabus opens with capabilities, models, or "what is AI," it's the wrong course\. 2. **Does it address building codes, HSW liability, and verification?** If those words don't appear, the course doesn't understand your liability surface\. 3. **Does it include structured practice and review on real projects?** One\-off webinars don't change behavior\. ASEE's data is clear on that\.[5](/blog/blog-ai-for-architects-course#ref-5) A note on budget\. Most firms have a finite training spend, and generic AI literacy courses are cheap and tempting\. But chasing pennies when you could be chasing dollars is exactly what generic training does to your training budget\. A $99 course that doesn't change a single workflow costs more than a $5,000 program that reshapes how your project teams work\. For a deeper look at the operational side, our [AI governance framework](/blog/ai-implementation-checklist/) walks through the verification and review structures regulated practices need\. Your firm's workflow is too unique for off\-the\-shelf training\. The course you actually need probably doesn't exist as a product— it has to be built around your codes, your tools, and your delivery process\. If mapping that out feels like a full\-time job on its own, that's the kind of problem an outside [implementation partner](https://dancumberlandlabs.com) can compress significantly\. Not by selling you a course\. By helping you build the one your firm actually needs\. ## FAQ ### What's wrong with "AI for Engineers" courses for architects? "AI for Engineers" courses are built around software engineering workflows— code generation, debugging, deployment, automation pipelines\. Architecture is design\-centric and regulated\. Generic engineering courses skip building codes, HSW liability, BIM integration, and the design judgment that defines architectural practice\. The capabilities you learn don't translate cleanly to your day\. ### How do successful architecture firms actually implement AI? They start with workflow mapping \(concept → design → code check → documentation → delivery\), identify the highest\-friction steps, and then evaluate which AI tools fit into those specific steps\. Implementation runs through structured practice and senior review, not one\-off training\. Firms doing this report 30–50% velocity gains on integrated workflows\.[4](/blog/blog-ai-for-architects-course#ref-4) ### What should an architecture\-specific AI course actually cover? At minimum— generative and parametric design integration \(Forma, Grasshopper, Dynamo\), BIM automation in Revit, code compliance verification patterns, visualization and rendering workflows, and structured judgment training on when to trust or override AI output\. Compliance, HSW, and explainability should be woven through every module, not bolted on at the end\. ### Why does liability matter so much in architecture AI training? Because the architect of record signs the drawings\. An AI\-generated detail that violates an energy code or an ADA requirement doesn't become the model's problem when it's built incorrectly— it becomes yours\. Training that doesn't teach verification, source\-checking, and override patterns is leaving the most important skill out\. ## References 1. American Institute of Architects \(AIA\), "Architects Are Excited About Potential, Concerns Abound" \(2025\) — [https://www\.aia\.org/aia\-architect/article/architects\-are\-excited\-about\-potential\-ai\-concerns\-abound](https://www.aia.org/aia-architect/article/architects-are-excited-about-potential-ai-concerns-abound) 2. NCARB, "How AI Is Reshaping the Architecture Industry" \(2025\) — [https://www\.ncarb\.org/blog/how\-ai\-reshaping\-the\-architecture\-industry](https://www.ncarb.org/blog/how-ai-reshaping-the-architecture-industry) 3. Cara Life Sciences, "Why Generic AI Systems Are Not Suitable for Life Sciences" \(2025\) — [https://www\.caralifesciences\.generiscorp\.com/why\-generic\-ai\-systems\-are\-not\-suitable\-for\-life\-sciences/](https://www.caralifesciences.generiscorp.com/why-generic-ai-systems-are-not-suitable-for-life-sciences/) 4. LogRocket, "Engineering Team AI Training: Building Skills, Not Just Tools" \(2025\) — [https://blog\.logrocket\.com/engineering\-team\-ai\-training/](https://blog.logrocket.com/engineering-team-ai-training/) 5. ASEE, "Examining the Opportunities and Challenges of Using Artificial Intelligence for Engineering Technical Writing Courses" \(2025\) — [https://peer\.asee\.org/examining\-the\-opportunities\-and\-challenges\-of\-using\-artificial\-intelligence\-for\-engineering\-technical\-writing\-courses\.pdf](https://peer.asee.org/examining-the-opportunities-and-challenges-of-using-artificial-intelligence-for-engineering-technical-writing-courses.pdf) 6. The Prompt Index, "Can AI Pass Engineering Classes? A Study Puts ChatGPT to the Test" \(2025\) — [https://www\.thepromptindex\.com/can\-ai\-pass\-engineering\-classes\-a\-study\-puts\-chatgpt\-to\-the\-test\.html](https://www.thepromptindex.com/can-ai-pass-engineering-classes-a-study-puts-chatgpt-to-the-test.html) 7. Autodesk, "AI in Architecture: Transforming Design and Delivery" \(2025\) — [https://www\.autodesk\.com/design\-make/articles/ai\-in\-architecture](https://www.autodesk.com/design-make/articles/ai-in-architecture) 8. Active AI, "AI for Architecture Firms: Use Cases, Benefits, and Implementation" \(2025\) — [https://www\.beactive\.ai/ai\-for\-architecture\-firms](https://www.beactive.ai/ai-for-architecture-firms) --- Source: https://dancumberlandlabs.com/blog/ai-for-architects-course/