How AI enters AEC Engineering
An engineer's take on where AI fits and where it doesn't
The AEC industry has traditionally been slow to adopt technology compared to other industries at scale1. As I enter my eighth year as a low voltage electrical engineer, the workflow I follow today is largely the same one I used when I started in this industry. While Autodesk has incrementally improved Revit over the years, MEP engineering processes have remained relatively stagnant. The nature of building design—requiring judgment, coordination, and stakeholder management, makes AI adoption uniquely challenging. This begs the question: how can AI be implemented in our workflow in a careful and meaningful way?
According to a recent study by UpCodes, 81% of AEC professionals use AI at least monthly. From my conversations with colleagues and peers across architecture, engineering, and contractor firms, most are engaging with AI at a surface level through web-based chat tools. Others have more organized AI task forces with pilot projects underway for full adoption at the enterprise level, but progress is slow due to hesitation around data security and integration value. Firms that do embrace AI prefer a slow, methodical approach, carefully integrating tools to stay competitive in a rapidly changing landscape. Ironically, firms that resist AI adoption may be exposing themselves to greater risk—individual employees may resort to uploading project data to public tools without oversight.
The spectrum of AI tool offerings in AEC runs wide. At the lowest barrier, engineers use simple AI chatbots as design aids, inputting project parameters to draft narratives, specifications, or scope summaries. Any task with a text output is the easiest entry point.
At the other extreme, AI startups are targeting full platform replacement: automated generative design that produces entire MEP layouts from code compliance alone. This approach promises to compress months of work into minutes, but struggles with the reality of client-driven design. Requirements change, floor plans shift, owners add scope. Design is iterative and human. Any solution must be flexible enough to manage continuous rounds of revision. Beyond the practical limitations, automated design poses a deeper structural risk—it removes the process of thinking through a solution. With AI doing the heavy lifting, how do we safeguard our ability to think? This fundamental shift creates a dangerous opening for inexperienced engineers to trust AI over their own judgment. It's a larger industry problem that none of us are fully prepared for.
The real opportunity lies somewhere in between—improving the weakest link in the design process: the busywork of translating design into documentation. AI should enhance the engineer’s ability to do better work, not simply replace or accelerate a narrow slice of it. Last week, I heard directly from clients asking for measurable improvements—they believe AI should help us build better and faster. Client demand signals rising expectations for firms to work more efficiently, and a real opportunity for those that adapt strategically.
The winning AI tool isn’t replacing engineers, but saving time between design decisions and model execution. I recently participated in piloting an AI tool, Arki, that manages details within Revit. This type of wedge software proved to be the easiest for engineers to integrate into existing workflows. It checks all the boxes for successful AI adoption: ease of use, demonstrable value, and minimal disruption to current workflow. By tackling one of the more tedious and repetitive aspects of BIM management, this tool solved a specific pain point and delivered measurable ROI for our team. That’s where the industry moves forward.
As firms grow more comfortable with enterprise AI software, appetite for improved output will grow alongside client expectations and the maturity of AI offerings. We are at a stage where management is curious but cautious. If AI tools continue developing at their current pace, those who are already comfortable with AI will only want more. Startups targeting complete workflow shifts will find their edge as both the solutions and firms’ openness mature. Just today, Endra—an AI startup automating MEP engineering—announced 50M Series A raise, an unprecedented signal of the industry’s readiness for a new chapter. The AEC industry may be closer to a paradigm shift than we think.
The industry won't transform overnight—and it shouldn't. The firms that win won't be the ones that adopt AI the fastest. They'll be the ones that adopt it most thoughtfully, solving real problems without dismantling what already works. The wedge is in the door. What gets built from here depends on whether we let the tools serve the work, or let the work serve the tools.
“Productivity Stagnation in the Construction Industry: An International Perspective” Goldman Sachs Research, 2 Feb. 2026, www.gspublishing.com/content/research/en/reports/2026/02/02/36f5c79a-3db6-48b3-abce-c7a732eea01a.html.