A structural question has a short version: will it hold, and who says so. Artificial intelligence (AI) is changing how engineers answer the first part. The profession has been direct about the second part.
The societies’ position
The American Society of Civil Engineers (ASCE) adopted Policy Statement 573 on July 18, 2024.1 It states that “AI cannot serve as a replacement for the professional judgement of a licensed Professional Engineer.”1 It also states that “AI cannot be held accountable.”1
The National Council of Structural Engineers Associations (NCSEA) reported on its research on April 6, 2026, and reached the same point from the firm’s side. It names three imperatives for firm leaders. The first is that “AI adoption must be governed, not improvised, with an emphasis on accountability, professional judgment, and trust.”2
What inspection tools can measure
On August 3, 2026, TechXplore published an announcement of a study from Seoul National University of Science and Technology.3 The researchers built a computer vision framework that tracks structural damage over time from drone images taken during routine bridge inspections.3 The system was validated over 120 days on one in-service concrete bridge.3 It tracked the progression of cracks, spalling (concrete that breaks away from the surface), and water leakage, and it measured damaged areas with a maximum error of 4.61% compared with manual measurement.3
The researchers state a limit. The method is best suited to relatively flat structural components, and its accuracy might differ on highly curved surfaces.3 The study is about bridges. It does not cover houses.
The useful idea for any structure is the comparison over time. One photograph shows a crack. Two measurements, months apart, show whether the crack is growing.
What a study of AI agents found
A preprint posted to arXiv on July 16, 2026 tested AI agents on structural engineering workflows.4 A preprint is a research paper that has not yet passed peer review. The authors built a benchmark of 150 controlled scenarios, and a trial succeeded only when every required check passed.4
General-purpose agents passed the model check in 87.0% of outcomes but reached only 22.0% end-to-end success.4 The authors report that their own system reached 82.9%.4 They built both the benchmark and that system, so an independent test is still needed.
The gap between 87.0% and 22.0% is the lesson. An AI agent can produce a model that looks right and still fail the complete engineering check.
What this means for an owner
- Measure movement over time. A structure that settles, cracks, or leaks needs repeat measurements. This applies to a slab house and to a house on pilings.
- Ask who is responsible for the answer. Where a project needs engineered drawings, a licensed engineer remains responsible for them. The ASCE policy explains why.
- Treat a fluent answer with care. A report that reads well is not proof that the analysis is complete.
We repair foundations, framing, and load paths, and we assess waterfront and pile-supported structures. We find the cause of the movement before the repair starts. Call (985) 255-2435.
American Society of Civil Engineers, Policy Statement 573, “Artificial intelligence and engineering responsibility,” adopted July 18, 2024. https://www.asce.org/advocacy/policy-statements/ps573---artificial-intelligence-and-engineering-responsibility ↩︎ ↩︎ ↩︎
NCSEA, “Structural Engineering Report Explores the Future of AI Adoption, Workforce, and Teams,” April 6, 2026. https://www.ncsea.com/news-post/structural-engineering-report-explores-the-future-of-ai-adoption-workforce-and-teams/ ↩︎
Seoul National University of Science and Technology, announcement of an AI framework for long-term bridge damage monitoring, published by TechXplore, August 3, 2026. https://techxplore.com/news/2026-08-ai-framework-term-bridge.html ↩︎ ↩︎ ↩︎ ↩︎ ↩︎
“StructureClaw: Traceable LLM Agents and an Executable Benchmark for Structural Engineering Workflows,” arXiv:2607.14896, submitted July 16, 2026. https://arxiv.org/abs/2607.14896 ↩︎ ↩︎ ↩︎ ↩︎