■ MODERATE RISK ■ Construction & Trades
AI will do more and more of the analysis and drafting, but a licensed human still has to stamp the drawings — and be the one who gets sued when the balcony fails. Structural engineering transforms; the professional engineer of record isn't going anywhere soon.
“AI models structural loads perfectly. Your hand calculations built character, not buildings.”
Our AI replacement risk score — how we score jobs
Day to day, structural engineers build analysis models in ETABS, SAP2000, or RISA, size beams and connections against ASCE 7 load combinations, coordinate endlessly with architects who moved a column again, review shop drawings, and visit sites to check that what's being built matches what was drawn. Then comes the defining act of the profession: sealing the drawings, which converts a pile of calculations into personal legal responsibility for the structure not falling down.
The computational middle of the job is exceptionally automatable. Generative design tools already produce optimized framing schemes; AI can size members, check code compliance, flag clashes in a BIM model, and draft connection details faster than any junior engineer. Model translation between architectural and structural software — currently hours of tedium — is being solved. Expect the analysis-and-drafting layer, historically the training ground for young engineers, to compress dramatically. Firms will run leaner, and a senior engineer with good tools will do what a five-person team did.
What resists is judgment plus liability. Software confidently analyzes whatever model you feed it, including a wrong one; the engineer's real skill is knowing which assumptions are lies — that the soil report is optimistic, that the contractor will substitute a cheaper connection, that this renovation's existing drawings don't match the building. Site conditions, existing structures, and construction-phase surprises are messy physical reality, not clean inputs. And every jurisdiction requires a licensed PE or SE to seal structural documents, a legal architecture that no AI vendor wants to inherit because inheriting it means inheriting the lawsuits. Our 32 prices in a profession that keeps its stamp while losing much of its spreadsheet.
Automatability: our editorial assessment of current and near-term AI capability
Analysis and drafting automation is accelerating now, and by the early 2030s expect substantially smaller project teams — the junior-engineer layer feels this first and hardest, since AI is eating exactly the tasks juniors learn on. Licensed engineers of record remain legally required and professionally central through 2040 and beyond, but the pipeline beneath them narrows, and the job tilts toward review, judgment, and site reality rather than model-building.
It can replace a lot of structural engineering labor — analysis, member sizing, drafting, code checks — but not the structural engineer of record. Law in every jurisdiction requires a licensed human to seal drawings and carry liability, and no software vendor is volunteering to be sued for a collapse. The role shifts from producing calculations to validating them.
Safer than most engineering-adjacent paths, with one warning: the entry-level years will look different. AI is automating the junior tasks — modeling, drafting, load runs — so early-career engineers must push faster toward licensure, site experience, and judgment-heavy work. The licensed profession is durable; the apprenticeship that used to lead there is being compressed.
The clean, computational middle: analysis model runs, member and connection sizing, code-compliance checking, drawing production, and clash detection. Generative tools that propose optimized framing schemes are already in use. The last things to go are existing-building assessment, construction-phase judgment calls, and the legal act of sealing drawings — messy reality and liability.
License first — everything else is negotiable. Then aim at the automation-resistant zones: renovation and retrofit, forensic investigation, seismic upgrade work, and construction-phase services. Learn the generative and AI tools well enough to audit their output, because 'checking the machine's work' becomes the core billable skill. Site fluency and client trust round out the moat.