Claude for Structural Engineers: The Practical Workflow

In 2024, Engineers and Geoscientists BC became the first provincial engineering regulator in Canada to issue a formal practice advisory on AI in professional practice, and PEO Ontario adopted it the following year. The guidance is written and the expectations are set. This course is the practical side of that conversation: how to apply Claude, Anthropic's AI assistant, to real structural engineering work while meeting the professional responsibilities that come with your engineering seal.

The course focuses on four high-value applications — drafting reports and documentation, writing specifications, researching and navigating codes, and generating calculation scripts — with live demos and exact prompts for each. It begins by building an accurate mental model of how a large language model actually works, so you understand why Claude can produce confident but incorrect output and where its real limits lie, including hallucination, no access to live data, unreliable inline arithmetic, and no ability to read CAD or geometry. From there it teaches structured prompting using role, context, task, and format, along with Claude's XML tag support, and shows how to set up a persistent Claude Project that loads your standards, jurisdiction, and output preferences automatically.

Throughout, the course keeps the engineer in the decision-making seat. It draws a clear line between using Claude to navigate and understand codes, which is research, and using it to determine code compliance, which is professional judgment that requires reading the primary source. It devotes a full module to professional ethics, liability, and the seal, mapping the EGBC advisory's principles of transparency, explainability, and interpretability onto a concrete review-and-documentation standard, and addresses confidentiality obligations around client data. It closes with a baseline AI governance policy for a firm and a 90-day adoption plan that starts with the lowest-risk, highest-value tasks and builds toward a verified, documented workflow.

This course is designed for practising structural engineers, from sole practitioners to firm staff and those supervising junior engineers, in Canadian practice. No programming background is required, though you must be able to read, test, and take responsibility for any output before it reaches a deliverable.

Learning objectives

By the end of this course, participants will be able to:

1. Explain how a large language model generates output and identify Claude's principal limitations, including hallucination, training cutoff, unreliable inline arithmetic, and lack of CAD or real-time data access.
2. Construct structured prompts using role, context, task, and format, and apply XML tags to improve the consistency and usability of Claude's output for technical work.
3. Apply Claude to core documentation tasks — assessment narratives, specifications, and RFI responses — while verifying every factual claim, code reference, and figure before use.
4. Distinguish research use of Claude for code navigation from code-compliance determination, and verify all code references against the primary source.
5. Configure a Claude Project with custom instructions and an appropriate knowledge base, observing copyright limits on licensed standards and confidentiality limits on client data.
6. Generate and validate Python, VBA, and AutoLISP scripts by testing outputs against known hand-calculated results and documenting the verification.
7. Apply the EGBC AI practice advisory's principles of transparency, explainability, and interpretability, and describe the review, documentation, and governance practices that preserve professional accountability under the engineering seal.