AI can help organize calculation notes, draft a client explanation, or locate a passage in material the firm is authorized to use. Treat its output as a starting point. A fluent answer can still rely on the wrong assumptions, units, code edition, boundary conditions, or load combinations. It may invent a reference or miss a detail that changes the result.
Set a clear boundary in the firm’s procedures: AI may assist with defined support tasks, while a qualified engineer independently checks technical content that could affect a deliverable. For calculations, check inputs, equations, units, methods, code basis, and results against the source record. For recommendations, confirm site conditions, criteria, constraints, and the professional reasoning behind the advice. The reviewer should be able to reproduce the result without relying on the model’s explanation.
Owners can put this into practice by choosing one low-risk task, such as turning approved notes into a plain-language draft. Write down the task’s permitted inputs and output, who reviews it, and what the reviewer must verify. Use a firm-approved system, remove unnecessary client details, and preserve the original source documents. Ask the technical lead to define how corrections are recorded and who signs off before anything reaches a client. If staff cannot trace a statement to a reliable source or independently verify it, leave it out.
Keep the review record with the project file. It can identify the AI-assisted draft, source materials, reviewer, checks performed, corrections, and final approval. If the calculation changes, retain the revised inputs and show how the change affected the result. This gives the project team a practical record of what was checked and supports the firm’s normal quality process.
Before adopting a tool, ask the firm’s attorney or security lead how its terms address retention, training, access, and deletion. Ask the client or check the contract when project information may be restricted. Start with a trial using approved, non-sensitive material. Have the technical lead review sample outputs, record recurring errors, and decide whether the task’s checking burden makes the workflow worthwhile.
AI does not hold a license or take professional responsibility. Licensed PE judgment and local codes govern.
This is general education, not engineering, legal, tax, or investment advice.
