The biggest civil engineering group in the US just put it in writing: AI can do a lot of the work, but it never gets the final say. A person does.
What happened
The American Society of Civil Engineers (ASCE) released AI RACE, short for AI Roadmap for Application in Civil Engineering. ASCE announced it on Oct. 19. Its board backed it on Oct. 17.
It's a plan for how AI should be used across a project's whole life: planning, design, construction, operations, and maintenance.
The main points:
- It has 11 pillars and moves from AI that helps you today to AI that runs more on its own later.
- It sets five "human red lines" that apply no matter how good the AI gets.
- Accountability never gets handed to the AI. AI can't seal, sign, or certify work.
- No AI output gets used without a check by a competent person, and the bigger the risk, the deeper the check.
- People set the limits on anything automated, and people keep the override.
- Where there's no licensed engineer on the hook, the roadmap says a named, competent person and the company using the AI stay accountable.
- ASCE plans training, certificates, and webinars, plus help for smaller firms and public agencies.
ASCE's 2026 president, Marsha Anderson Bomar, put it simply: "AI is a tool."
Why it matters to you
Read that red line again: where nobody has a license on the line, a named person is still accountable.
On a lot of jobs, that named person is the lead, the foreman, or the PM.
So if AI writes your daily report, your lookahead, or your punch list, it's still your report, your lookahead, and your punch list. Your name is on it. "The AI said so" won't hold up with your super, your PM, or an inspector.
This is also where the industry is going. The roadmap talks about AI in project controls, scheduling, and field data capture. Those land on your desk as you move up. The people who know how to check AI work fast are the people who get trusted with it.
How to use it on the job
1. Use AI for the first draft, never the last word. Let it clean up notes or build a checklist. Then read every line before it goes out.
2. Check harder when the stakes go up. A typo in a crew text is no big deal. A wrong torque value, quantity, or date on a safety or quality record is. Match your check to the risk.
3. Keep a quick record. For anything that matters, note what you asked the AI, what you changed, and who checked it. ASCE recommends a one-page "AI-Use Record" for engineers. A short note in your daily log does the same job for a lead.
4. Know your company's rules. Your company's policies and safety program still come first. If there's no AI policy yet, ask your PM before you put job data into a tool.
Try this prompt the next time AI writes something for you:
Here is a draft I got from an AI tool: [paste draft].
I'm the crew lead and my name goes on this. List every number, date, quantity, spec, and safety claim in it so I can check each one.
Flag anything that looks guessed or that I should confirm with the plans, specs, or my PM. Keep it short.
The bottom line
AI is getting a seat on the job. It's not getting the pen. If your name is on it, you check it.
More: The daily report your PM actually reads · AI and field apps worth using
Source: ASCE, Oct. 19, 2026 · AI RACE roadmap (PDF)