✉ Envelope #50: NCSEA AI Townhall Recap

Author

Andy Lin
May 02, 2025

Good morning! Andy from Back of the Envelope here.

In case you missed it, the NCSEA Foundation recently hosted an AI-focused town hall for structural engineers. I couldn’t make the live session, but I watched the full replay so you don’t have to (unless you want to).

According to NCSEA, over 800 attendees joined the event!

Things kicked off with a few audience polls (no screen share of the results, but the audio revealed enough).

Then came introductions: host Emily Guglielmo, the AI Grant Team (linked below), followed by a quick look at NCSEA’s AI strategic roadmap.

The remaining 45 minutes or so were structured around Q&A, covering four main topics:

  1. Professional Practice

  2. Education & Training

  3. Workflow, Collaboration & Technology

  4. Other AI Questions

Let’s dive in!
(Estimated read time: ~6 minutes… oops slightly over 5, but it’s because of all the actual questions from the webinar!)

Audience Polls

Poll #1: How big is your firm, and how long have you been practicing? (00:00:44)

Responses were well-distributed across firm sizes. Most participants had 11–15 or 26+ years of experience.

Poll #2: "Have you used AI to assist in structural analysis, design or documentation tasks yet?" (00:02:24)

  • ~8% said "yes, regularly"

  • ~13% said "no, and never will"

  • The rest said either dabbling or planning to start soon

My take: The question was probably a little vague… people interpret "AI" in different ways. E.g., I use ChatGPT/NotebookLM/Gemini/Grok regularly, but not for traditional calcs or documentation.

Poll #3: "What skill do you think will be most important for structural engineers in this AI era?" (00:03:41)

Winner: Critical thinking and engineering judgment.

My take: This is likely common sense for most of us right now but we should continue to reinforce this to the new generation of engineers who will almost certainly grow up with AI.

Poll #4: "What worries you the most about AI?" (00:04:42)

Top concern: safety and liability. Then loss of traditional skills, then ethics and job security. About 9% weren’t worried at all.

Poll #5: "In 10 years, what percent of the work you do today do you think will be AI-assisted?" (00:05:37)

Most common answer: 10–50%.

Topic 1: Professional Practice

1/ "What are the legal ramifications of using AI and how can we ensure engineers use AI properly?" (00:16:21)

  • Engineers of record still hold responsibility.

  • AI must be transparent, explainable, and thoroughly validated.

  • Use AI to inform decisions, not to make them.

2/ "How does the use of AI affect intellectual-property rights?" (00:20:43)

  • Use enterprise-grade AI platforms that guarantee data privacy.

  • The biggest risk is staff unknowingly exposing data in public tools like ChatGPT.

My take: Enterprise ChatGPT requires 150 seats at $60/month… probably not practical for most firms. Teaching staff to disable AI training in tools or use team plans is more realistic.

There may be other alternatives like Microsoft Copilot (which uses OpenAI in their backend currently, but not always the latest model). Would love to know how people are approaching this.

3/ "Assuming we’ve figured it out and we’re letting our engineers use AI securely… what are the implications of AI on the professional practice, and specifically how will AI affect how we train young engineers?" (00:23:55)

  • Senior engineers will shift to more reviewer/strategy roles.

  • Juniors still need foundational skills (hand calcs, engineering reasoning).

  • Stronger mentoring and QA/QC processes will be essential.

My take: In many ways, this builds on what we’re already doing… just with greater intentionality and structure.

4/ "What are practical applications of AI in structural engineering today?" (00:29:55)

As consumers (non-coders):

  • Drafting proposals, reports, and field notes

  • Summarizing RFPs

  • Writing RFI responses and emails

  • Early-phase concepting and design options

As developers (coders/technical users):

  • ML models predicting member strength

  • Training LLMs on internal docs to answer firm-specific queries

  • Research underway: generating shear wall layouts from plans

My take: This is the one that everyone (myself included) is actively exploring. I haven’t played around with early-phase concept/design options though. Anyone has? Let me know your thoughts.

Topic 2: Education & Training

5/ "How can engineers get started learning AI and what are the best resources for training—coding class, course, example problem?" (00:33:52)

  • Begin with ChatGPT, Claude, or Perplexity.

  • Use them for tasks you already do (summarizing, planning, writing).

  • If you're technical, explore free courses and challenges. (links at the end)

Also subscribe to Back of the Envelope! I’ll be sharing more AI tips and use cases.

6/ "What AI education do you think should be the standard for practicing engineers?" (00:39:28)

  • If you're a user: learn the strengths, limitations, and how to validate outputs.

  • If you're building tools: get familiar with APIs, basic coding, and how LLMs work.

Topic 3: Workflow, Collaboration & Technology

7/ "Can you share some examples of how AI can help engineers who use Revit, CSI, Bentley via their APIs?" (00:42:43)

  • If you know a bit of Python or C#, LLMs can write API scripts to automate common tasks.

One of the panelist mentioned, “we have a lot of information, but we are just not quite ready to share it today"… The recording doesn’t show who was speaking at the time, so I am not sure who said it. Does anyone know? I’d like to reach out and follow-up to see if I could share some of the findings.

On LinkedIn, I’ve come across a few people experimenting with things like connecting ETABS to Claude using MCP (model context protocol, which is gaining traction lately).

In one example, the user typed a prompt like “model out my named; use frame members,” and ETABS did just that (something to that effect, I don’t remember the exact thing). Regardless, it’s an interesting first step.

I’ve also seen demos where people hook up Revit to an LLM so you can “talk” to your model… asking it questions directly and getting responses.

At first glance, some of the stuff just seem like you could do them way faster and more efficiently by just doing the actual thing… like we normally would, instead of “talking to it and hoping it’ll do the thing.” But I suppose we do need to start somewhere to build it up from there.

8/ "How could AI be used in seismic design of structures?" (00:46:44)

  • Research use cases include:

    • Nonlinear response prediction (ML-based)

    • Real-time seismic forecasting

    • Automated shear wall layout generation

  • Promising, but still in early research phases.

9/ "Thinking about a consumer vs. producer lens: can you speak to the maintenance burden of those tools that are developed in-house versus off-the-shelf tools?" (00:51:37)

  • Building tools = control + customization, but high maintenance.

  • Buying tools = easier to adopt, but limited flexibility.

Topic 4: Other AI Questions

10/ "How soon do you think AI will replace some of our current work, particularly in the residential field or other repetitive tasks?" (00:55:02)

  • Repetitive and low-risk tasks are most likely to be impacted first.

  • But full replacement is still years away.

11/ "How can you create your own AI tools or agents" (00:55:53)

  • The panel dropped links in chat to low-code resources (links at the end)

  • Main idea: You don’t need to be a full-stack developer to start building small AI helpers.

That’s it for now. I heard there were lots more questions in the live chat, but I wasn’t there so I missed it.

If any of this resonated (or didn’t), let me know.

And if you're trying out AI in your firm, I’d love to hear what’s working (or what’s frustrating).

Until next time!

PS.

If you are working on a steel building and you haven’t heard of Durafuse, you might want to check it out.

It is a proprietary steel moment frame connection that offers high ductility and fast recovery after an earthquake.

And they’ve got your back with full design team support related to the connections: calcs, plan check comments, RFI responses, and shop drawing reviews.

Click here to learn more: https://go.sehq.co/durafuse

PPS.

Links:

Came across this… it feels pretty relevant to the current state of generative AI:

source: Tarek Ghazzaoui (The BIM Consultant)

Hit reply. Or else…

Or else… nothing. It just makes my day to hear from you. Let me know what you thought of this week’s newsletter by replying. I read every response!

Oh, and tell others to sign up here.

Reply

or to participate.