Only 27% of architecture, engineering, and construction firms use AI.
Twenty-seven percent.
Goldman Sachs ranks engineering as one of the top three industries most likely to be transformed by AI. And three out of four firms haven't even started.
But here's the number that should actually scare you.
Of the 27% already using AI? Ninety-four percent are increasing their usage this year.
That's not a slow adoption curve. That's a gap ripping open in real time. One group is accelerating. The other is standing still, telling themselves "we'll look into it next quarter."
There is no next quarter. The quarter was last year.
"AI can't do structural engineering."
You're right. It can't.
AI won't design your steel connection. It won't assess your cracked lintel. It won't make the judgement call on whether that existing foundation can take the load from a two-storey extension.
And that's not the point.
The point is everything else you do that isn't structural engineering. Which, if you're honest, is about 40% of your week.
Fee proposals. Meeting notes. Status update emails. Site visit write-ups. Chasing information. Formatting reports. Searching your own server for a calculation you know exists somewhere in a folder called "Projects 2023 FINAL v2 (USE THIS ONE)."
AI is absurdly good at all of this. Right now. Today. Not in some future version. Today.
A fee proposal draft that used to take 45 minutes? 45 seconds. A site visit report? Dictate your notes into your phone, paste them into Claude, and you've got a first draft before you've left the car park. Meeting minutes? Done before the Teams call has even ended.
You're not using AI for engineering. You're using it to stop doing everything that isn't engineering.
The real reason you haven't started.
It's not scepticism. I know it's not, because when I show engineers what AI can actually do with their admin, they don't push back. Their eyes go wide. They ask "why didn't I know about this?"
The real reason is time.
You don't have time to experiment. You've got 6 engineers, a stack of deadlines, and no slack in the system. "Try out AI tools" sits on the same list as "reorganise the server" and "update the fee template." Both have been there since 2019.
You can't adopt new tools when your team is drowning in manual process. The firms adopting AI aren't smarter than you. They already freed up the time to try things. They fixed their operations first, and AI became the obvious next step.
If your engineers are spending 15 hours a week on admin, they don't have 30 minutes to learn something new. The admin has to go first.
Here's what AI actually looks like at a small firm.
Not robots designing buildings. Not generative design creating entire floor plates. Forget all of that for now.
At a 10-person structural engineering practice, AI looks like this:
Your senior engineer is prepping for a QA review. Instead of spending 3 hours cross-checking the calc pack against the basis of design, they feed both documents to AI and get a flagged list of discrepancies in 10 minutes. They still review it. They still make the call. But the grunt work just got cut by 80%.
Your principal needs to write a fee proposal for a residential scheme. Instead of opening last year's template and manually adjusting everything, they describe the project to AI and get a draft in the firm's tone, with appropriate fee stages, in under a minute.
An architect emails asking for a status update on three projects. Instead of your engineer spending 20 minutes checking trackers and composing a reply, AI pulls the latest from your project system and drafts the response. Your engineer reads it, tweaks one sentence, sends.
None of this is science fiction. This is what the 27% are already doing. Every day.
By 2031, 41% of construction workers will have retired.
That's from Deloitte's 2026 outlook. Forty-one percent. And only 10% of the current workforce is under 25.
The industry is losing people faster than it can replace them. That's not going to change. There is no magic pipeline of young engineers about to flood the market.
So the firms that make their existing team 20% or 30% more productive with AI will win more work, respond faster, and retain better people. Because those people will spend their days doing actual engineering instead of copying numbers between spreadsheets.
The firms that don't? They'll keep haemorrhaging talent to competitors who've figured this out. And they'll keep blaming the "skills shortage" for a problem that's really about how they run their practice.
You have two choices.
Choice one: keep doing what you're doing. Lose 15 hours per engineer per week to admin. Watch the 27% pull further ahead. Tell yourself you'll "look into AI" when things calm down. Things won't calm down.
Choice two: fix the operations first. Get proper systems in place. Claw back the wasted hours. Then use that freed-up time to bring AI into your workflow, starting with admin, not engineering.
The second choice is harder to start and easier to sustain.
The first choice is easier to start and impossible to sustain.
Pick one.
The industry already picked for you if you're taking too long to decide.