The Biggest Lie in AEC Right Now

Anthropic's AI report just made the case for a career in construction.

👋 Hey, Kyle here! Welcome to The Influential Project Manager, a weekly newsletter covering the essentials of successful project leadership.

Today’s Overview:

• Anthropic's new labor report shows who's actually getting disrupted by AI — and it's not construction.

• Why high-paid desk workers face more AI exposure than the trades (and what the data says)

• How to turn construction's "low exposure" into an unfair competitive advantage.

---

The decisions stay human. AI handles the rest.

The PE stamp has to be a human. The owner's sign-off has to be a human. The super's call on whether a lift is safe today, the architect of record's certification, the GC's signature on the pay app — these don't go away. They can't. Liability, code, and trust all sit with a named person.

But every one of those decisions sits inside a web of coordination — owner, GC, subs, architects, engineers, inspectors — all moving in parallel, all waiting on each other. Pulling the submittal package, validating completeness, routing for review, chasing late responses, reconciling versions, building the audit log. None of it requires a person chasing it, and most of it eats the week.

Moxo orchestrates that work across AI agents, the systems you already run, and the humans who need to take action. Agents prep, validate, route, and chase. The PM, the architect, the owner's rep arrive at every decision faster, with full context, and ready to review.

The accountable human stays in the chain. The grunt work doesn't.

See how it works at Moxo.com.

---

🤥 The Biggest Lie in AEC Right Now

Filed under: Construction, AI

Hey there,

It seems every few weeks someone is telling me:

“AI is replacing everyone…. SaaS is dead…. Agents are taking over…. Every white-collar job is on borrowed time.”

College students are changing their majors due to AI fears about job displacement (Axios).

Your LinkedIn feed is a wall of panic. Your group chat is asking: "Are we next?"

If you're a PM, superintendent, or project engineer, you've had the thought: “What does this mean for me?”

I get it. So much noise. For the most part, I've ignored it, because most of it is speculation dressed up as insight.

Then a few weeks ago, something different dropped.

Anthropic, the company behind Claude, one of the most widely used AI systems on the planet, published a major labor market research report. No predictions. No hype. It includes actual observed data on what AI is doing to real jobs, in real workflows, right now.

For anyone working in AEC, the findings are worth slowing down for.

The jobs getting hit hardest? They're not ours.

Here’s what we’re covering:

1. What the Anthropic report actually measures

2. Who’s exposed right now (the data will surprise you)

3. Why construction stays low — and why that matters

4. The real opportunity hiding in plain sight

Let’s go.

---

1. What the report measures

Anthropic uses observed exposure to estimate how much of a role’s work is being touched by AI in practice, with extra weight given to tasks that are fully automated (not just assisted).

Construction comes in around 15% theoretical exposure and about 2% observed coverage. That puts us near the bottom of the chart:

Figure 2: Theoretical capability and observed exposure by occupational category. Share of job tasks that LLMs could theoretically perform (blue area) and our own job coverage measure derived from usage data (red area).

The red area, depicting LLM use from the Anthropic Economic Index, shows how people are using Claude in professional settings. The coverage shows AI is far from reaching its theoretical capabilities. For instance, Claude currently covers just 33% of all tasks in the Computer & Math category.

As capabilities advance, adoption spreads, and deployment deepens, the red area will grow to cover the blue. There is a large uncovered area too; many tasks, of course, remain beyond AI’s reach—from physical agricultural work like pruning trees and operating farm machinery to legal tasks like representing clients in court.

---

2. Who’s exposed right now

The report shows high coverage in roles like computer programming (around 75% observed coverage), plus customer service and a lot of analyst work.

Figure 3: Most exposed occupations. Top ten most exposed occupations using our task coverage measure.

The part that caught my attention is who those workers are:

• Pay: the most exposed group averages $32.69/hr vs. $22.23/hr for the least exposed (about 47% higher)

• Education: graduate degree holders are 17.4% of the most exposed group vs. 4.5% of the least exposed

• Work setting: more office-based, more screen time, more repeatable knowledge work

You can see it in early labor signals too. Entry into high-exposure occupations for workers ages 22–25 is down ~14% since ChatGPT launched. BLS projections in the report also show a small but consistent relationship between higher exposure and lower projected employment growth through 2034.

The most vulnerable workers in America right now earn $32/hr, hold graduate degrees, and sit at desks.

Most people had that backwards.

---

3. Why construction stays low

That 2% number reflects something specific about the work.

AI processes patterns in structured data and produces outputs at scale. That's why it's moving fast through software engineering, customer service, and financial analysis. Those jobs run on information that can be digitized, repeated, and automated.

The inputs on a job site work differently.

Every project is a one-time prototype built by a different team, in a different location, under different conditions. The data is messy. Every project starts from scratch.

Think about what a great superintendent actually does:

• Reads a room of five trade foremen with competing priorities and gets them aligned before anyone walks off the job

• Spots the HVAC rough-in conflict before the plumber cuts the wrong pipe

• Makes a call to keep work moving with 60% of the information they wish they had

• Rebuilds trust with a client after a bad week, without a script

All of it runs on high-judgment, sequencing, relationships, and physical presence.

AI can absolutely help with parts of the work, and it already is. But the core outcomes still depend on people who can think clearly under pressure, in messy conditions, with real stakes.

Construction sits at 2% because the work is hard to replicate.

The safest place to be in the job market right now? Construction.

---

The real opportunity

Here's where I want to push your thinking:

Safety from AI disruption is a baseline. The actual opportunity sits higher.

The opportunity is building an AI-native team inside the most human-intensive, judgment-driven industry on the planet.

Automate the tedious. Amplify the high-judgment work. Get better outcomes than either humans or AI could produce working separately.

Here's what that looks like in practice. I've been building out a system in Notion with dedicated AI agents for each major administrative workflow:

1. Change order management — An agent flags missing backup, checks scope alignment, and drafts the initial response. I step in for final checks and negotiation.

2. Scope writing — An agent generates the first draft from project notes and prior work. I refine and sign off.

3. Meeting notes and action items — Captured and organized automatically so the team always knows who owns what.

4. Project reporting — Owner reports drafted from live project data so my PM walks in prepared, not still writing.

5. RFI and submittal tracking — Clean and current so the PE spends time coordinating, not chasing paper.

6. Supply Chain Management - Agents build submittal, Connects activities, submittals, and materials into one system.

7. Prioritization and communication — Open items ranked by urgency and impact. Stakeholder updates, visual summaries, and project snapshots drafted and ready to send.

If you pair that leverage with field judgment, you get more predictable outcomes with the same headcount.

AEC sits in the best market position of any professional category right now!

P.S. We’re hiring. If you’re a PM, PX, or Superintendent that want to build on a team that runs projects this way, contact me now!

---

Here’s what to take with you:

1. High-paid, desk-based, repeatable knowledge work carries the most risk. The data says so, and the early labor signals confirm it.

2. Construction’s core skills resist automation. Sequencing, relationships, decision-making under pressure — those run on judgment, not data.

3. Safety from disruption is a starting point. The teams that win build AI into their workflow to amplify what their people already do well.

4. Start with the admin drag. RFIs, meeting notes, scope writing, and change management — these are the places to deploy AI first.

Until next week,

Kyle

---

Find me on LinkedIn, Threads. or Book a 1:1 Call.

Ready to level up? Here are 4 ways I can help you:

1. Get ProjectOS. The AI-powered Notion project operating system I use to manage $300M+ in projects. Set up in 1 day, not 3 years. Save 10+ hours weekly.

2. Join Antifragile Project Management. Leadership development course and community built on the 7 Archetypes Framework. Transform from project witness to project leader. Earn promotions and reach senior leadership levels.

3. Read the Book. No Bullsh*t Project: A Project Manager’s guide to successful project leadership. Available on Amazon.

4. Promote Your Business. Get your brand in front of 8,600+ highly engaged construction leaders and decision-makers with a 51% open rate.