Augment, Don't Replace: What AEM's John Somers Sees Coming for Construction

Close your eyes and picture an "autonomous job site."

I'll guess what you see. A hundred machines moving dirt. No operators. Nobody in a trailer watching the screens. Just steel doing its thing.

That's not what's happening. And according to John Somers, it's not what's coming anytime soon either.

John is Vice President of the Construction and Utility sector at the Association of Equipment Manufacturers (AEM). He's been at AEM for more than 19 years... seven in marketing, about twelve in the construction sector. He spends his days between OEMs, contractors, policymakers, and the people writing the safety manuals for 50 different machine types. If anyone has a clear view of where autonomy and AI are actually landing in heavy equipment, it's John.

This was his second time on the mic with me, and the first for Hat in the Loop. Here's what stuck.

"Reluctant" is the wrong word

The construction industry has a reputation. Slow. Stubborn. Reluctant to adopt technology.

John hates that word, and after this conversation, so do I.

His point is simple. There's a big difference between pushing back on something and needing to see the ROI land in a timeframe that makes sense. Contractors work on thin margins and fixed timelines, and they carry real risk every time they try something new on a live job. That isn't reluctance. That's discipline.

And right now, the reasons to try something new keep stacking up. Associated Builders and Contractors and others have put the construction labor gap somewhere between 400,000 and 600,000 workers. When you've won the bid, you've got a completion date, and you don't have enough people to run the equipment... that's when the roller gets an autonomy kit.

Which brings us back to that imaginary job site. What's really happening today looks a lot more like this:

→ A tandem roller compacting soil or asphalt on its own, because nobody could be found to sit in the seat going back and forth all day

→ Demolition robots going into spaces no human should be in

→ Multiple people still in control of the process, and most machines still run by an operator

John called it augmented. Supplemented. Humans are still running the show.

We also talked about why construction trails other industries here, and it comes down to something every contractor already knows... no two job sites are the same. A factory is a controlled environment. You can perfect a process in one building and copy it to the next. Amazon does exactly that. Agriculture has a version of the same advantage, because a row-crop farmer works the same plot of land every year.

A job site? Different soil, different rock, different plan, every single time. That's a much harder problem for a machine to learn, and it's why the people who understand how equipment is really operated matter so much right now. John told a story about an autonomy startup proudly showing an OEM their kit digging a hole. The OEM's reaction was basically, "My four-year-old could dig it better." Wrong movements, wrong timing. The machine was picking up dirt, technically. But nobody with real operating experience had taught it how.

That's the gap. And it's why training content, operator knowledge, and simulator data are suddenly some of the most valuable assets in this industry.

Safety isn't on the list. It is the list.

Anyone who knows me knows product safety and technical communication are core values for me, not talking points. So I was glad John went there without me pushing.

His take: we shouldn't be building or harvesting anything unless we're doing it safely. Otherwise, what's the point?

What's changed over his 19 years isn't how much safety matters. It's how much ground it covers. The regulatory space keeps getting more active and more complex, and a lot of it isn't about the instant physical harm we usually picture. Emissions. Chemicals and materials in products. Air quality. All of it traces back to protecting people.

Then there's technology itself. John brought up an example I can't stop noticing now. Delivery vans used to beep when backing up. Everyone got so used to the beep that they tuned it out. So now a lot of them make a different, harsher sound. Solve one problem, and a new one shows up next to it. Every new piece of tech on a job site carries that same question... what happens over here when we change something over there?

But the part that hit me hardest was about people.

John oversees AEM's safety materials program, the manuals for 50 different machines. And he said they're asking whether there should be general guidance for the person who isn't operating anything at all. The one who just laced up boots and strapped on a hard hat for the first time and walked onto a job site.

What do they need to know? And his follow-up: what do they need to know that maybe I don't even know?

I've spent 20+ years in training and adult education, and that's the gap I've seen over and over. We train people on the machine. We rarely train them on the place. As the workforce turns over, with retirements on one end and new hires on the other, that gap only gets wider.

It's also exactly the kind of conversation that happens at AEM's Product Safety and Stewardship Conference, where engineers, product managers, compliance folks, and technical communicators get on the same page and learn from each other. John called that shared learning one of the most important things AEM offers. I agree. The 2027 conference runs April 19–22 at the Radisson Blu, Mall of America, and if safety is part of your job, it should be on your calendar.

The ATM paradox

This is the part of the conversation I'm still chewing on.

When ATMs started showing up in banks, the fear was predictable. Tellers were done. People could get their own cash. Why would a bank need all those employees?

That's not what happened. Banks got more efficient. They opened more branches. They hired more people. And the people working there stopped spending their whole day counting fives and tens. They became relationship managers. They handled loans and accounts and spent real time with customers.

There are seven banks within walking distance of my house. Seven.

John's point is that the same pattern is coming to manufacturing and construction. Some jobs will be displaced. Some will change completely. Pretending otherwise doesn't help anyone. But if your thinking is "this is what I do today, so I'm finished tomorrow," you're looking at it the wrong way. The job evolves. You can still be working toward the same outcome, an excavator rolling out the door at the end of the day, just from a different seat.

You can already see it:

→ There aren't enough welders, so robots handle more of the welding. Someone still has to oversee those robots and check the welds.

→ Factories are being redesigned around automation, and the ones I saw at RES and Fabtech this year actually gave humans more space to move, with sensor zones and instant stops when a person gets close.

→ Upskilled workers end up monitoring machines instead of sitting in a single cab all shift.

I'll be honest about my own version of this. What I do for a living probably won't look the same in five or six years. It also didn't look like this five years ago. I've been using agentic workflows heavily for the last couple of years, and if you took them away from me today, I genuinely wouldn't know how to do my job the old way anymore. Work that used to take me six months now takes three or four days.

That's not a threat. That's the ATM paradox playing out on one person's desk.

My dad would have told me to stay far away from manufacturing. He saw the bad side of it. If my kids want to go into manufacturing, I'll back them 100%. It's safer, cleaner, more efficient, and more high-tech than most people realize. If you want to work in tech, manufacturing is a great place to be.

Why AEM Annual feels different this year

AEM Annual is my favorite conference every year, and John explained why this one is shaping up as the most future-looking yet.

It isn't a product showcase. Members aren't on stage pitching what they sell. It's enterprise strategy and tactics, built for the people deciding where their companies go next. Two sessions stood out:

→ A member panel on real AI adoption. Equipment manufacturers sharing how they've actually implemented AI inside their teams and what's worked. On the prep call, one panelist said something that I think is the whole message right now... by now everyone should understand AI is a tool. The next step is going back and rethinking how you do your job so it's fully integrated. Shout-out to Chris Morbitzer, who's running that panel.

→ Where the money is going. JP Morgan's head of global investment opportunities will talk about how investor priorities, funding sources, and financial structures are shaping innovation. More construction organizations are starting venture arms or backing startups directly, because the tools they need aren't arriving fast enough. If you want to know what's coming, follow the money.

The takeaway

When I asked John for his shout-out, he didn't pick one person. He pointed to the contractor groups, like Matt Abels and the technology and innovation work at ABC, and AEMP's push on cameras, proximity sensors, and other safety technology. These groups are bringing their members' real needs straight to the OEMs and the startups building the next generation of equipment.

That's the pattern across this whole conversation. Autonomy and AI aren't replacing the people who know this industry. They depend on them. The operator who knows what a good dig looks like. The safety lead who knows where the guard belongs. The trainer who knows what a first-day worker needs to hear.

This was all complicated before any of the new tech showed up. Most of the answers come from making sure the people with that knowledge are in the loop.

The full conversation with John Somers is up now on Hat in the Loop. Give it a listen, and then tell me... what's the biggest misconception you hear about autonomy in your corner of the industry?