Big iron is getting smarter. Your ops should too.
Gravis just raised $200M to retrofit excavators with self-driving AI. Here's what that actually signals for owner-operators.
The headlines this week were full of robots and billions of dollars. Most of it isn't coming to your jobsite anytime soon. But one number buried in the research stopped me cold: 41% of the U.S. construction workforce is expected to retire by 2031. That's not a tech story. That's a succession and capacity story, and the clock is already running.
What happened
Gravis Robotics closed a $200 million round from SoftBank to put self-driving AI systems on existing excavators - not new machines, retrofitted ones. The idea is that you don't need to buy new iron; you bolt autonomy onto what you already have. Separately, Caterpillar announced a pilot with Nvidia to build an AI assistive system into mid-size mini-excavators. And a McKinsey report from July made the rounds again this week, estimating that AI could automate 39% of nonphysical work in construction - estimating, scheduling, document review, coordination - for firms that actually control their own project data.
The Wix and Bluehost AI agent announcements also made noise this week. They're aimed at website businesses and retail. I'll name them here and set them aside - they're not relevant to how you run a construction company.
Why it matters for your shop
The autonomous excavator story is real, but it's a 5-to-10-year story for most owner-operators in the $5M - $90M range. You're not going to retrofit your Cat 320 this fall. That's not a knock - it's just honest about where the technology sits and what it costs to integrate. The Gravis raise is a signal, not a product you can buy.
What is relevant right now is the McKinsey framing, and I want to be careful not to overstate it. Their finding that 39% of nonphysical construction work could be automated is meaningful - but only for firms that have their project data in order. If your estimates live in one place, your schedules in another, your RFIs in a third, and your field notes in someone's text messages, no AI tool bridges that. The firms that will see real gains from AI in the next two years are the ones that already have disciplined handoffs and consolidated information.
The 2026 Civil Contractor Report surveyed 150 earthmoving firms and found 35% are running software that won't share data with other software, and nearly a third still make critical decisions by phone call. If that sounds familiar, the AI conversation is premature. The prerequisite is connected data, not smarter tools on top of broken ones.
Here's what I actually think the labor retirement wave means for operators at your scale: the experienced people who carry institutional knowledge in their heads are leaving the industry faster than you can hire replacements. The answer isn't autonomous excavators. It's documenting what your best people know - how a job gets handed off from estimating to field ops, what the project manager checks before a sub walks on site, how change orders actually flow - and building that into your systems before those people are gone.
One thing to try this week
Pick one handoff in your business - estimating to project management is usually the leakiest - and spend 30 minutes with your next meeting using Granola to capture the conversation automatically. Then take the transcript into Claude and ask it to identify every assumption or decision that was made verbally but never written down. You'll probably find three to five things that exist only in someone's head. That's your first documentation priority. Not because AI told you to. Because the person who knows it might not be around in three years.
- Austin
We help growing businesses build systems that last.
If this resonates with your current situation, a 30-minute conversation is the fastest way to know if we can help.