Caterpillar figured out something most companies deploying AI are still struggling with: the model is the easy part. Getting it into daily operations is where the work actually lives. According to TechCrunch AI, the industrial giant is now taking decades of hard-won lessons from automating mining sites and applying them to its broader AI push, from field technician tools to legacy code modernization.
That’s a useful signal. When a 100-year-old equipment maker with 118,000 employees says the hard part isn’t the tech, it’s the workflow, everyone deploying AI should pay attention.
From haul trucks to jobsites
Caterpillar’s autonomy story started in mining, where labor shortages and dangerous conditions made automation an obvious win. The company now sells automated haul trucks, drills, underground loaders, dozers, and remote-controlled gear, plus a software command center and fleet management to run it all.
Now it’s moving that playbook into messier environments. “Now we’re in this super exciting time where we can take all of that learning from mining and bring it into much more dynamic environments, jobsites, quarries, and construction sites,” CTO Jaime Mineart told TechCrunch AI during a fireside chat at the Ai4 conference in Las Vegas.
What stands out here is the sequencing. Caterpillar didn’t start with the hardest, most unpredictable sites. It started where automation was easiest to justify, proved it out, then expanded. That’s the opposite of how a lot of companies approach AI, chasing the flashiest use case first and stalling on integration.
The data moat nobody talks about
One example Mineart highlighted is the Cat AI Assistant. A field technician standing next to a machine can use voice commands to pull up repair procedures, troubleshoot problems, and identify parts before starting a repair. Customers, operators, and technicians are already using it.
The assistant runs on Caterpillar’s proprietary data. And the numbers explain why that matters: about 1.6 million connected assets globally and more than 16 petabytes of structured data. That’s the real moat. Any competitor can license a frontier model. Almost none can replicate 16 petabytes of machine data collected over decades.
Caterpillar is also using AI to scan sites and build digital twins in manufacturing, and internally to modernize legacy code, generate and test software, and catch defects earlier. Standard enterprise stuff by now, but worth noting the company treats its own operations as a testing ground.
Why the workflow is the real bottleneck
Mineart kept coming back to one point. “The hard part about autonomy and about physical AI is incorporating that technology into the customer jobsite and into the workflows,” she said.
Deploying an autonomous machine isn’t the same as transforming a site. People have to change how they work. Roles shift. An operator who used to control one machine might end up overseeing several from a remote command center. Caterpillar leans on experienced operators to help train the AI, turning institutional knowledge into training data.
That shift creates a new problem: retraining the workforce. The company plans to spend $100 million over five years training its 118,000 employees in AI, autonomy, and robotics. Read that as a signal of where the constraint really is. It’s people, not algorithms.
What practitioners should take from this
A few practical lessons from Caterpillar’s approach:
- Start where the ROI is obvious. Prove the system in a controlled setting before pushing into chaotic ones.
- Guard your proprietary data. Your operational data is the durable advantage, not the model.
- Budget for change management. If you’re spending on models but not on retraining people, you’ve mismatched the problem.
- Turn expert knowledge into training input. Your veterans are a data source, not just a cost center.
There’s a business tailwind underneath all of this. Caterpillar’s Q2 revenue hit an all-time high of $20.5 billion, helped by demand for power-generation equipment used in data centers. That division’s sales jumped 72% to $3.10 billion, and CEO Joe Creed said “no one is slowing down” on cloud and generative AI infrastructure.
So Caterpillar is playing both sides of the AI boom. It’s selling the picks and shovels that power data centers, and it’s using AI to run its own business better. The company that spent decades automating the physical world may have quietly built one of the more grounded AI strategies out there. Full details are in the original TechCrunch AI report.