OpenAI's Enterprise Push Is Impressive, But Let's Talk About What It Actually Means for the Factory Floor

I called my old colleague Frank at Siemens last week to get his take. He's still in the trenches, working on PLC integration for a tier-one automotive supplier. His view? These large language models are useful for documentation, for maintenance scheduling, maybe for some quality inspection tasks where you've got time to think. But real-time control? We're not there. We might never be there with this architecture.

Where the Enterprise AI Actually Lands

The Uber partnership is instructive. OpenAI's working with them on driver earnings optimisation and booking assistance. That's a marketplace problem, not a physical automation problem. The latency requirements are completely different. If your ride-hailing app takes an extra 500 milliseconds to suggest a route, nobody dies. Nobody even notices.

Same with Parloa's customer service agents. They're building voice AI that can handle enterprise call centres at scale. Impressive, genuinely useful, probably going to put a lot of offshore call centre workers out of jobs. But it's not robotics. It's not even close to robotics.

The India expansion is interesting, actually. OpenAI's building local infrastructure there, which suggests they understand that latency matters enough to warrant regional data centres. That's a start. But a data centre in Mumbai doesn't help a warehouse robot in Chennai if the control architecture still assumes cloud connectivity.

What This Means for Industrial Automation

I'm not saying OpenAI's enterprise push is irrelevant to our industry. It's not. The applications they're listing (predictive maintenance, quality documentation, supply chain optimisation) are all real use cases. I've seen demos. Some of them are genuinely better than what we had five years ago.

But the coverage treats this like AI is about to take over manufacturing. It isn't. The hard problems in industrial robotics aren't language problems. They're physics problems, reliability problems, integration problems. When I was commissioning lines in the early 2000s, the challenge wasn't making the robot smarter. It was making sure the gripper didn't drop parts, the vision system could handle lighting variations, and the whole thing would run for 18 months without a major fault.

Those problems haven't gone away. They've gotten harder, if anything, because everyone wants more flexibility, more product variants, faster changeovers.

So yes, celebrate the million customers. It's a real achievement. But let's not pretend this is the moment AI conquers the shop floor. We're still waiting for that, and I suspect we'll be waiting a while longer.

Sarah Williams · 5 hours ago · 2 min