When AI Writes Code, But Engineers Still Call the Shots

Not long ago, Mark Reitzel asked an AI to build him a web version of a Japanese card game his father used to own. Ten minutes later — poof — a working application appeared on his screen. Functional, polished, and ready to go.

Impressive? Absolutely. But for Reitzel, who directs product management at Mitsubishi Electric Iconics Digital Solutions, the real question wasn't about card games. It was about industrial automation. If AI can whip up a game in ten minutes, what happens when you point it at a factory floor?

The answer, as it turns out, is complicated.

Here's the thing: AI is fantastic at generating code. Dashboards, reports, visualizations — these come together in minutes flat. But industrial software isn't a dashboard. It's the nervous system of a manufacturing plant, the thing that keeps assembly lines humming, turbines spinning, and safety systems doing their job.

Generating the first version of an application? That's getting easier by the day. Delivering software that plant managers can actually trust to run their operations 24/7? That's a whole different ballgame.

Take a simple example. A plant manager asks for an operational report. AI generates it in minutes. Everyone's happy. Then the manager says, "Great — now send me that report every single day." Suddenly, a one-off task becomes mission-critical software that people depend on. Platforms evolve, security threats emerge, requirements shift. And someone has to keep all the plates spinning.

That someone is an engineer — not just any engineer, but one who understands industrial architectures, cybersecurity, regulatory compliance, and the messy reality of keeping legacy systems talking to new ones. AI doesn't know if your data flow exposes a vulnerability. AI doesn't worry about whether that API call will still work after next quarter's platform update. AI doesn't lose sleep over what happens when the network goes down at 3 a.m.

Experienced engineers rarely build industrial software from scratch anyway. They build on proven platforms — the kind that have been battle-tested through years of real-world deployment. These platforms come with validated architectures, redundancy, cybersecurity protections, and long-term support. AI can generate an application on top of that foundation in record time. But that application is just one layer of the solution.

Speed is seductive. But in the industrial world, speed alone doesn't determine whether software is ready for prime time. The real question isn't "How fast can we build this?" It's "Who's going to keep it running five years from now?"

AI writes the code. Engineers own the outcome. And in industrial automation, that distinction makes all the difference.