PLC Programming in the Age of AI (Part 2) The People Who Use AI Best Are Still the Same High Performers

Posted by on



After publishing the previous article, several friends told me it was “a bit too painfully accurate.” So today, I’ll continue along the same line.

By now, most of us have probably noticed that many people around us are already using AI to assist with PLC projects. Some highly capable engineers have gone even further—they are using AI in mechanical design, UI design, PLC architecture, and code generation. Others with less experience may still be experimenting, but even they can use AI to generate function blocks.

Under different goals and expectations, many people can confidently rate the results of their AI-assisted work at around “90 points.”

But have you noticed an interesting pattern?

No matter which AI model—or combination of models—is used, the people who achieve the most impactful results tend to be the same group who were already strong engineers before AI. They were either the ones who used to write solid code manually, or they were successful team leaders.

This phenomenon is worth thinking about.

To some extent, it suggests that the effectiveness of AI is not determined solely by the capability of the model itself. Instead, it depends heavily on the abilities the user already possesses.

So what exactly is this ability?

Is it an ability shaped by the working environment, or is it primarily a personal capability?

If we start attributing it to organizational constraints, management structure, or process design, the discussion quickly becomes too broad. But we can approach it from a simpler perspective.

Consider the automation project currently on your desk. Whether you are implementing it yourself or assigning it to others, ask a few practical questions:

  • How do you define work objectives?

  • How do you design the workflow?

  • How do you track project milestones?

  • How do you obtain timely and accurate feedback on project status?

Or to put it more bluntly:

How do you avoid ending up with a pile of “spaghetti code”?

(And if this question has never crossed your mind before, then perhaps we are not exactly in the same profession.)

Writing a single function block with AI is probably one of the easiest tasks today. Both experts and beginners can often achieve reasonable results. The reason is simple: writing a function block is inherently a low-management problem.

  • The system coupling is minimal.

  • The functional boundaries are narrow.

  • The scope is local rather than systemic.

However, writing an entire PLC program is far more complex. That becomes a system-level problem. And when you move from writing a PLC program to developing an entire product, the problem expands further into a multi-domain, global system problem.

Instead of trying to define this “core ability” precisely, it may be enough to recognize that modern AI models can already help individuals improve their understanding of it. At the very least, they can enhance our awareness of architecture, system thinking, and project organization.



For example, it is now possible to have meaningful discussions with AI about the software architecture of automation systems across different platforms and industries. One can even discuss product-level architecture and management strategies. I won’t include screenshots here, but the capability is already quite impressive.

For now, the best mindset may simply be:

“Finish first, then perfect.”

AI tools are important not just because they increase productivity. More importantly, they create a rare moment of capability equalization.

In the past, it was almost impossible for an individual engineer to have access to such a highly capable assistant—one that can follow instructions, iterate on ideas, and participate in problem-solving at this level.