Currently, most people use AI to chat. I use it to ship industrial automation code, make market analysis, lead preparation, write our partner handbook or make presentations.
I recently started working with Claude Code. And suddenly, “Agentic AI” isn’t just a buzzword anymore. It feels tangible.
It’s one of those moments that feels as transformative as when ChatGPT first launched.
What I’ve Been Using It For#
Here are some of the things I’ve been doing with Claude Code:
- Improving my Hugo-based website with new content, updates
- Building an app for the ctrlX OS Ecosystem
- Python scripts for home automation projects
- Preparation for fair visits
- Market analysis including competitor monitoring
- Company information gathering for partnership inquiries
- Presentation preparation for a product manager community meeting
I asked Claude Code if we should participate in a new fair for us and if this justifies the investments. It made a clear argumentation ready for management why we should attend this fair, made a suggestion for possible leads and made a plan to prepare for the fair.
We also made a market analysis. A comparable report would easily have cost several thousand euros just two years ago.
Building a ctrlX OS App#
Could an AI coding agent help me build a real app for an industrial control system?
It can. 🏭 👷
I built a Python snap that reads live ctrlX OS system metrics and a PLC variable from the ctrlX Data Layer and displays them in a web dashboard.
Nothing fancy on the surface. But under the hood, it’s already tapping into core elements of ctrlX OS, like the ctrlX Data Layer. This is the data backbone of our open automation platform, uniting all controller data in one semantic layer.
Was it smooth sailing? Not exactly.
The actual app was made in 5 minutes, but the integration into ctrlX OS took a bit longer.
Library dependencies and the early adopter version of ctrlX OS 4.4 and the ctrlX OS App Building Environment (virtual machine with Linux Ubuntu Core) kept us debugging for 2 hours.
Claude Code asked me to deliver different system logs to it. It wanted to see how apps on ctrlX OS are integrated into the system and learn from this for my app. We went back and forth for a while.
But that’s part of the learning.
My takeaway: Claude Code lowers the barrier drastically to build even for industrial control systems! I have never made an app for ctrlX OS and now I’m empowered to do so!
Is the code perfect? Definitely not, see my GitHub repo.
Was it fun? Absolutely!
My colleagues also made this experience using GitHub Copilot and posted a How-to in our ctrlX OS Community.
Why This Matters#
It’s really fascinating to finally understand the buzzword “Agentic AI” in practice. Instead of just chatting with an AI, you’re working alongside an agent that can actually execute tasks, navigate your file system, run commands, and build things with you.
The shift from “AI that talks” to “AI that does” is significant – and tools like Claude Code make that shift real.
It also has the potential to change the Industrial Automation world. I have created an app for a controls system, Claude Code also proposed code examples in IEC 61131-3 Structured Text.
Will we have the code conversion tool in 6 months that could change everything? Many machine builders and operators are stuck in legacy PLC systems due to the investments in code that they have made over many years.
Could AI coding agents soon leverage this treasure and make the code transferable into newer, better and more secure automation platforms? These code investments block many users from moving currently.
What I’m Curious About#
I’m planning to test different coding agents, not only Claude Code. I will work with Codex from OpenAI and try also OpenCode with local models.
👉 What are your experiences with AI agents? Comment on my LinkedIn post
