I've been working with the platform for a decade now, and I'm a bit of an AI holdout. For a variety of reasons, ranging from retaining expertise to destroying the planet, I've mostly resisted the call to use AI for anything and everything - call me a Luddite. My first and last prompt was written in 2023.
But I also realize that AI is here to stay in some capacity, so I am curious to hear your Ignition AI Stories. If you've got a favorite success or failure, cautionary tale, what have you, I'd love to read it.
What are your favorite use cases? Things you've produced that may not have been possible otherwise? Time savings?
There is lots of talk about having AI develop whole views or handling debugging, both things I see as a means of closing the loop when developing / maintaining expertise. Have any of you run into a scenario where you churned something out with AI that then became a headache to support?
Have you found a happy medium where you're learning through a partnership with the AI model?
In saying that, i do use it to help make basic SQL queries, or for long tedious but simple tasks. Things that are easy to verify, and also low risk if it goes wrong. Like today i had it make a dataset memory tag that stores fault codes, their descriptions and their remedies as specified in the parts manual. Would have easily taken me a couple hours to type it all out, but instead i did the first few rows to set the pattern, then gave it to the AI to finish out.
I'm currently using Claude Code to accomplish something I've been wanting to for years, but never tackled because it is kind of a major undertaking...
Our MES has always been divided into two main projects, one for Production, and one for Secondary, but over time there has been a lot of overlap where both projects have their own version of essentially the same interface, and so their common functionality has been inside a third project that they both inherit from. I am constantly having to juggle my development across all three.
Many users regularly use both projects, and so shortly after I started working here I created a "wrapper" project that uses a side menu to navigate between the two, loading them inside iFrames. This has been clunky, and although I did manage a hacky workaround that automatically logs users into one project when they've already logged into the other, it is still kind of a pain to keep switching between the two.
I'm nearly finished having Claude consolidate it all into just one project, something I only started yesterday, and which otherwise would have taken me several days, if not weeks to pull off. Claude is quite knowledgeable about Perspective, and is able to build me patch ZIPs that I then import into my Designer to test.
I also recently used CC to build an Organizational Chart for Perspective that is on the Exchange. It uses Markdown JavaScript injection, which is an endangered species, and so I plan on getting help from Claude to convert it into a proper Module.
As the number of tools required in my toolbelt seems to be every growing AIs best use (for me) right now is as a code assistant and better contextualized search engine. Ive also used it to make mock ups for interfaces and generate simulated datasets to make demos out of.
I'm not a heavy AI user but I'm not against it. I mostly use it to look up obscure errors and to troubleshoot syntax errors in more complicated SVGs. I don't think it's going anywhere and it will only get better as people feed human-made tools into models to expand functionality.
I'm not convinced the "destroying the planet" narrative is based on reality. My wife's an environmental geologist. She's actively working on at least 3 data centers I know about. We don't discuss NDA stuff with each other so I am speculating that her involvement in the projects indicates that they are trying to minimize environmental impacts.
I'm also not convinced that AI is the silver bullet that will solve all problems and eliminate the need for people to do real work anymore. At least not anytime soon. We'll see where it goes. It's certainly getting better.
Here are some Ignition-related use cases I can point out:
The help site now has an AI feature that can be used to find answers (lower right corner).
Throwing error messages into AI can be helpful with troubleshooting.
AI can help with tag migrations from legacy platforms into Ignition. We may not have to write bespoke tag migration tools anymore. It's especially useful if you have tedious data entry requirements.
AI can be used when migrating custom recipe systems from legacy platforms to Ignition. One of my colleagues was able to read a binary recipe file from a legacy platform and port all the recipes into Ignition.
AI is getting to the point where it can make SCADAs that don't have P&ID screens pretty well. Most of my customers don't want those but I think it will prove to be really relevant on greenfield projects for systems that have large volumes of the same type of equipment (data centers, etc). It's very hard to convince people who have P&ID screens for their entire process that they don't actually want them (they usually do). I have seen some AI-made P&ID screens too but I wasn't impressed with what I've seen so far.
I think this was the use case I had for it in 2023 when I was interacting with it, and the results left me wanting. But its been a busy 3 years for AI models.
Wow! That sounds like a huge lift, super satisfying to knock out, and something I would be cautious about being able to support. Are you feeling comfortable with supporting what Claude is churning out?
Those all seem grounded and attainable. I've not done a migration in a long time, but I would have loved to have had the help decoding the madness that is an early 00s scada system.
I've seen the AI button on the docs website, but having been interfacing with it for so long, I've yet to touch it. The docs just feel intuitive and familiar. If it were a paper document, it would be worn, torn, and well loved.
Yes, but I intend to keep testing it the rest of the week, and only deploy first thing Monday when I'll have plenty of time to be around and address any issues I/we may have overlooked. At the first sign of any serious trouble, I can easily revert the changes. It's mostly just reorganizing and rewiring the connections between views, scripts and named queries, with the only visible changes being in menu navigation.
I do trust Claude, but also do my best to verify. Over this past year, he's written over 60,000 lines for my own personal projects, and I haven't even looked at 99% of it.
Had lots of success with it. Especially been handy for creating dev / QC tools. For example I had it create a tagQC library where you can point it at a folder and it scans every tag and provides a report of how many tags are assigned to each taggroup, what OPC UA server and device they poll, what history provider, history sample mode ,etc. It also has fields so you can provide what you expect and it'll report only on exception and provide list of tagpaths of all unexpected tags so you can track them down. Been very handy when mirroring one equipment package to another and QC checking someone's work in seconds.
