Anyone have personal experience using these “AI Operator” type things? I know of Cube and Facil.ai but I have yet to see them action. Not “in action” as in the theory of their use, but the actual set up process.
My assumption is that they work like other front-ends that collate data from multiple BAS vendors. Get a network connection to your supervisors, pull in your BACnet points, and the AI can use the data.
I’m not convinced of their usefullness for smaller shops. Maybe if I’m a JCI rep for a tri-state area, but if my shop has less than maybe 50 active sites at one time, and not all of them want remote access, could I reasonably fit an A.I. solution?
So I have a few NDA’s in place so I cannot talk about under the hood on some of these but I can talk about AI in general for solutions:
Some of then talk to the data directly through drivers they wrote specifically.
Facil.ai has a few large high-profile clients that they monitor and control their systems and those companies are happy with their energy savings. Your multi-site scenario would be a good candidate for something like that. I can get you in contact with people at facil.ai and have them speak to you directly. Though I don’t know if they talk to controls companies rather the customers of the controls companies.
There’s a big push for AI to help control and facilitate energy savings and I’ve seen it do fairly well in some scenarios.
AS AN ASIDE:
I have a lot of concerns about AI itself. While I think it’s a great tool, I truly feel you should know how stuff works before getting AI involved. I know it can be a good learning tool as well, but I think if you start with AI and don’t truly learn things on your own in applications, you’re going to lose sight of what’s in front. I use AI really for one aspect for me, and that’s color schemes and true styling. I’m 100% color blind and so my eye to make things pop doesn’t really exist. Therefore, I use it to make those calculations for me. Otherwise, the rest of the coding I do comes from my mind. I know people who get their code working and then use AI to pump out other code based off that, which I also think is fine. Especially if you know how to fix it afterwards.
I feel people will start relying on AI so much that it’s going to cause us to stop truly thinking for ourselves and it’s going to be a crutch rather than a power tool.
Some of then talk to the data directly through drivers they wrote specifically.
Do you mean they have drivers to talk through the communication trunk? Or that they have drivers to specifically interpret data from the supervisors, but it’s still an IP connection. I know you can’t go into specifics, but it makes me wonder lol
I feel people will start relying on AI so much that it’s going to cause us to stop truly thinking for ourselves and it’s going to be a crutch rather than a power tool.
People have a false assumption that AI will hijack the thought process of the future workforce and prevent them from being valuable in the same way previous generations were. I’ll tell you with absolute certainty, the upcoming workforce in the trades will need AI to figure out how to change a local IP address on Windows. It’s that dire out there.
But hey, maybe if we made all the programs and workflows like a phone app, they’ll understand that, right? Hell no. Ask a 16 year old how to turn off adaptive brightness on their phone. You’d be surprised how long it would take them to figure it out. The level of tech illiteracy has only become exponential, and as much as I hate it, LLMs trained on controls might be the only way to trust your average zoomer to start screwing with a site’s configuration.
To answer your first question. They have drivers talk to the communication trunk. No need for a Niagara Supervisor at all. A few others will use the supervisor in place, but others tie directly into the communication trunk.
For your second statement, I don’t think AI will replace most jobs. What I would like to see if people to be self-sufficient to start or at least have enough knowledge to guide the AI somewhere. Now, certain things that I think that could be out of their depth? Sure, AI can give them an answer.
What I’ve been dealing with recently are people who use LLMs willy-nilly and then ask me to fix their code because AI didn’t have all the information that it could’ve been given. So at that point, I wish people would try stuff on their own first (and there are a lot of free utilities to assist with the learning process, should they want). I’m not saying they can’t use Google or anything like that to find answers, but I’d want them to have some kind of knowledge to use things as intended.
I agree that AI won’t “replace” jobs, or at least jobs that are skilled enough to make a noticeable impact. Malaysians getting work through Fiverr might not be as skilled in Photoshop anymore, but they’ll start running Stable-Diffusion AI farms instead. It’ll be fine in that regard.
My point is more that the younger generation has grown up with technology that is so incredibly low skill to use, that when it comes time to funnel them into the workforce, perhaps LLMs will need to fill in for that gap.
Not but 15 years ago, if you were a 12 year old kid and wanted to mod a computer game, you had to probably know how to read JSON a little bit and understand file structures. And know how to effectively search forums for the information you’re lacking. Nowadays, everything is a singular launcher program and if your mod file doesn’t work, you’re shit out of luck because you’re not supposed to access files or change any configurations.
BAS is more complex than a phone UI, but our coming workforce is unfamiliar with computer concepts.
You’re absolutely right! Great points! 
