Quality managing editor Michelle Banger and Ed Goffin of Pleora Technologies discuss his recent article for Quality called “Cutting Through the Complexity of AI.” 

Michelle: So can you tell us a little bit about this article and how you came up with the topic?

Ed: I think from covering connectivity solutions to some of the work that we're doing in AI. And in this article, I was really trying to focus more on the end user experience. And there's a lot of hype around AI and a lot of excitement around AI. And sometimes that means people get a little bit overwhelmed, whether that's the terminology or more deployment strategies. So in our case, we're working with a number of small, medium, and larger size manufacturers, all trying to come up with their own deployment strategy for how can I use AI in my manual processes or some of my automated processes. Yeah, that sounds good. I thought the very helpful stuff was kind of talking about what you've seen work and maybe what doesn't work. Could you give us an example of maybe best-case scenario for adopting AI? Sure. So I think part of what happens is a lot of people get excited by AI and they see a lot of perfect use cases for how technology can be deployed in the field and sort of build expectations around that perfect use case. And no matter how much we can try to simplify AI, there are still some complexities, right? Like you need to have a certain amount of images, both good and bad, to train an algorithm. It does take time and money. For some manufacturers, it can mean, you're changing a process that's time-honored and all your employees understand. So there's a lot of even just communication to your employees for why you're trying to bring in these new technologies. And those are the types of things that create barriers for deployment.

The other big thing is just being a little bit too ambitious in what your deployment plan is, especially if you've got a process that is very manual and requires a lot of human decisions, to take that process and go to fully automated is quite a leap. So what we're finding best case scenario is break down that leap into a couple of more incremental projects like use some basic AI functionality to add decision support to a human operator process. You can start to benefit from the technology and that it can make a decision consistent and reliable. You can start to gain some trust from your employees around using this new process in your facility. And maybe more important, you can start to gain some data around some of the processes that exist so that you can then formulate, you know, a longer range automation plan.

Michelle: I thought it was just very helpful when you said don't aim to automate an entire, whole manufacturing line right away, just pick one difficult section and, you know, tackle that first.

Ed: I mean, we've been doing work with one manufacturer and they have a 10-year plan for automation. And eventually at the end of this 10 years, their hope is to, you know, deploy robotics but they recognize that that is a long-term investment that will require both time and money. So they've broken it down into these very logical steps of I would like to add decision support to this human decision process over the next couple of years and then gain the data from that to then automate the next step in my process so that in a decade I'm where I wanna be. Wow, that seems so long-term with AI. Who knows where AI will be in 10 years. Yeah, I mean, it'll be curious to see how their plan comes to fruition, right? But they know where they want to be. And they know that in their case, they're dealing with some labor issues that they foresee coming within the next decade. It's going to be harder to find employees to do this job. So that's part of what's driving their plan.