In Stay Sharp Episode 115, AI + PLM—The Data Readiness Reality Check, co-hosts Jonathan Scott and Juliann Grant tackle practical approaches to AI with PLM. They’re joined by Graham Law, who is a PLM Solution Architect for Razorleaf, as well as a PLM industry veteran and former IT manager. That background makes him well qualified to discuss the challenging intersection of IT, PLM, and AI.
Good Data is Critical for AI
No matter what industry you’re in, your AI must be working with good quality, relevant data, because the only way AI can accelerate the way your business works is if it’s got the right data to draw upon. For engineering and manufacturing, that means good quality engineering and manufacturing data. Graham defines “AI-ready data for engineering” as data that’s clean, structured, rich in content, relational, and interconnected. He says, “Engineers already spend a massive amount of time searching for data, recreating things that they’ve lost that already exists. AI can help solve that, but only if you give it the right foundation.”
The challenge of feeding AI good data starts with gathering and connecting data from different silos, but it doesn’t end there, Graham says. “On top of that, even in the PLM system itself, you can get inconsistent naming. People call a bolt several different things. You’ve got missing metadata. Most PLM systems have got duplicates all over the place and obsolete parts that are still floating around.” With AI perhaps more than any other tool or human, what you risk with bad data is the supremely confident incorrect answer.
How to Become AI-Ready
The best place to start to become AI-ready is with a data audit—to get control of your engineering data. Graham explains, “That’s normalizing it, classifying it, and cleaning up any relationships between your parts, your CAD changes, and all that kind of stuff. Basically all the stuff that gives engineering data meaning.” Jonathan suggests that doing so will reveal your process to you, “Understand where your data is today, where the gaps are, the problems are, and then think about how you’re going to tackle it. That can lead you to a plan.”
The next step is governance, two ways. The first is controlling how your data is being created—“Stop making bad stuff to start with,” says Jonathan—so that it’s ready to be fed to AI. The second part of governance is establishing access controls and security for your data and deciding what AI is allowed to see. Graham’s opinion is that a PLM system is an important tool to help you do that. “You need it for your structure. You need it to connect all your engineering data to things like manufacturing, quality, service suppliers, etc.,” he says.
But Graham also urges organizations to not be overwhelmed and try to “boil the ocean.” Graham advises starting small, with something like summarizing ongoing change requests or automating part classification. Even onboarding new engineers. “An interesting one,” Graham says, “is finding insights from the data that as a human, you’d probably miss because you can’t obviously handle that amount of data. The AI can do it and maybe flag things that are wrong. Maybe this set of parts keeps failing. Customers don’t like this color, let’s change it to blue. All sorts of crazy things.” Think about what you want as a result, and plan tasks for your AI to tackle accordingly. Graham adds, “You don’t need to boil the ocean and you don’t need everything to be perfect before you embrace AI. Maybe just expose it to a certain product, to a certain data set.”
Jonathan agrees with easing into your use of AI, and he compares implementing AI to using search capabilities, suggesting you get comfortable and confident with the basics, then try something a little more advanced. “Everybody’s situation is different,” he says. “Everybody’s digital thread is different. There are a lot of common things that people use, but when you get into their specific case, it starts to get pretty unique.” Just like one size doesn’t fit all, Graham says one approach won’t fit all companies. With one exception: “If you want to embrace AI, start prepping your data now,” he says.
Learn More About Making your Engineering Data AI-Ready
The full podcast contains more discussion and details on a variety of topics, including the Ferrari fuel analogy that will help you understand the importance of AI-ready data, why you’ve got to be careful using existing enterprise libraries to populate AI, who in the organization needs to be involved with the data audit and data cleaning, and more.
Be sure to check out Stay Sharp Episode 115: AI + PLM—The Data Readiness Reality Check and join us each week for a new podcast.
Follow the Razorleaf Podcast, Stay Sharp in Digital Engineering on:
