In Stay Sharp Episode 90, Unleashing Unstructured Data: AI’s Manufacturing Revolution, co-hosts Jonathan Scott and Juliann Grant are joined by Patrick Harrigan, the VP of Partnerships at CADDi, a manufacturing technology company specializing in AI platforms to manage manufacturing knowledge. Patrick shares his insights on how AI can help transform unstructured data into useful intelligence to positively impact your manufacturing business.
Defining Structured and Unstructured Data
Structured data is what most of us work with every day. It exists in digital form, it’s organized in some manner, and it can be easily searched or analyzed. Think about files in your computer’s file structure or in your CAD, PLM, ERP, or other enterprise platform. Unstructured data isn’t any of that. It has no regular structure or format. It might be qualitative or require special techniques for analysis, like audio or video files, notes on paper, or similar. It might be the knowledge from decades of shop floor experience that’s locked in senior workers’ heads. Industry estimates 80% or more of a manufacturer’s data is unstructured—that’s a huge volume of assets you’re not leveraging.
“It’s everyone’s dream to get that data into a digital, structured format so that we can do amazing things with it around digital twins and digital threads,” Patrick says. “And really the first mile is converting this unstructured data to a usable format.” Once that’s done, making the information useful is about adding context, which turns it into structured data. Jonathan explains, “It’s the connections to other systems. It’s the connections to other pieces of data. Those links are the structure. The metadata, the data models, those are the structure that gives everything context that we don’t have.”
What’s enabling this true digital transformation is the technology we have today. Information-gathering used to be a very manual process, with analysts working to draw conclusions. “Now, with the evolution of AI and how rapidly these tools are growing,” Patrick says, “you have bots that are constantly crawling these data sets and automatically drawing these correlations and providing this structure, which has rapidly accelerated our ability to work with structured data sets.” Juliann follows that thought to its logical conclusion, noting, “All this data becomes feeders to help really accelerate some of the plans for digital twins and digital threads.”
Managing Unstructured Data
To further differentiate structured and unstructured data, Patrick describes two kinds of systems that manage information. Systems of record generate net-new data that’s housed in the applications, such as PLM or ERP. He says, “A defined data model comes out of the box with those systems.”
Systems of insight, on the other hand, aren’t usually producing net-new data. “It’s more leveraging existing data from systems of record, drawing those correlations and providing that structure that we referred to previously, to ultimately serve up analysis and insights into these systems of record,” Patrick explains. The idea is a complementary system or platform—not a replacement—that can pull data from your enterprise systems of record, a little bit from here, a little from there, and so on. With that collection, manufacturers can drive a net-new insight and also drive their businesses.
Where AI Fits In
As mentioned, the first challenge is collecting your unstructured data. AI, which will be used at the end to derive insights from the data, can help with collection too. Technology like optical character recognition (OCR)—the ability to scrape PDFs, images, paper docs to extract text and numerical values—is foundational. But there have also been many advances around extracting shapes and geometries from images or paper documents.
Once you’ve captured the data, you can harness the power of AI to analyze everything exactly when you need it. Currently, almost all manufacturing operations still require a human to make a critical decision at some point. “How AI can really bring value to manufacturers in the short term,” Patrick says, “is the ability to serve up these insights in context when an end user needs them. It’s the ability to inject a recommendation, a corrective action, a suggestion into the existing workflow that I’m trying to achieve or the problem I’m trying to solve.” If we can do that, Patrick adds, “I think we can really accelerate manufacturing as a whole.”
Learn More About How to Accelerate Your Manufacturing Operations
The full podcast contains more discussion and details on a variety of topics, including how to ensure the AI you’re using is as close to 100% accurate as possible, how to find the sweet spot so the insights you get are accurate and valid, why it’s critical to break down organizational data silos, how an iceberg relates to your organization’s data, and more.
Be sure to check out Stay Sharp Episode 90: Unstructured Data Unleashed – AI’s Manufacturing Revolution, and join us each week for a new podcast.
Follow the Razorleaf Podcast, Stay Sharp in Digital Engineering on:
