At the MLOps Sweden meetup last week, hosted at AI Sweden, I had two conversations that surfaced something worth spending time on.
The first was with an insurance executive who described a big AI challenge: decades of claims documents, underwriting reports, and customer correspondence sitting in filing cabinets and legacy systems. "We know there's gold in there," he said. "We just can't mine it."
The second was a startup founder presenting later that evening. His company helps dental practices transform the chaos of patient information - X-ray images, handwritten notes, voice memos, scanned documents - into structured patient records that are fed into their systems. What used to take hours of admin work now happens in minutes.
Same insight, different industries: the real AI opportunity isn't in the data you already have organized. It's in the 90% you've never been able to use.
Here's a number that should get your attention: up to 90% of enterprise data is unstructured - text documents, images, videos, audio recordings, emails, PDFs, handwritten forms. This data has been accumulating for decades, but its messy nature made analysis nearly impossible at scale.
Traditional systems were built for structured data - neat rows and columns that fit in databases. When you tried to do something with a scanned contract, a recorded customer call, or an engineering report with embedded diagrams, you hit a wall. The information was there, but locked away.
Generative AI changes this equation fundamentally.
The breakthrough isn't that AI can now "read" documents - OCR has existed for decades. The breakthrough is that AI can understand context, extract meaning, and transform messy, inconsistent, multi-format data into structured information that integrates with your systems.
In insurance, AI now processes the artifacts that actually land on an underwriter's desk: scanned handwritten claims, 50-page risk reports with embedded tables, policy documents with inconsistent formatting. Early adopters report 30-40% reductions in processing time and significant improvements in risk assessment accuracy.
In healthcare, AI scribes turn real-time conversations and exam findings into structured clinical notes. Voice recordings, images, and handwritten notes become chart-ready documentation - cutting documentation time by more than half.
The pattern is the same across industries: data that was too messy to process is becoming accessible. Customer calls become analyzable text. Technical drawings become searchable databases. Decades of paper files become queryable knowledge.
According to MIT Sloan Management Review, if 2025 was the year organizations realized GenAI has a value-realization problem, 2026 is the year of doing something about it. The shift is from individual productivity tools to enterprise-level deployment.
One of the clearest paths to enterprise value is unlocking the unstructured data that's been accumulating for years.
It might seem like a counterintuitive insight, but your competitive advantage might not come from having better AI than your competitors. It might come from having more data that's finally accessible. Every organization has this dormant asset - the question is who moves first to activate it.
This week, inventory the unstructured data in one department. What's sitting in email archives, shared drives, paper files, or legacy systems that nobody can easily search or analyze?
Then ask: if this data were structured and accessible, what decisions would we make differently? That's your unstructured data goldmine - and generative AI just gave you the tools to mine it.