In this short explainer, Box CTO Ben Kus breaks down the next frontier of enterprise data: combining unstructured and structured information so you can use the best of both worlds. Generative AI has already made it possible to extract structured information from unstructured content, pulling key clauses from contracts, surfacing insights from research proposals, and converting visual assets into queryable data. But the real breakthrough comes when you stop treating these two data types as separate problems. Ben explains what it looks like when you keep unstructured and structured data together: you can find all contracts related to a specific topic and immediately surface the most important limitation of liability clause. You can pull a set of research proposals and read through them to understand their key insights on a given subject, all within a single workflow. The power of querying, sorting, and filtering structured data combines with the ability to read, understand, and reason over unstructured content. For enterprise teams, this is the shift that unlocks the full value of your content. AI is only as good as the content it can access, and most of that content is unstructured. FAQs Q: What is unstructured data, and why does it matter for AI? A: Unstructured data includes content like contracts, research proposals, images, videos, and blueprints, anything that doesn't live in a traditional database or spreadsheet. It matters for AI because most enterprise knowledge is stored in this format, and AI can only deliver value when it can access and reason over that content. Q: How does generative AI work with unstructured data? A: Generative AI can read unstructured content and extract structured information from it — for example, pulling key clauses from a contract or identifying the core findings in a research proposal. That extracted data can then feed into analytics systems alongside your existing structured data. Q: What does it mean to "combine" unstructured and structured data? A: Rather than converting unstructured content into structured data and discarding the original, the goal is to keep both together. This lets you use the querying, filtering, and sorting power of structured systems while also being able to read and understand the full unstructured source so that you get the context, not just the extracted fields. Q: What are practical examples of this combined approach? A: Ben Kus gives two examples: finding all contracts related to a specific topic and then surfacing the most important limitation of liability clause; and pulling a set of research proposals and reading through them to understand their key insights on a given subject. Both require the ability to query at scale and understand content in depth at the same time.

 51040      2