Box CEO on how generative AI can turn the mostly unstructured data inside enterprises into valuable knowledge, enabling greater automation of workflows and more
Aaron Levie / @levie : X: @hunglee , @villi , @adambrotman , and @tomwarren X: Hung Lee / @hunglee : Structured vs Unstructured data - AI will collapse the distinction and make it all accessible knowledge 👇🏽 @villi : This is really exciting. When I was at Box we were hoping NLP models can help us extract basic metadata on files like business contracts. Today, all of your unstructured data can come to life and become a strategic asset. Adam Brotman / @adambrotman : “for the first time ever, generative AI actually lets us talk to our unstructured data. Multimodal models especially allow us to process this content using a computer and essentially perform any task that a human can, but at infinite scale and speed.” Tom Warren / @tomwarren : good perspective on why AI matters so much to businesses. While there's a backlash on the creative side to generative AI, that's just a part of what AI models unlock. AI can and will be a powerful tool for a lot more than just generating content
Context & Ripple Effects
The discussion extends an established enterprise-AI use case: applying models to work artifacts rather than only generating content. Earlier coverage found generative AI already reshaping work in contracts, professional services, filmmaking, and programming through task-specific workplace adoption.
Box’s framing pushes that logic toward the enterprise file repository: documents, contracts, and other multimodal records can be queried and classified as knowledge. It also fits reports that AI-assisted RFP work raised sales-team productivity at major software companies, including AI-assisted responses to customer proposals.
First-order effects
- Enterprise teams can use generative and multimodal models to extract metadata, retrieve information, and summarize content from previously hard-to-query files, making more existing records usable in workflows.
- Box and similar enterprise-content platforms gain a clearer AI value proposition: turning stored unstructured material into an operational input rather than treating it solely as archived content.
Second-order effects
- Workflow-software vendors face pressure to embed AI search, extraction, and automation around the documents their customers already hold; contract-management software is a particularly relevant adjacent category, where AI expectations have already spurred deal activity.
- The value shifts from merely storing files to governing access, context, and reliable outputs from them, increasing the importance of integrations between content repositories and workflow tools.
Third-order effects
- If these deployments prove dependable, the boundary between structured databases and document-heavy systems may matter less to end users, with AI acting as the interface that converts both into workflow-ready knowledge.
- Enterprise AI competition could increasingly center on distribution through systems of record and on the quality of permissions, retrieval, and process integration—not just access to a general-purpose model.
The trend: Generative AI is moving from standalone assistants toward workflow-native systems that make enterprise content searchable, interpretable, and actionable.