I've been thinking a lot about this gap between "having data" and "having knowledge," and this article puts words to something I've felt firsthand while exploring AI tools at work.
Universities (and honestly most large organizations) sit on mountains of data: course catalogs, policies, research records, student systems. But an AI agent pulling from all that still won't know that "add/drop deadline" means something different for grad students than undergrads, or that one department's "advisor" is another's "coordinator." That context lives in people's heads and in years of institutional practice, not in a database schema.
The article's point that knowledge is really about meaning in context, not just volume of data, reframes a lot of the AI agent hype for me. It's not a matter of just plugging an LLM into our systems and calling it done. The hard, unglamorous work is capturing institutional knowledge in a way machines can actually use, which is a lot closer to the work of documentation, information architecture, and change management than it is to flashy model demos.
As someone moving from web dev into AI work, this is reassuring in a way.
https://www.technologyreview.com/2026/10/05/1145580/connecting-ai-agents-to-enterprise-knowledge/ #AI #EnterpriseAI #HigherEd