As the company pushed to make data a shared asset across business and engineering teams, its existing metadata setup could not keep pace. Out-of-the-box catalog fields captured generic technical detail, tables, columns, types, but not the business context that determined whether a dataset was fit for a given use case. Analysts and engineers who wanted to build on existing data assets, or feed them into emerging AI agents, had no reliable way to tell what a dataset was for, who owned it, or how current it was, without asking around.
Enriching metadata to answer those questions was a manual, recurring task split across the data team, consuming hours every week that could have gone into higher value work. With growing internal demand for self-service access to trusted data, from business teams and from AI agents alike, the company needed a way to package data as discoverable, well-described products rather than raw tables.