Classifai.

What it does

Classifai classifies structured data — the tables, columns, and fields inside your databases and warehouses — and turns that classification into documentation you can actually govern from. It produces a data dictionary and a set of field definitions, delivered as a reviewed PDF, and a platform to keep both current as the estate moves underneath you.

It exists because most institutions cannot answer a simple question about their own data: what is in this column, who owns it, what sensitivity does it carry, and is that still true this quarter. Catalogue tools assume someone already documented the answer. Usually nobody did.

How it works

You bring the taxonomy — your codes, your sensitivity levels, the edge cases your data team argues about. Classifai reads schemas, samples values, and proposes a classification for every field, with the reasoning attached: what evidence in the data supported the label, which rule or definition it matched, and how confident the system is. Reviewers accept, correct, or override, and corrections carry forward rather than being lost on the next run.

The output is two things at once: a document your governance function can act on, and a live set of definitions your systems can query.

Deployment & governance

Runs as a hosted platform, as an API your own pipelines call, or entirely on-prem where residency rules require it. Every classification is logged with its inputs, model, version, and rationale, so an audit is a query rather than an archaeology project. Nothing about it is domain-specific — the same pipeline runs against a health registry, a payments ledger, or a national asset register, because the taxonomy is yours, not ours.