Extracting UBOs from documents (OCR)

Last updated: June 17, 2026

Overview

Ondorse can read ultimate beneficial owner (UBO) information directly from uploaded documents, helping you fill in UBO data faster and keep declarative data consistent with supporting documents.

This is built around three capabilities that all use the same underlying extraction model:

Who this is for

These feature are most useful for two situations:

  • Highly regulated clients who need to demonstrate coherence between declared data and the underlying documents: a pure compliance need (e.g. being audit-ready for the ACPR).

  • Clients with entities outside France, where UBO data isn't easily accessible from registries. Here the benefit is mainly operational: less manual data entry and fewer back-and-forths with the end customer.

Supported scope

UBO extraction is tested and supported for the following countries:

๐Ÿ‡ง๐Ÿ‡ช Belgium ยท ๐Ÿ‡ช๐Ÿ‡ธ Spain ยท ๐Ÿ‡ณ๐Ÿ‡ฑ Netherlands ยท ๐Ÿ‡ฎ๐Ÿ‡น Italy ยท ๐Ÿ‡ฉ๐Ÿ‡ช Germany

Important limitations

Please read these before relying on the feature in production:

  • Test phase recommended. AI extraction isn't considered fully production-ready for a given client until it's been validated on a large enough set of that client's real documents. We strongly recommend running a test phase on your own document set before depending on it.

  • Shareholder graph reliability varies. Shareholder graphs come in many formats, so extraction reliability on non-standard graphs is lower.

  • "Compare with content" is scoped to UBO extracts. Extending it to other document types is a feature request.

Getting started

If you'd like to trial UBO extraction on your account, contact your Ondorse point of contact or reach out via support. We'll help you set up a test phase and review results together before switching it on for live cases.