- Document AI reads certificates of incorporation, shareholder registers, trust deeds, and registry extracts, and returns owners, roles, and holdings as structured data.
- Entity resolution matches the same person or company across documents, registries, and your customer records, even when names, spellings, and addresses differ.
- Each link in the ownership chain points back to the page it came from, so your analyst can check it before approving onboarding or a review.
- Everything runs in four deployment options: on-premises, air-gapped, private cloud, and cloud, so client documents stay in your environment.
The real owner sits three layers down
Your onboarding and financial crime teams have to know who stands behind a business client. The answer is often spread across a holding company in one country, a trust in another, and a nominee director in a third.
Your analysts read the documents, key the names into a spreadsheet, search registries, and draw the structure by hand. The same person appears as slightly different spellings, and a company they control, and matching them depends on who does the review.
Shell companies and layered structures are built to exploit that. A missed link can hide a sanctioned party, a politically exposed person, or a mule network behind a clean-looking client, and the loss lands on your books.
How beneficial ownership intelligence works
Document AI reads the documents, entity resolution links the names across sources, and your analyst sees the ownership chain with the source behind each link.
Step 1: read the documents
Extractors use intelligent character recognition: the model reads each page the way a person would, including scans, tables, stamps, and handwriting. One schema covers names, roles, ownership percentages, share classes, addresses, dates, and registration numbers, across many layouts and languages.
Step 2: resolve the entities
Named entity recognition finds the people and companies on each page. H2O Driverless AI can train a classification model on name pairs your analysts have already confirmed, and score new pairs as the same entity or not, with reason codes your team can read.
Step 3: build the ownership chain
Matched entities and their holdings form an ownership graph. The chain runs from your client to the natural persons at the top, and each link carries the document and page it came from.
Step 4: flag and review
Your rules decide what needs a closer look: an owner above your ownership or control threshold, a circular structure, a nominee pattern, or a name that also appears on your screening results. Your analyst reviews the flags and approves, corrects, or escalates, and the file keeps the evidence.
Where it fits in your financial crime program
Beneficial ownership sits between onboarding, KYC, and screening. H2O works beside the KYC, screening, and case management systems you run today; they stay the record, and this adds the reading and matching they depend on. H2O makes no claim of compliance with beneficial ownership rules, FATF recommendations, or BSA/AML obligations: your compliance team and your regulator decide that.