AI Auto Data Ingestion (Statements + Emails)

Decoder

PDF statements and email trade confirmations entered manually one by one — Ginkgo AI auto-extracts transaction fields and validates them into the ledger with rule checks.

SCENARIO

Business Scenario

Family offices receive PDF statements from 5-10 private banks / brokers every month (different templates / multi-page / bilingual Chinese-English) plus a large volume of trade confirmation emails (different formats / with attachments / bilingual). Traditional manual entry by PMs / relationship managers is slow and error-prone — PMs often spend 50% of month-end time on data entry. Errors and omissions are routine, discovered only at month-end reconciliation when it is already too late.

MECHANISM

Mechanism · Core Engine

Decoder

Dual-track parallel AI ingestion — (1) PDF path: AI large model recognises PDF statement layouts (UBS / CS / Citi and other templates), auto-extracts transaction fields (date / instrument / quantity / price / FX rate) + rule validation before structuring into the database. (2) Email path: AI large model recognises email trade intent (buy / sell / subscription / redemption / rebalancing / dividend etc.), extracts transaction fields + rule validation before auto-posting to the ledger.