Introducing Trust Reader: Personr's New Tool for Automated Deed Extraction
Every trust and SMSF deed hides its key details in hundreds of pages of legal text.
Personr's new Trust Reader finds them in seconds, designed to automatically extract all material parties from trust deeds and self-managed super fund (SMSF) deeds.
At launch, it exceeds the roughly 90% global accuracy benchmark, reaching 92% or higher.
For trusts, Trust Reader identifies trustees, settlors, beneficiaries, appointors and protectors.
For SMSFs, it identifies individual or corporate trustees, members, and any other roles relevant to ownership and control.
What It Does
When someone sets up a family trust or a self-managed super fund, all the important details are contained in one legal document: the deed.
These documents can run hundreds of pages long, and now, Trust Reader can read them for you.
Every extraction includes page-level source references for a full audit trail, along with a confidence score on each field to guide human oversight.
"Australia has more than 600,000 self-managed super funds, plus countless family trusts, and every single one has a deed that someone has to read at some point," said Nick Ahrens, Personr co-founder.
"Our tool does that reading in seconds instead of hours, cutting manual work by more than half and making mistakes far less likely."
Why It Matters
Trusts and SMSFs have no public register.
Unlike a company, you can't simply look up who owns or benefits from them; the only way to find out is to read the deed itself.
These documents are often hundreds of pages long and difficult to navigate, which slows the process and increases the risk of human error.
With Trust Reader, accountants, banks, and financial advisers save valuable time they can redirect toward higher-value work.
The Technical Side
Trust Reader handles digital or scanned PDFs up to 300 pages, using OCR combined with large language model extraction.
It returns structured JSON via a REST API in under 60 seconds per document.
Key capabilities include:
Batch uploads for processing multiple documents at once
Configurable role taxonomy to handle non-standard trust structures
An initial fraud-signal layer that flags metadata tampering and OCR anomalies
