Family Docket AI
Rotman School of Management · University of Toronto

Legal document intelligence · Research prototype

The hard part of family law isn't reading the documents. It's knowing which law applies.

A contested family matter buries people in paper — receipts, statements, correspondence, court filings. Family Docket AI reads that pile, organizes it, and points each document to the law it actually bears on.

Ontario family law Built for self-represented litigants Synthetic data only · not legal advice

Hundreds of documents. One matter. No time.

A single contested family matter in Ontario can generate hundreds of documents that all have to be gathered, sorted, disclosed, and tied to the specific issues in dispute — support, parenting time, the matrimonial home, the division of property.

People with lawyers pay by the hour for this work. People without lawyers — a growing share of family court — do it by hand, late at night, or not at all. The organizing is where matters stall before the law is even reached.

A majority of family litigants — priced out of the help they need.

40–57%
of people in family court appear without a lawyer — up to 80% in urban centres
$25k–$100k+
typical legal cost to each side of a family case that reaches trial
~$7.7B
spent out of pocket every year by Canadians facing legal problems
~1 in 2
adults face a serious legal problem in any three-year period

Family matters run on documents — hundreds of receipts, statements, and filings that must be gathered, organized, and tied to the issues in dispute. A litigant can have the stronger case on the merits and still lose it because their disclosure is incomplete, late, or disorganized. That is the gap Family Docket AI is built to close.

Sources: Department of Justice Canada · National Self-Represented Litigants Project · Canadian Forum on Civil Justice.

From a folder of documents to an organized, law-linked record.

Every document moves through the same five steps — automatically.

01

Ingest

Read the text out of each uploaded document, whatever its format.

02

Classify

Identify what kind of document it is and the family-law issue it concerns.

03

Extract

Pull out the details that matter: parties, dates, amounts, a plain summary.

04

Map to law

Link each document to the provisions of family law it likely bears on.

The hard part
05

Organize

Produce a structured record a person can actually review and file with.

Reading a document is easy for AI. Knowing the law is not.

Modern language models read and summarize documents reliably. But asking a search engine "which statute governs this?" is a different problem — and getting it confidently wrong is worse than staying silent.

Asked for the law governing a couple's matrimonial home, naive legal search returned the Income Tax Act — correctly cited, and completely wrong.

A real, properly-cited Canadian statute. Also not the law that divides a family home. For someone without a lawyer, a plausible wrong citation is a trap.

Our approach: encode the legal judgment, don't leave it to keyword search.

Family Docket AI routes every document through a curated, lawyer-reviewed map from each family-law issue to the provisions that actually govern it — because in Ontario, which statute applies can turn on something no keyword match can decide, like whether a couple was married. That map is the defensible core of the project, and the part a lawyer keeps sharp.

An honest research prototype — and a clear next mile.

Working today

Read, classify, extract

An end-to-end pipeline that classifies synthetic test documents and pulls out their structured details, measured against a hand-built answer key.

The research frontier

Mapping documents to law

The step that makes the tool valuable — and hard. We have a measured result on why naive search fails, and a curated map as the fix under active development.

The discipline

Synthetic data only

No real family-law documents, from any matter, ever. Every test document was authored for this project — keeping the work clear of privacy and ethics obligations.

Bridging engineering, law, and AI.

SO

Saba Owji

MBA Candidate · Rotman School of Management, University of Toronto

Family Docket AI is the work of Saba Owji, whose background spans civil engineering and the management of real estate and capital projects — disciplines built on turning overwhelming volumes of documentation into something orderly and accountable.

The project grew out of an independent, faculty-supervised research study that bridges her graduate law coursework — taken on exchange at Imperial College London — with her Rotman MBA. It brings a builder's instinct for systems to a problem the legal system too often leaves to hand and to chance: giving people without a lawyer a real chance to organize their own case.

The study proves the idea. The ambition reaches further — to build Family Docket AI into a tool that self-represented litigants and family-law practices can actually rely on.