Harvey builds a shared fact record for litigation

Overview

Updated 1 hour ago

A long lawsuit generates millions of documents and a factual record that has to be rebuilt whenever a new lawyer joins the team. Harvey now sells a system that extracts facts from those documents, links each one to the exact line it came from, and keeps the record current as new files arrive.

The company showed the capability at its FORUM Chicago event and opened it to early access. Litigators are about one-third of Harvey's customers, which is why the firm built a dedicated product for them rather than another general-purpose chat window.

Why it matters

If a firm can trust a machine-built fact record, months of document review collapse into a query, and every draft inherits it.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

One-third

Share of Harvey's customers who are litigators

Harvey cites this as the reason for building a litigation-specific capability.

Millions

Documents the fact graph is built to scan per matter

Harvey's stated scale for a single matter's document set.

Harvey products carrying the Jus Mundi arbitration agent

Assistant, Agents, Vault and Workflows, per the partnership announcement.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Winston Weinberg

Co-founder and chief executive of Harvey

Leading the company's expansion from research and drafting into matter-level workflow

Gabriel Pereyra

Co-founder and president of Harvey

Overseeing product development with design partners

Richard Jeens

Disputes partner and member of the innovation committee at Slaughter and May

Design partner on the fact graph

Organizations Involved

Harvey

Legal AI company

Shipping fact graph to early access; building with design partners

Legal AI company whose products cover research, drafting, and now matter-level fact management for disputes and investigations.

Slaughter and May

Law firm

Design partner testing the fact graph on disputes and investigations

London-founded firm whose disputes and investigations practice is helping shape Harvey's litigation workflows.

Jus Mundi

Legal research platform

Arbitration agent integrated into Harvey's products

International arbitration and cross-border disputes research platform whose agent now runs inside Harvey.

Timeline

2 events Latest: Yesterday

November 2026
Jus Mundi arbitration agent placed inside Harvey
Latest Partnership

Arbitration research runs in Harvey's Assistant, Agents, Vault and Workflows.
Yesterday
Harvey unveils fact graph at FORUM Chicago
Product launch

Company shows a shared, source-linked fact record for disputes and internal investigations.
3 days ago

Scenarios

Litigation teams make the fact record their working file

Possible Resolves by Q2 2027

Discussed by: Harvey and design partner Slaughter and May

Early-access firms feed new matters into the fact graph, and chronologies and witness outlines get drafted from the shared record instead of from scratch. Adoption would show up as named customers and case studies rather than waitlist announcements.

Extraction errors keep lawyers checking every link by hand

Possible Resolves by Q2 2027

Discussed by: Independent commentary on the early-access launch, including Hanging Context

The system's value depends on facts being pulled correctly and inconsistencies flagged often enough that lawyers stop ignoring the warnings. If error rates stay high, firms keep the manual record and use the tool only for first-pass review.

A rival ships a comparable matter record and closes the gap

Uncertain Resolves by End of 2027

Discussed by: Legal technology trade coverage

Fact extraction linked to sources is a software problem, not a proprietary dataset. Competitors selling document review or research tools could add a shared matter record, turning the fact graph into one feature among several rather than a reason to switch vendors.

Historical Context

2 moments from history that rhyme with this story — and how they unfolded.

1973-1975

Lexis and Westlaw go online (1973-1975)

Full-text legal research moved from printed indexes to terminals, letting lawyers search case law directly instead of paging through digests.

Then

Large firms adopted terminals slowly, and hourly billing for research time became a subject of client complaints.

Now

Search became the default way lawyers find authority, and the databases became the commercial core of the legal publishing industry.

Why this matters now

Search answered 'where is the case.' The fact graph attempts the harder question of what the documents in one matter collectively establish, which search never tried to hold.

February 2012

Da Silva Moore v. Proskauer (2012)

A federal magistrate judge in New York approved using predictive coding, a machine-learning method, to screen a large document set in an employment discrimination case over the objection of the opposing party.

Then

The ruling gave litigators a citable precedent for machine-assisted review of evidence.

Now

Technology-assisted review became routine in large document productions, with courts and parties arguing over transparency and validation.

Why this matters now

It set the terms of the argument now repeating: a machine can review the documents, but the process has to be defensible and auditable. Harvey's source-line linking is aimed at that requirement.