About

Litigation runs on facts. Software should treat them that way.

Every case is won or lost on its record — thousands of pages of pleadings, medical records, transcripts, and productions that hold the facts, the dates, and the contradictions that decide outcomes. The teams that master the record win. The tools for mastering it haven't kept up.

FactMarshal exists to close that gap. We build AI litigation intelligence with one non-negotiable design rule: the system shows its work. Every fact we extract is pinned to the page it came from. Verbatim facts are machine-verified against the source. Everything else goes through human review built for speed. The chronologies, contradiction reports, and answers our assistant produces are assembled from that verified record — never from a model's unsupported say-so.

How we build

Built by practitioners. FactMarshal was founded by a litigator who practiced for more than a decade before a line of it was written — someone who has signed the filings, sat through the privilege fights, and lived with what a missed fact costs — and a machine-learning PhD who has spent fifteen years making extraction systems accurate enough to rely on.
Citation-first, chat-second. Fluent text is easy; verifiable records are the product.
Human in the loop, by design. AI does the reading; your team does the deciding — with tooling that makes verification as fast as reading.
Privilege as architecture. Firm-level tenant isolation and matter-level access roles are built into the schema, not bolted on. Client files are never used to train models.

We're a small team, and the two people below are responsible for what the product is. It's opinionated because our experience made us opinionated: about verification, about citations, and about never asking a lawyer to trust what they can't check.

The founders

Co-founder · Litigation

Amanda Coop

Amanda has spent more than a decade in litigation, on every side of the injury case. She spent seven years as staff counsel for one of the country's largest insurers, trying bench and jury cases as first and second chair. She then founded her own plaintiff-side firm focused on medical malpractice and premises liability, and now practices as a partner at a Nashville litigation firm, where she leads the office. She is also a Rule 31 listed mediator with the Tennessee Supreme Court's ADR Commission.

That range — defense counsel, plaintiff's counsel, neutral — is why FactMarshal treats the record the way it does. The verification queue, the contradiction reports, the citation on every answer: each exists because she has watched cases turn on a fact nobody flagged. She holds a JD from Appalachian School of Law, is admitted in Tennessee and Georgia, and speaks regularly on the use of AI in legal practice.

Co-founder · Machine learning

Robert Coop, PhD

Robert has spent more than fifteen years making machine-learning systems accurate enough to act on. He earned his PhD at the University of Tennessee, Knoxville — doctoral work on how neural networks forget that is still used as a baseline in the field — and started out in research at Oak Ridge National Laboratory. In industry, he founded and led the AI and machine-learning team at Stanley Black & Decker, served as Chief AI Officer at Advent Health Partners, where the work ran on reading medical records at scale, and led AI at WorkWave.

FactMarshal's extraction pipeline is his answer to the problem that has run through all of it: model output is only useful when a professional can stake their name on it. Machine verification for verbatim facts, human review for everything else, and a citation on every answer — because "trust me" is not an engineering spec.

Want to see it on a real matter?

Request a demo