Defensibility in the Age of GenAI Review: What Every Practitioner Needs to Know

Generative AI has changed what is possible in document review, but it has also raised new questions about how legal teams can trust, validate, and defend the results. How do you know a GenAI review system is working? How do you explain its decisions to opposing counsel, a court, or your client?

Join Bennett Borden, a practicing litigator and data scientist with more than 25 years of experience, and George Socha, a recognized authority on legal technology and discovery standards, for a candid conversation about defensibility in GenAI-assisted review. Drawing on Bennett's recent independent evaluation of Reveal's aji GenAI review system, the discussion will walk through the practical questions practitioners should be asking before adopting any GenAI tool, along with the empirical evidence that helps answer them.

What You'll Take Away

• A framework for evaluating whether a GenAI review workflow is defensible under professional and discovery standards
• How to approach prompt or "definition" creation, calibration, and validation in a repeatable, testable way
• What agreement rates, precision, and recall mean in practice when working with a GenAI system
• How transparent reasoning and in-text citations strengthen your ability to explain results to clients, opposing counsel, and courts
• Lessons from an independent empirical evaluation of a GenAI review system tested against more than 33,000 documents

Whether you are considering your first GenAI review project or refining an established workflow, this session will give you a grounded, practitioner-focused view of what defensibility looks like today.