In recent weeks, Kilpatrick’s E-Discovery Of Counsel, Nicole Allen, had the distinguished opportunity to serve as a panelist at The Masters Conference Legal in Chicago. Nicole participated in Relativity’s panel titled “Generative AI in E-Discovery: How to Test, Trust, and Thrive in a New AI Era.” The session featured an in-depth exploration of the evolving landscape of generative artificial intelligence within e-discovery.
Key takeaways from the panel include:
1. AI in E-Discovery is the future, but “best practices” still apply. Leveraging cutting-edge AI technologies allows us to be forward-looking andinnovative, but that does not mean we can skip the steps that ensure validation, reliability and trust. In fact, the integration of AI should serve to strengthen these core tenets—reminding us that technology is most valuable when it is coupled with the same care, scrutiny, and ethical standards that have always guided our work.
2. Why do we validate? Validating e-discovery results has always been a cornerstone of sound legal practice, and that does not change with the introduction of AI. Validating e-discovery AI is essential to ensure the tool delivers accurate and reliable results, reducing the risk of errors or missed documents. Should you need to show your work to courts or opposing parties, validation demonstrates that your use of AI is well-founded. It also acts as a quality control measure, builds trust among teams and clients, and supports continuous improvement by highlighting areas for refinement.
3. Practice Makes Progress: Start Smart, Test Thoroughly. Test any new AI tool by starting with settled matters and a small, manageable set of documents. Iteratively test and refine your prompts, always comparing the AI’s output to a human review to spot and address gaps. Adjust prompts or set aside tricky documents as needed, repeating this process until your results meet your quality standards. As you gain confidence, gradually expand your sample size. Use the same careful, step-by-step approach for live reviews. The more cycles you run with small sets, the stronger your validation rates will be, both quantitatively and qualitatively. High validation rates early on are a strong indicator the tool will perform reliably at scale.
4. To Disclose or Not to Disclose. There is not a universal requirement for disclosure of AI use in e-discovery, but transparency is often expected and may be required by the jurisdiction, court order, local rules, or case protocols. Where there is no specific guidance, you must determine whether transparency will help avoid disputes, support defensibility, and potentially strengthen your legal strategy or position in negotiations.
5. Communicate with your client. Always obtain the client’s permission before using AI tools and clearly discuss expectations up front. Keeping clients informed about any changes ensures transparency, builds trust, and helps avoid misunderstandings—supporting strong, defensible results and a collaborative attorney-client relationship.
6. Not all documents are equally suited for AI-assisted review. Documents that are particularly sensitive, complex, or nuanced materials may still require careful human analysis to ensure accuracy and context are not lost. While AI can greatly enhance efficiency and consistency in e-discovery, human review remains essential for quality control and for making judgment calls that technology alone cannot reliably address.
