Blog Post
Measure Twice, Cut Once: Generative AI for E-Discovery in Public Sector Disputes
Senior Managing Director,
FTI Consulting
Like nearly every industry, the public sector is considering whether and how they may be able to use artificial intelligence in their day-to-day work. In e-discovery, practical use cases are already in play. Public sector legal teams can build upon these to achieve new efficiencies and to lay a foundation for future adoption of new forms of AI and generative AI as they become more reliable and sophisticated.
The use of AI isn’t new in e-discovery. Predictive coding, continuous active learning and machine translation are examples of “traditional” AI, which has been reliably and defensibly used in courts for more than a decade. Generative AI adoption in e-discovery is also moving quickly, and whilst it may have a long way to go before it can reach the same degree of acceptance as traditional AI, with proper human oversight, it has the potential to accelerate certain steps in the e-discovery process, such as document summarisation for early case assessment, enhancing OCR output and classification (review) of documents.
The key is to approach generative AI with the same benchmarks and rigour that should be applied to every aspect of e-discovery workflow and technology. Public sector legal teams can improve their use of traditional AI by focusing on reasonableness, proportionality, transparency, explainability and accuracy, and begin to explore opportunities for generative AI to support additional efficiencies. So, what can this look like in practice?
As noted, one key example is to leverage generative AI in the early case assessment phase. Using targeted and tested prompts, legal teams can uncover and extract facts and determine relevant themes, keywords, and individuals, as well as insights that can help counsel form their strategy early on in a matter.
Generative AI’s ability to summarise documents and create chronologies based on information within a document set can also help legal teams improve efficiency in e-discovery while keeping a human in the loop. Targeted prompts are important; the better the prompt, the more accurate the results will be. The results of a summarisation exercise can support document prioritisation for review and also help the case team verify results for quality control. It’s a helpful tool to evaluate whether the AI results say the same thing as the lawyer who completed an analysis. Differences can bring to light new issues that may need to be explored, or mistakes in the review process that may need to be addressed.
Processing and analysing low-quality scanned documents or illegible handwriting in hard copy documents converted to PDF are common challenges in e-discovery. OCR is a helpful tool for dealing with these issues, and generative AI shows potential for enhancement. Large-language models have the power to predict the next likely word in context or carry on a sequence of text or characters, which can help fill gaps and clarify hard-to-read scanned documents and images.
What next?
When implementing AI in e-discovery, legal teams should focus on foundational principles of quality, proportionality and defensibility and experiment with AI solutions that can support their workflows without adding undue risk. Working with experts who understand the best practices of e-discovery and the technical nuances involved in applying AI, legal teams can follow these steps to incorporate technology into their workflows and verify its reliability:
- Targeted sampling of a small sample of documents to test inputs and outputs.
- Initial test of prompts used against the targeted sample to validate the results further.
- Keep running the prompts to iron out deficiencies and achieve a defensible level of accuracy.
- Evaluate and validate results using a repeatable and explainable methodology, suitable for the task at hand.
- Apply broadly to summarise more documents and continue quality control.
Public sector legal teams will continue to face data challenges in e-discovery and investigations. A sensible and reliable approach to leveraging technology in these matters is critical. Likewise, teams should look to experienced, reliable technical experts who can help navigate these issues and ensure defensible, accurate, timely discovery of actionable information, no matter the type of case or the type of technology being used to respond to it.
Related topics:
The views expressed herein are those of the author(s) and not necessarily the views of FTI Consulting, its management, its subsidiaries, its affiliates, or its other professionals.