Transforming eDiscovery in a GenAI World

As our survey shows, document review in eDiscovery has made progress through centralized approaches, but it still requires heavy effort and oversight. Its high cost is cited as the top challenge, more than twice as likely to be raised as concerns about work quality. But for 35% of senior legal leaders, adopting new technology or using it effectively still is a major eDiscovery obstacle; new-tech adoption is over three times more likely to be a challenge than data volume. Tech skills gaps trouble 22% of leaders, and 21%, particularly in nonprofit organizations, report insufficient time to complete document review.

Leaders are looking to AI for quality insights, cost reduction, and speed improvements. So, it isn’t a surprise to find that summarization is the leading GenAI use case (identification of personally identifiable information/protected health information, privilege review, deposition prep, first-pass review), with 57% of legal leaders now use GenAI for summarization. A strong majority expect near-term gains—87% believe GenAI can improve eDiscovery efficiency within 24 months—and 47% expect improvements within 12 months. Only 2% disagree that GenAI can improve efficiency, and 6% have no plans to use it.

The unresolved questions around legal AI ethics keep firms from rushing in: 93% of senior legal leaders are concerned about ethical and privacy risks from GenAI in eDiscovery, and 92% see security risks. The same share (92%) worry about AI hallucinations. Reputational damage ranks as the single greatest concern for 28% of respondents, while only 6% are primarily worried about court acceptance of GenAI outputs. Organizational views on security risk differ by sector: 54% of public organizations, 45% of private organizations, and 38% of government organizations say GenAI poses a significant security risk.

Forty percent of respondents prefer an in-house delivery model using an organization-specific large language model (LLM); 58% report their legal team already has an organization-specific LLM, and 35% of those without one are planning to build one. That said, the report notes a practical alternative: Go-to-market solutions built for eDiscovery can help protect proprietary data, offer third-party oversight, and often accelerate time-to-market.