The State of AI in eDiscovery: From Hype to Real-World Impact

Summary: According to our second annual State of AI in eDiscovery report, AI skepticism is being replaced with curiosity. Discover the AI adoption and application trends experts think will have the biggest impact on eDiscovery and your work. Leaders from Meta and Redgrave recently joined Lighthouse for an enlightening conversation—here are the highlights.

‍

AI is accelerating in legal—in adoption, advancements in understanding, and maturing views of both the opportunities and risks it presents. During our recent webinar, State of AI in eDiscovery, we brought together industry experts—Robert Keeling, Partner at Redgrave; Megan Ferraro, Director and Associate General Counsel at Meta; and Karl Sobylak, Senior Director of Product Management at Lighthouse—to unpack the findings of our second annual report, and more importantly, to talk candidly about how AI is actually being used in eDiscovery today.

What we heard confirms what many in the space are already seeing: skepticism is down, experimentation is up, and the real-world impact of AI is starting to take shape.
‍

Less hype, more action

Just a year ago, legal teams were still asking, “Is this real?” Now, the conversation has shifted to, “How can we use this?” Megan shared that the dialogue between in-house teams and service providers has evolved—clients are no longer just listening; they’re shaping the direction of AI tools. Meanwhile, Robert pointed out that attorneys are starting to adopt AI into their day-to-day workflows in very targeted, meaningful ways.

The fear that generative AI would upend legal jobs is cooling. Instead, overall sentiment toward AI has become more positive, with a 38% increase in our survey respondents who believe AI has a net positive impact on the legal industry. Teams are now recognizing the opportunity: automating the repetitive so attorneys can focus on the strategic.
‍

What’s driving adoption

According to our survey, the top motivators for adopting AI are team productivity (67%) and freeing attorneys to focus on high-value work (56%). That tracks with what we’re hearing across the board. Megan emphasized the need for more consistency in document review—something AI can deliver at scale. Robert highlighted use cases like Microsoft Copilot, which his team is now using to boost internal project management. Like Robert’s team, many others are taking advantage of enterprise AI tools: Our survey found that there has been a 95% increase in the number of respondents reporting they work for a company that has implemented an enterprise AI solution such as Microsoft Copilot.

Karl explained that as we find success with AI there is a “flywheel of wanting more of it.” As legal teams begin to see tangible results from AI—faster turnaround, less cost, more accuracy—they’re asking what else is possible.
‍

Accelerating experimentation and evaluation

One of the big takeaways from the conversation was that the best AI solutions often start with a simple question: “What if we tried applying AI to this specific challenge?”

Robert stressed the value of experimentation. His team pilots GenAI on samples to benchmark performance and ROI before rolling out more broadly. Megan noted that traditional three- to five-year tech evaluation cycles no longer work in a world where tools evolve monthly. Now, they’re gathering feedback and evaluating tools on a six-month cadence.

Our survey respondents also expressed an increased interest in evaluating the ROI of AI. While team and attorney productivity remain the leading motivation for AI use, recognition of AI’s positive impact on ROI of legal spend increased significantly year over year.
‍

Trust and transparency still rule

While enthusiasm is growing, trust and transparency are still top of mind. Data security and output reliability remain core concerns—they were the two top areas of perceived risk related to AI in our survey—especially as AI moves from back-office support to client-facing tasks. Megan and Robert both agreed: Vendors need to lead with transparency and proactively address information security concerns.

And while courts are beginning to acknowledge AI’s role, Robert noted that we still need more judicial validation of AI-assisted review to cement its defensibility.

Karl noted, however, that there is a good precedent for using technology in eDiscovery and ensuring quality results (with non-AI analytics, for example), and that there are opportunities to augment existing technology and workflows with AI.
‍

The road ahead: specialization and scale

Looking forward, the panel sees the future of AI in legal as highly specialized. Instead of one AI to rule them all, we’ll see task-specific models: one for summarizing case files, another for categorizing privilege, another for translations. And that’s a good thing.

As Karl pointed out, the combination of targeted models, workflow alignment, and transparent measurement will define the next phase of AI maturity in legal.
‍

Final thought

AI isn’t a silver bullet. But it’s no longer a question of “if” AI will be used in eDiscovery. It’s a question of “how” legal teams will harness it to drive better outcomes—faster, smarter, and more defensibly.

Want to dig deeper into the how we’re thinking about and using AI? Visit the Lighthouse AI page.