In the News

Gen AI’s Place in E-discovery is Far From Certain: A Chat With Legalweek Speaker Robert Keeling

| Legaltech News

In a recent Legaltech News article, “Gen AI’s Place in E-discovery is Far From Certain: A Chat With Legalweek Speaker Robert Keeling,” Robert Keeling examines the varying reception of generative artificial intelligence (GenAI) compared to Technology-Assisted Review (TAR) within the eDiscovery field and the potential challenges to its adoption.

Gen AI’s Place in E-discovery is Far From Certain: A Chat With Legalweek Speaker Robert Keeling

By: Rhys Dipshan

When it comes to AI adoption by the eDiscovery community, what’s past isn’t always prologue.  While technology assisted review (TAR), which was essentially machine learning, received a frosty reception by eDiscovery practitioners, GenAI is more readily welcomed. But widespread adoption isn’t guaranteed, and all depends on if the technology can overcome potentially significant roadblocks.

Legaltech News Caught up with Redgrave LLP Partner Robert Keeling to discuss why GenAI is finding a receptive user base in the eDiscovery world, what challenges to adoption could lie ahead, and what needs to change for the technology to fundamentally transform the Electronic Discovery Reference Model (EDRM) process.

Keeling will speak at the “Navigating the AI Revolution: Strategic Insights and Innovations” session on Wednesday, March 26 at Legalweek 2025 in New York.

This conversation has been edited for length and clarity.

You lead Redgrave’s Second Request practice, which helps clients handle Hart-Scott-Rodino (HSR) requests from the DOJ and FTC seeking more information on potential mergers.  Given the cuts to federal agencies, how do you believe your practice will change in the short and long term?

So from my perspective, we are not seeing any significant changes at least in the short term, and that’s partly because the extent we’re working on antitrust litigations, those litigations were already filed, and same with antitrust investigations that they were already ongoing.

I think the question is whether we will see a decreased number of second requests or decreased number of antitrust litigations being brought by the government, at least in the short term.  I think that is still to be determined.

But overall, the government, regardless of the administrations, has been taking an increased emphasis and focus on antitrust enforcement, and from kind of a medium and long term perspective, I suspect that overall, we will not see significant changes in the overall level of enforcement by the U.S. regulatory agencies.

Do you believe gen AI is a technology that eDiscovery attorneys will welcome, unlike the industry’s experience with TAR?

I believe that we will see a significantly faster adoption of the use of GenAI in eDiscovery compared to TAR.  I think there’s a number of reasons for that, but the biggest reason is that TAR unfortunately suffered from this perception that it was a so called black box, and it was very difficult for practitioners, for lawyers, for judges, to understand how it was working and how it was kind of making determinations on documents or categories of documents.

The advantage with GenAI are several. One, people are using it outside of their daily practice.  They’re using GenAI for fun … Their kids are using it for homework. ... they’re seeing it, not just in specific cases, but in their normal day to day, like if they run a Google search. … it’s becoming more part of what is accessible to people outside of their practice. … so I think you will see there’s less of a suspicion or barriers to entry of using GenAI.

The other thing that we’re seeing is that with prompts, because you’re using language and words and paragraphs and giving instructions, it’s more similar to how you would maybe instruct a review team, which people have a lot of experience with, as opposed to TAR …

A lot has been said about GenAI’s potential in eDiscovery.  What do you think the impact of the technology will be in the eDiscovery process?

I think short term we’re already seeing it be very effective on things like summarizing deposition transcripts … or looking at expert testimony or past expert testimony, same with looking at smaller sets of documents, and categorizing them by issue, and those have kind of immediate benefits…

Where we’re still having challenges is document review and privilege review at scale, really because of a pricing issue, the cost.  I say that while acknowledging that the cost has come down tremendously over the past 18 months, and so the price is right now at a level where I think it is useful and accessible…

As the price continues to decrease, we will see an even greater adoption of the use of GenAI for kind of larger scale applications like responsiveness review or privilege review.

What is your biggest concern with the use of GenAI in eDiscovery, or perhaps one that you think isn’t getting enough attention?

One concern is, because it works pretty well, there is a risk for people to rely only on the GenAI search and not perhaps other searches to either identify [responsive] information or privileged information, because [while] it tends to work well, it’s not perfect.

The other is, by far the greatest risk to adoption for GenAI—if requesting parties ask for the prompt, or if courts order the prompt, because that is very much work product.  And if there are court decisions that order the turning over of the prompts that were used for determining responsiveness, I think we will see then people stop using GenAI because of a fear of turning over work product.

Where do you see the best opportunities for emerging technologies, including GenAI, in the eDiscovery field?

So I think the biggest opportunity and change may be the introduction of either GenAI or [large language models] into the back end of Microsoft Purview or Google’s enterprise application suite.

For example, the use of Copilot and bringing it into Purview, and that would give potentially the ability to run queries on the company systems regardless of custodian and to be able to look across larger numbers of people to find the documents most relevant to the particular suit.  Then [we could] only export those documents, as opposed to exporting large mailboxes … That is the promise to dramatically cut costs, and [it] will have an impact on every kind of aspect of the EDRM.

What do you hope attendees take away from your Legalweek panel?

Our panel is on AI strategic insights and innovations.  [What] we’re going to be trying to do is really focus on practical advice and what me and the other panelists are doing with AI, as opposed to maybe talk about more theoretical applications and uses.  Instead, we’re really going to try and provide insights as to how GenAI can be used, and AI generally can be used, on matters right now, and hopefully the audience will be able to get some really good takeaways from that.

 

Related people

Explore Our Team