Generative AI is reshaping legal practice, with lawyers and other legal professionals increasingly reporting time savings and productivity gains. Tools like Everlaw^AI Assistant are helping legal teams approach litigation and other data-intensive matters in vastly more efficient ways. Everlaw Founder and CEO AJ Shankar recently addressed advancements in GenAI capabilities at an “Ask Me Anything” community roundtable, where he provided his perspective on the state of the legal industry and what the future holds for ediscovery.

In the AMA discussion, AJ answered member questions about how document review may evolve and what professionals need to consider for responsible use of GenAI. Attendees also got a preview of upcoming Everlaw features, including the new Everlaw^AI Deep Dive tool (formerly known as Project Query), available in open beta since Aug. 12, which helps users quickly analyze large document sets by asking questions in plain language and getting answers with citations to easily verify information.

[Ed. Note: AMAs, community roundtables, and other events are exclusive to members of the Everlaw Ediscovery Community. To join, sign up here. The following highlights have been edited for clarity and brevity.]

What are the top development priorities for Everlaw in the next couple of years? 

Number one is to ensure that our core discovery workflows remain at the very cutting edge. By that I mean the ones that help you ingest, review, and produce data. Number two is to continue to invest in other, adjacent product areas key for litigation and investigations success. For example, we’re making a ton of improvements to our narrative-building feature, Storybuilder, over the next couple of years, as well as legal holds, and collections from modern data sources. 

Of course, we need to ensure that we are embedding GenAI features wherever possible in these workflows to give you the tools that you need to succeed. We want GenAI improvements to be available to you so you can figure out the right usage pattern for you and your risk tolerance, your client needs, and so on. That means that we’ve got to have all the best tech out there.

What features are you most excited about in the next few months? 

I’m very excited that we’ll make Deep Dive, formerly known as Project Query, available. This new tool gives you the ability to ask questions of your entire corpus and get answers with citations. Our closed beta users have told us this is a game-changer for litigators. I’m also really excited about the file-view feature in DataViz of your file path — it’s one of those small quality-of-life improvements that we want to keep ensuring we don’t forget about. We have a host more coming out in terms of refinements to user experience that we know will make your life better. 

We also have a bunch of work on our chat experience like launching a fully native chat viewer that allows you to filter down by user, to redact individual messages, to have more control over segmentation, to produce with more granularity and rigor, to search by individual message. All that is going to make Everlaw the number-one tool for handling your chat data. 

Alongside that is a number of entity improvements. We’re continuing to consolidate our viewpoint of the people in the case, so that you can understand individuals across all their communication channels, whether email, chat, or text. We already now can automatically identify all the emails that a person alias as a person might be using that’ll help you search for them better. 

We’re going to add in phone numbers and chat IDs.

What are the next key generative AI features or enhancements planned for Everlaw^AI Assistant?

We’re really excited to continue driving adoption and to build upon the core functionality of the platform. 

For features like Coding Suggestions, we have a number of improvements to make it easier to set up and configure your coding suggestions and run them. 

Our extractions tool, which probably not enough people know about, is insanely powerful. 

We’re boosting that up to allow you to extract things like numbers and dates, and sort them and edit them. 

That will let you pull out regular information from a whole set of documents, such as all the email addresses or Social Security numbers or clauses of a particular type, or titles of a bunch of documents. Beyond that we’ll have more ways to interpret the data you’ve extracted.

We’re also working on Storybuilder improvements, making our drafts more first-class-AI oriented. 

How can misuse of AI become a professional liability?

Generally speaking, I would say that use of AI, like any other tool, should ultimately conform to all the ethical obligations and the legal obligations you might have. 

Ultimately, whatever the AI does, we all, as humans, are responsible for the output in the work. We should never be turning in an AI’s work without making sure that we put our stamp of approval on it. When these things go wrong, it’s when people haven’t done that.

In our legal system we are responsible for these outputs. There is a real risk in us blindly following these tools. 

That’s why we, at Everlaw as a company, have been very intentional about the functionality we have laid out so that we make it very easy for you to verify the work. We really encourage you to check your work where appropriate, and we want you to develop your own risk tolerance and process for using the tool so that you are comfortable with the outcomes. 

As a futurist, what do you think document review in the US could look like in three to five years?

There’s a big range which is dependent on the rate of improvement of frontier AI model functionality. Starting more conservatively, there’s a world where Everlaw looks much the way it does today, where we just have really, really good tooling for every step of your process. 

The second, more aggressive world view is one where you’re merely orchestrating the process, and we have an AI-accessible platform where you orchestrate the ingestion of data, then the whole system gets automated and processes your data, largely driven by AI — with certain human components to do things like QA checking. At that point, you’re really an orchestrator, and we’re relying on your expertise and your judgment to ensure that the outputs are good.

The most aggressive viewpoint is one where this is approaching singularity-type stuff, where one corporation sues the other, and they would never trust any humans to litigate – because we’re not smart enough – and you'd have AI agents that are all fully running it, getting data in and out of Everlaw and doing the review and all the analysis they want to do. They’re better at negotiating than a human, so you’re going to not rely on humans to negotiate.

Ultimately, it’s going to come down to model capabilities and the risk tolerances of people. I think there’s a middle ground where we have clear human orchestration and oversight where humans are still very much in the loop for critical decisions, and are ultimately a part of the process of getting to justice.

How is Everlaw ensuring that the insights of litigation support professionals who drive adoption and manage real workflows are helping to shape product innovation?

We focus on getting a lot of coverage of use cases, not just of the super high-powered attorneys whom we want to serve, but all the folks that are ensuring that the nuts and bolts in the machinery work well every day. We’re really intentional about this; It’s one of our core company values to have respect and understanding of all our users, including the non-attorneys.

We try to work really closely with all our users — all the ones that are excited and interested in giving us feedback. How do we actually get this information? There's avenues like the community here; we log all the comments and suggestions coming in. We also will work with the key contacts in every one of our clients; we go out and solicit your feedback and advice and consolidate the user’s pain points.

We also track support interaction. So if you’re a lead support person and you’re asking questions of our support team that we think could benefit from making a feature request, we’ll log that and track who’s asking for it. 

We also have a survey — with one coming out pretty soon — where we ask very open-ended questions of everybody.

And then of course for specific features we’re working on, we interview many users and often have a beta-testing cohort to help refine the experience.

Petra Pasternak is a writer and editor focused on the ways that technology makes the work of legal professionals better and more productive. Before Everlaw, Petra covered the business of law as a reporter for ALM and worked for two Am Law 100 firms.