Reimagining Workflows for Generative AI
Highlights from the Everlaw Summit '25 Keynote
by Casey Sullivan
AI is everywhere. From the sparkle in your emails to the video in your social media feed to the results of your Google search, AI can feel inescapable. In many cities, the San Francisco Bay Area especially, you can’t complete your commute without an AI inundation.
“I can understand there’s some fatigue,” Everlaw Founder and CEO AJ Shankar acknowledged as he kicked off his keynote address to more than 500 legal professionals at this year’s Everlaw Summit. But there is a good reason for the attention paid to AI. “The technology continues to advance apace, with new capabilities unlocked every year,” AJ explained. “It’s transforming the practice of law faster than any prior technology.”
So, yes, this keynote would be another AI keynote, AJ said. And for good reason: With testimonials from customers driving genuine impact with generative AI, the announcement of the forthcoming public release of EverlawAI Deep Dive, and updates to AI pricing to make usage seamless and costs predictable, AJ’s keynote showed how AI is already changing the shape of litigation today.
Of course, “We know that each of your AI journeys is different,” AJ elaborated. Everlaw’s goal, he promised, is to support you wherever you are on that journey. “The reason I talk so much about AI is not to shill it, or to coerce you to use it. It’s to educate you so you can make an informed decision about what’s best for you and your organization.”
“We spend time thinking deeply about how new technologies like GenAI are going to impact the core workflows and goals of litigation and investigations,” AJ said. “I hope we can inspire you to rethink your own workflows to incorporate AI tools where they make sense for your journey.”
Watch AJ’s full keynote below and read on for more highlights.
Play this video on Vimeo
Integrating AI Throughout the Everlaw Experience
As GenAI continues to improve rapidly, it is already helping legal teams automate tasks like document translations, summaries and sentiment analysis, and even do first-pass review on thousands of documents in a single go.
For maximum impact, legal teams need to understand not only how AI works, but also how it fits into their litigation and investigations workflows.
As always, “Our commitment is to ensure you have access to the best technology, no matter what it may be – AI or not, big or small – and that you’re fully trained and enabled by this tech,” AJ shared. That is accomplished with a tool where all code is developed in-house, where every experience is built by a research and development team of nearly 200 people, and where “we’ve never offshored our engineering and we’ve never acquired another company,” AJ noted.
It’s a team that thinks deeply about how GenAI will affect both the goals of litigation and investigations and how that work gets done. “We’re not adding AI functionality opportunistically, to capture news cycles or do big fundraises,” AJ explained.
“I can confidently say that we’re adding new AI features because they work.”
That means AI-assisted experiences are increasingly embedded in day-to-day workflows.
“Eventually, ignoring AI is going to feel like ignoring search. It’ll be an essential technology,” AJ said to the packed ballroom of legal pros from law firms, corporations, and the public sector gathered at San Francisco’s Palace Hotel. “Our job is to ensure that you always have access to the best technology and that you’re fully trained and enabled.”
A Responsible Approach to AI Development
To deliver great functionality that you can trust requires a strong commitment to responsible development, AJ noted. That means always protecting your data, ensuring that all the data that goes to model providers is never used to train their models and is never retained after the query has been answered.
It means minimizing the risk of hallucinations by avoiding the built-in knowledge in LLMs, “a huge source of hallucinations in the legal field,” AJ explained, and instead focusing on “the four corners of your documents, providing the AI with ground truth at the time we ask it a question.”
It means a clean and logical integration into current workflows, with precise use cases developed through robust testing, and the necessary context to inform the LLM, “which means that using our AI is often as simple as clicking a button,” AJ said.
It means ensuring that validation is quick and easy. “We understand that any work product is ultimately yours to own,” AJ added. “So we make it easy for you to validate the AI’s output, whether by providing Bates-numbered citations or direct quotations from the source material.”
Finally, it means taking the time to do it right. “We are much more interested in being the best than being the first,” AJ pledged, explaining that “we typically beta test our AI features, gathering thousands of items of feedback over many months, and iterating on the experience before putting them in the wild.” The “balance of prudence and pragmatism” is designed to give users the best AI functionality, and the confidence that these tools can be relied upon.
To show how legal teams are doing just that, AJ invited representatives from three Everlaw customers – Benesch, GEICO, and Vorys – onto the stage to share how they reimagine their workflows through the combination of GenAI-powered tools and core Everlaw features.
Benesch: Collaborating on Coding Suggestions in Storybuilder
EverlawAI Assistant’s Coding Suggestions, which uses the reasoning capabilities of large language models to expedite the review determinations process, helps surface relevant documents based on prior review decisions. The GenAI feature has shown it can perform at accuracy levels that match or exceed those of human reviewers, significantly speeding up the manual review process.
Legal teams at Am Law 200 firm Benesch have used Coding Suggestions for first-pass document review, to understand productions from opposing counsel, and for privilege review, said the firm’s Director of Litigation Support Steven Delaney. To date, Coding Suggestions has performed with more consistency and at a higher speed than first-level human reviewers, he said.
Play this video on Vimeo
The key to success is working closely with the attorneys who know the case and can refine the prompts that Coding Suggestions needs to pinpoint the correct information.
The case teams “need to help define Coding Suggestion prompts for each issue,” Delaney explained. But “the great news is that they just need to review five to 10 documents per issue.” That iterative process, focused on a smaller number of documents, “helps you to refine much more quickly,” Delaney said.
The Benesch litigation support team collaborates with attorneys in Storybuilder Drafts on the issues and prompt revisions, and to document how prompt choices were made, along with the results they yielded – all with version control.
“This is the key time to be taking advantage of this. Right now, you can get ahead of the opposing side by using this technology at this time.”- Steven Delaney, Benesch
“We love that we can have different attorneys working on different prompts concurrently in Storybuilder,” Delaney said. “It is great to have it all in one place for the team.”
Delaney ended with a reminder for the audience: “This is the key time to be taking advantage of this. Not everyone is using it right now.”
“If you want to get a competitive advantage, you don’t get a competitive advantage by doing what everyone else is doing. The best you can get then is parity,” Delaney explained.
“Right now, you can actually get ahead of the opposing side by using this technology at this time.”
GEICO: Combining GenAI and Traditional AI
Ed Valio, Director, Ediscovery and Records Management at leading insurer GEICO, joined AJ on stage to talk about a unique, non-litigation use case for Everlaw that his team devised to assist the organization’s contracts group.
To find an answer to an “unusual question,” they needed to evaluate tens of thousands of contracts on a very short timeline. “Having the full end-to-end workflow in Everlaw, with AI in the right places at the right time, is really what made this process so simple to execute and allowed us to move so quickly,” Valio explained.
Play this video on Vimeo
After uploading all 50,000 contracts into Everlaw, Valio began by leveraging custom extractions, an EverlawAI feature that allows you to quickly, and in bulk, pull relevant information such as contract signatories or notice periods, from your document set. Valio paired that information with high-value search terms to first identify relevant information, then used Coding Suggestions to evaluate potential relevance.
"Don’t keep yourself in a silo. The more we apply these capabilities outside of ediscovery, the more strategic and indispensable we can become.” - Ed Valio, GEICO
The team used that output to seed a Predictive Coding model to score the full contract set to help find any contracts that might also fit the profile they were looking for, but didn’t contain the high-value search terms.
These highly scored documents and the sample set were delivered to colleagues on the contracts team.
Valio’s team was able to turn around the entire project in 48 hours and provide a definitive answer to the business.
It turned out that just one document out of the 50,000 contracts had the exact characteristics they were looking for. “That allowed us to understand the situation very quickly and clearly across that whole set,” Valio said.
Valio closed by emphasizing the strategic value of these new workflows across functions. “Don’t keep yourself in a silo. The tools and the skills that we as ‘ediscovery people’ use every day aren’t just valuable for litigation and investigations, they can solve high-value business problems across an organization,” Valio explained. “The more we apply these capabilities outside of ediscovery, the more strategic and indispensable we can become.”
Vorys: Using Deep Dive Across the Litigation Lifecycle
EverlawAI Deep Dive received enthusiastic endorsements from customers in the open beta program. The tool lets case teams, including partners and other senior attorneys, engage with large data sets directly by asking questions, just as they would with a colleague, to quickly surface verifiable insights.
AJ welcomed Julie K. Brown, Director of Practice Technology at Vorys, on stage to share her experience. Brown, who leads a highly skilled legal technologist team at the Am Law 200 firm, said Deep Dive wowed the attorneys during the beta pilot. After a successful initial test, the firm has used Deep Dive at various points in a case lifecycle, including for investigations, quality control, and deposition and trial prep.
Play this video on Vimeo
Early on in one investigation, the Vorys team applied Deep Dive to figure out who the key individuals were with knowledge about a non-compete agreement and received a detailed list of not only the people, but also explanations of how they relate to it.
Vorys also used Deep Dive for quality control in a case involving 2 million documents and a one-week production deadline. To ensure defensibility, the team used traditional search and Predictive Coding to move quickly through the data set to production. Deep Dive was an effective validation tool, she said, letting attorneys ask about information they may have missed.
“The attorneys were just in awe when they saw the results.” - Julie Brown, Vorys
For depositions and trial prep, the team was able to use Deep Dive to identify key documents and any potential gaps by asking, for example, what a deponent knew about a topic, if there were discrepancies between documents, and who else may have known about the topic.
Demonstrating the value of Deep Dive was easy. In their initial test of Deep Dive, Brown explained, “The attorneys were just in awe when they saw the results.”
“The answers were spot on,” Brown said. “Based on just that one test, the word spread. If attorneys like something, they’re going to go tell all their friends. Ever since that, we haven’t had a challenge getting additional cases [into Deep Dive]. In fact, most of the time attorneys are coming to us saying, ‘Can you add my case to Deep Dive?’”
Now, when she introduces Deep Dive to new teams, she begins with a simple question: “Are you ready to see some magic?”
Deep Dive is scheduled to be made available to the broad public in December 2025.
Getting More Value from EverlawAI
To make sure that all users can take advantage of AI tools in their workflows, AJ also announced key pricing changes designed to make EverlawAI more accessible than ever before.
“I’m happy to announce that we are now including Writing Assistant and single-document Review Assistant tasks with your core Everlaw subscription,” he said. “No, we are not forcing you to pay more for a new ‘standard’ GB rate,” AJ added. “Our rates are staying the same.”
With AI included, your reviewers can now summarize a long, complex document, make better coding determinations for individual docs, and complete topic analysis and extractions. Case teams can add documents to Storybuilder, have them automatically summarized or incorporated into a memo, an outline, a table of key facts, and more.
“The bottom line is that we think there is powerful functionality here in Storybuilder and Writing Assistant,” AJ concluded. “We want you to try it, and if it delivers, to start using it in every case. At no extra charge.”
AJ also announced that Coding Suggestions, Everlaw’s most popular batch action, will be reduced in price by more than 40% starting in October.
Purchasing AI will now be easier than before, AJ explained, with Staging Drive, ECA, active, suspend, and AI credits to use for batch translations, batch Coding Suggestions, or Deep Dive soon available in a single, unified contract.
“This isn’t a one-time promotion – it is real and permanent,” AJ said. “The change is consistent with our belief that, as your best technology partner, our job is to enable you to use the best cutting-edge technology out there.”
More New Tools – And the Future of Legal Workflows
Everlaw will also soon add two brand-new features to the EverlawAI toolbox: the Depositions Q&A tool and the Privilege Descriptions tool. With these tools, users can interrogate their deposition documents for comprehensive insights – or subtle inconsistencies – with synthesized answers citing back to source testimony for easy verification.
Meanwhile, Privilege Descriptions will take some of the pain out of creating extensive privilege logs and similar documentation, with AI-generated, grounded explanations of why a document is marked privileged.
With generative AI tools now woven throughout the Everlaw experience, “We think that soon enough,” AJ elaborated,” using these tools is going to be second nature to you. Your workflows will incorporate them, and they’ll improve the quality and speed of your work.”
Making sure those workflows are effective, repeatable, and defensible is essential, AJ said. “In some industries, it doesn’t matter how you get a result. But not our industry. You have obligations to your clients, to an opposing party, to the court. We need to ensure that when you do work, it stands up to inspection.”
That’s especially true, AJ continued, “in a future when AI agents are running rampant.”
“We think AI has a very important role to play in the future of ediscovery,” AJ acknowledged, “but it needs to be harnessed, and orchestrated, by humans.”
Meanwhile, in the present, many complex workflows are managed in spreadsheets or ad hoc.
“What if we could address both problems with one solution?”
That’s the impetus behind Everlaw Workflows, which will allow users to create and execute customized workflows. “With this tool,” AJ explained, “you can construct complex workflows and have your documents automatically flow through them, with reporting on each step.”
The steps of the workflow can be anything in Everlaw, from first-pass review to running search term reports to applying batch AI actions and more. You can add in AI where it makes sense, leveraging Coding Suggestions, automatic translations, custom extractions, and more.
Users will also be able to trigger workflows automatically, such as when new documents are added to a matter, and manually, adding in sampled quality control, conditional branching, and gating to ensure security and reliability.
“Imagine this future,” AJ posited. “Instead of getting in the guts of Everlaw, you’re orchestrating outcomes. Your colleagues can step in at exactly the right time to add value in a defensible, repeatable way. You can use AI to solve problems when the budget requires it or the client or stakeholder allows it. You can see exactly where each of your documents is at any time: which are moving along, which are stuck, which will require additional resources to hit your deadlines.”
“Much like I demonstrated Deep Dive on this stage last year,” AJ reminded the audience, “Everlaw Workflows are in their early stages. But success is built around partnership between us and you. If you would like to be a design partner on this, reach out to your customer success manager.”
And be sure to return for Everlaw Summit ‘26.
If you’re interested in learning more about generative AI, or want to see how EverlawAI Assistant can improve your workflows, request a demo today.
Casey Sullivan is an attorney and writer based out of San Francisco, where he leads Everlaw’s content team. His writing on ediscovery and litigation has been read by thousands and cited by federal courts.