We use AI every day. Or, perhaps more correctly, machines are learning more about us every day—we just know how to use it to our advantage. Netflix recommends movies based on movies we’ve watched and liked before. Spam filters let us ignore thousands of pointless emails. And of course, Amazon’s targeted advertisements—carefully curated to your buying history and demographic information, among other things—make purchases (and repeat purchases) easy.
In e-discovery and other areas of legal operations, that’s true as well. Many forms of legal technology utilize some degree of AI. Technology Assisted Review (TAR) and predictive coding have become standard tools of the trade, but are still under-utilized among legal teams. Exterro’s 2019 In-House Benchmarking Report found that only 9% of respondents thought that TAR helped save time and costs—shocking, considering review is the most expensive and time-consuming e-discovery stage. Another survey by the Corporate Legal Operations Consortium (CLOC) found that only 12% of teams used AI technology.
The motivation of investing in AI should be that it can execute legal tasks faster, more accurately, and more efficiently than humans. Deep learning and predictive capabilities that can cut down the time and costs associated with tasks like identifying responsive data among large troves are improving at a fantastic rate.
So how can attorneys and in-house legal teams begin to embrace AI use? The answer is to seek AI technologies that can help answer current and forward-looking legal and regulatory hurdles to create efficiency and save time even as the regulatory environment is changing. Namely:
- How do we automate mundane tasks?
- How do we get to the facts of a matter as quickly and defensibly as possible?
- How do we deal with personal data to ensure compliance?
In the Future, Lawyers Will Be Part Robot
Not literally, but for many attorneys and in-house professionals, technology should be lock-step with processes. This requires a deeper understand of specifically how the technology can save time and money, and therefore why it should be used. The thought of learning a new technology when a current process works just fine can be scary, but ultimately it should help individuals do their jobs better—and it may become a necessity as legal teams are asked to do more with less.
“You’re going to see attorneys using their own [software] more to get things done,” predicts Ronald Hedges, Senior Counsel for Dentons LLP and former U.S. Magistrate Judge for the District of New Jersey. “Technology is affecting the profession but the question is where exactly it’s going to hit.”
Considering that AI is capable of automating many of the tasks that once required teams of people, it may be expected that headcount is one of the areas that is most affected by acquiring AI.
“I expect we’re going to see attempts to use technology to reduce staff functions more than anything else,” Hedges added.
The concept of the robot lawyer has been around for quite some time. In fact, an MIT and UNC report, “Can Robots Be Lawyers? Computers, Lawyers, and the Practice of Law,” disputed many of the primary arguments that automation and AI would replace the work that attorneys and legal teams currently perform. They found that AI is most suited to perform functions that comprise only a small fraction of what lawyers do. Utilized properly, AI is best suited to increase productivity and create time savings, not reduce headcount.
And judges like Hon. Charmiane Claxton, U.S. Magistrate Judge for the Western District of Tennessee, don’t see “robot lawyers” as a reality.
“There have already been some attempts to use AI to assist in winnowing through e-discovery requests, but it’s kind of high-or-miss, so you still need a human to double-check what you’re looking at,” says Claxton. “So I don’t think we’re ever going to get to a point where it’s ever going to be a complete replacement in culling through and preparing e-discovery responses.
As comfort and awareness of these capabilities increases, AI can help more attorneys and legal teams automate the most mundane and time-consuming tasks to free up resources for more important work.
Predictive Coding—and New AI Tech—Will Lead to The Creation of New Standards
Predictive coding is a machine learning process that uses software to take keyword searches and logic, entered by people, for the purpose of finding responsive documents. It typically applies to much larger datasets to reduce the number of irrelevant and non-responsive documents that need to be reviewed manually.
But predictive coding doesn’t represent the most forward-looking technology: The ability to smart label during the review stage, for example, leverages the latest advancements in Deep Learning. By incorporating techniques of active learning, the system is enabled to learn from the reviewers as they review documents, and suggests labels for the rest of the non-reviewed documents without the need for dedicated human-machine training sessions.
These may be the latest emerging standards, but David Cohen, Head of E-Discovery for Reed Smith, says that going forward in this case means going backward a bit: recognizing the value of predictive coding, and establishing judicial standards.
“I remember 10 years ago when predictive coding was new in the market, people predicted it was going to replace human review,” says Cohen. “That certainly hasn’t happened, and one of the reasons it hasn’t happened is because it’s so complicated in terms of what standards apply to it—how are you going to get the other side to accept it? How are you going to get the court to accept it?”
Hon. Xavier Rodriguez, U.S. District Judge for the Western District of Texas, has a number in mind.
“We’re expecting 85% accurate for a lot of predictive coding,” says Rodriguez.
Which begs the question: Why isn’t there a standard yet? As AI continues to evolve, and more sophisticated concepts begin to permeate into e-discovery platforms, accuracy and reliability increase. Eventually, they’ll reach the point where they become commonplace in negotiations.
“It’s not just about the technology, it’s about coming up with standards and making everybody feel comfortable that the AI is accomplishing what it needs to accomplish,” says Cohen. “And of course in discovery, that means finding the stuff that is relevant, and getting rid of the stuff that isn’t relevant.”