Legal work, well known for its long hours and heavy workloads, is being revolutionized by artificial intelligence — specifically generative AI, which is one of the most transformative technological advancements in recent history.
A 2023 study by researchers at Princeton University, the University of Pennsylvania and New York University found that “legal services” is among the industries most exposed to occupational change from generative AI. Another report, published in 2023 by economists at Goldman Sachs, estimated that 44 percent of legal work could be automated by emerging AI tools.
AI tools can now handle the endless reading, summarizing, creating and filing of documents ordinarily relegated to junior attorneys or paralegals. And even if it doesn’t replace lawyers entirely, the emergence of generative AI stands to radically alter the nature of practicing law, as well as its business model.
“It’s going to have an impact. We’re already starting to see that,” Pablo Arredondo, co-founder and chief innovation officer at legal AI company Casetext, told Built In. “How exactly it plays out remains to be seen. There’s a lot of choices we all have to make as a profession that’s going to help dictate that.”
Legal work has been identified as a ripe target for AI disruption many times before, with warnings of artificial intelligence ousting lawyers dating as far back as 2011. But over the years, this technology has served more as an aide to lawyers rather than a replacement, helping them quickly and efficiently identify, sort and classify words in lengthy documents.
The latest generative AI craze, with ChatGPT at the forefront, is stirring up these conversations once again. Large language models, or the underlying system of platforms like ChatGPT, are now capable of reading, analyzing and summarizing large swaths of text — automating the bulk of what lawyers, paralegals and other professionals in the industry ordinarily spend their time doing.
GPT-4, the large language model serving as the backbone of ChatGPT and several other generative AI systems, managed to score in the 90th percentile on the Uniform Bar Exam. This pushes the boundaries on not only what AI is capable of doing on its own, but what it can enable in the legal world.
“Passing the bar exam is a great illustration of some of [GPT-4’s] power and nuance, especially the fact that it could generate essays. It was sort of a vivid illustration of the progress,” Arredondo, who co-authored the GPT-4 study, said. “What really matters is attorneys can use this in their day-to-day work and get better outcomes for their clients.”
Law-focused AI software like Casetext, Latch, Harvey and others have taken GPT-4 and other language models and fine-tuned them for their specific kind of legal work. They can help lawyers research relevant case law, statutes, regulations and legal opinions, sifting through thousands of pages in a matter of minutes to not only find documents, but summarize them and highlight important passages.
AI tools can also help with contract analysis (a fundamental part of the legal profession), where they review contracts and provide insights into potential risks or compliance issues. Some can even generate legal documents themselves, a process in which a lawyer inputs specific requirements or key details into the tool and the system generates a preliminary draft, which can then be reviewed and refined by human professionals.
“Generative AI deals in the stock and trade of lawyers, which is words,” Ben Allgrove, an attorney and chief innovation officer at international law firm Baker McKenzie, told Built In. He predicts AI will cause a “reshuffling” of the legal industry, where the use of generative AI will be an expectation, not an experiment.
“I think it would be negligent or unprofessional for any lawyer not to investigate how they can use technologies and advances to deliver a better product,” he added, “in the same way that if a lawyer didn’t use an electronic database to search for cases would be negligent.”
In some instances, this investigation has already begun. Several prominent law firms, as well as legal departments within major companies, are putting this technology to the test.
In February 2023, international law firm Allen & Overy entered into an exclusive partnership with a tool called Harvey, which provides thousands of the firm’s attorneys with assistance in legal research, drafting documents and contract analysis. LexisNexis, a long-time data and analytics provider in the legal industry, announced in May 2023 that it created a new generative AI platform using GPT technology and teamed up with some of the biggest law firms in the United States, including Baker McKenzie, Foley & Lardner and Reed Smith. And several Fortune 50 companies, like Microsoft and Ford Motor, as well as top law firms like DLA Piper and Kirkland & Ellis, have tested a tool created by Casetext called CoCounsel.
This software isn’t doing anything that attorneys or paralegals can’t do themselves. It’s just able to do it faster and more efficiently than any human is physically capable of doing it — posing both time and cost-saving potential.
“There has been a lot of waste in law, and there’s a lot of room for efficiency,” José Padilla, a corporate attorney and founder of legal AI company LegalMente, told Built In. “AI will definitely help in that.”
Despite AI’s offerings, it is unlikely that these tools will outright replace human attorneys — at least not right now. Providing legal advice is more than just predicting words, which is all generative AI is capable of at the moment. And beyond all the reading and rote memorization, practicing law well requires a lot of uniquely human skills.
“Being a lawyer is more than just fulfilling a task,” Allgrove said. “It’s not the law that gets the deal across the line, it’s the negotiating skills. It’s the ability to read the person on the other side of the table.” As of today, AI can’t do that.
A much more likely outcome of this generative AI revolution is a change in the business of law — namely, the billable hour, which remains the dominant business model in legal work.
“Ultimately, it breaks the billable-hour model,” Allgrove said. “Once you get into a world where your law firms are needing to invest larger and larger in the technology to provide the product at the end of the day — not the time — then we’re going to have to find different models which enable us to get a return on our capital and our effort that is fair and that’s not based on charging for time.”
That being said, if AI was to do away with anyone, it would likely be the people working on what Hamid Kohan, CEO of legal AI company Legal Soft, calls “transactional law.” That includes tax resolution, immigration and estate planning — the sort of stuff that doesn’t require a lot of litigation. It mainly involves finding correct documents and filing paperwork, which a machine can do now for a much lower price than a human attorney.
It could also replace the “junior attorney drudgery,” Padilla said, like reading and summarizing thousands of pages of text. But these tools will likely always require a human eye.
“I think there’s always a role for a lawyer,” Padilla added. “Legal expertise is more than just what AI can offer, which is just kind of brute force legal knowledge. There needs to be expertise on top of that knowledge.”
Maybe AI won’t replace all lawyers outright. But can it still provide sound legal advice? At the moment, the answer is “not really.”
“It can approximate, and it can be very convincing,” Allgrove said. “But I have failed to see any system which can accurately apply law to a fact pattern in order to provide legal advice.”
In practice, giving reliable and accurate legal advice requires a lot of knowledge of the law, but also a nuanced understanding of the context in which the law exists, Padilla said. “As attorneys, we’re regulating human behavior. Because of that, human behavior and human interaction is a lot of what we’re about. Sometimes the legal answer is not the right answer, sometimes there’s a way to solve the problem of human interaction without the cold hand of the law.”
AI doesn’t understand that nuance, and could draw conclusions that may not produce the best result for a client.
“In law, there are a lot of gray lines. And you’ve got to make sure that the AI understands all the gray areas of the law. …It does not,” Kohan, who’s also author of How to Scale Your Stupid AI Law Firm, said. “It makes mistakes, and when you’re dealing with human lives like that you can’t make mistakes.”
Indeed, the propensity of AI to make up information — a phenomenon experts call “hallucinating” — is perhaps this technology’s biggest hurdle towards full adoption in the legal industry, a profession that hinges on finding and weighing facts.
This was made clear in 2023, when a New York attorney got in hot water for using ChatGPT to do legal research for a personal injury case. He crafted a 10-page brief, citing more than half a dozen previous court decisions. All of them turned out to be completely fabricated by the chatbot. The attorney was sanctioned by a federal judge and fined $5,000, along with a colleague.
The incident, which arose out of an otherwise obscure, run-of-the-mill lawsuit, captured the attention of both the tech world and the legal world — serving as a sort of punchline and cautionary tale of the dangers of artificial intelligence. Since then, a handful of federal judges have issued orders requiring lawyers to disclose when they’ve used generative AI to create legal documents.
Still, law firms and legal departments around the world are continuing to take advantage of a growing number of software to keep up with the demands of case management, research and the drafting of legal documents.
There are lots of legal AI tools on the market today. Some are meant to help attorneys streamline their research tasks and draft documents, others are designed to cut out the need for attorneys altogether. And many of them have been growing fast over the last couple years.
Harvey uses a combination of data analytics, machine learning and natural language processing to automate various aspects of practicing law, including contract analysis, regulatory compliance, due diligence and litigation. The platform is built on top of GPT-4, OpenAI’s large language model, which was then enhanced for legal work. In fact, OpenAI’s Startup Fund was among several big name investors to back Harvey with a $21 million Series A round in April 2023.
Founded in 2022, Harvey has already made inroads at a couple of major firms. The company entered a “strategic alliance” with accounting giant PwC in March 2023, and entered an exclusive launch partnership with international law firm Allen & Overy in February 2023. A&O is reportedly the seventh largest law firm in the world, and has made Harvey available to some 3,500 of its lawyers across 43 offices. The technology promises to be a “game-changer” for not just the firm, but for the entire legal industry, according to David Wakeling, head of A&O’s markets innovation group.
DoNotPay first launched in 2015 as “the world’s first robot lawyer,” initially generating the text users need to file complaints and cancel subscriptions. Over the years, the company has expanded its services, using artificial intelligence to help people register a trademark, file a police report and even sue people. In 2022, it also rolled out a chatbot that can help users negotiate bills using GPT-3 — the same large language model first used by OpenAI’s ChatGPT — engaging in back-and-forth conversations with various companies through live, text-based chats.
In 2023, DoNotPay made headlines for its plan to use its chatbot to help a person fight a speeding ticket directly in a physical courtroom. According to the company’s CEO and co-founder Joshua Browder, the intent was to have the AI “listen” to the case and then generate responses using GPT-3. The client would then hear those responses through a pair of AirPods and repeat them verbatim to the presiding judge.
The scheme received a lot of buzz — and a lot of flak, including from state bar prosecutors, Browder said in a tweet, which prompted him to “postpone” the court case amid alleged threats of jail time. The company is also facing a proposed class action lawsuit that claims the site offers “substandard legal documents” and provides “unauthorized legal services.”
Latch uses GPT-4 to simplify the contact review and redlining process for lawyers. It offers case management, time tracking, billing, and client communication tools, too. The software is a Microsoft Word add-in, meaning lawyers can use Latch directly from within Microsoft Word. And it integrates with various cloud storage services, making it easy for law firms to centralize their data and collaborate with each other on cases.
Latch launched in April 2023, and in just one month the startup had a wait list of more than 80 companies, co-founder Min-Kyu Jung told the Wall Street Journal. Now companies can get partial access to it within a day of signing an agreement with the startup.