Can Artificial Intelligence Solve Big Law's $60B Question?
ALM Media
computer brain
(Image: Shutterstock)
For law firms and their clients, the promises and the challenges of new technologies were both on display earlier this month at a Georgia State University College of Law symposium on turning artificial intelligence into legal intelligence.
Three lawyers who turned to careers in legal technology, Joe Tiano of Legal Decoder, Alison Grounds of Troutman Sanders’ eMerge subsidiary and Ed Walters of Fastcase, discussed AI's impact on billing, e-discovery, legal research and beyond. Firms and in-house lawyers realize that artificial intelligence shaping the future of legal practice, they said, even if they are embracing it with varying degrees of enthusiasm.
One sign of the shift, said Tiano, Legal Decoder’s CEO, is that he initially sold his software to companies to track and manage their legal bills—but now firms have started using it to improve their internal processes.
They want to track how their attorneys and staff are spending time, he said, so they can streamline their processes and provide better pricing to clients—ideally making legal services costs both more predictable and affordable.
Clients are generally happy with the lawyering from their outside counsel, but they are not happy with the pricing, said Tiano, who left the partnership at Pillsbury Winthrop Shaw Pittman five years ago to found Legal Decoder.
Alison Grounds (from left), Michelle Hook Dewey, Patrick Flanagan, Joe Tiano and moderator Natalie Kelly during a panel at GSU's symposium. (Courtesy photo: GSU College of Law)
In fact, Tiano considers “pricing uncertainty,” the single-biggest challenge facing the legal industry. According to his research, there is $60 billion worth of pricing uncertainty in Big Law—the difference, he said, between hours billed and realization rates.
Wachtell doesn’t have to compete on price, Tiano said, but Am Law firms in the middle of the pack need to differentiate themselves. “You don’t want to compete on price,” he said. “That’s a race to the bottom.”
But they can compete on price predictability.
When he was in private practice, Tiano said, clients would ask him how much an M&A transaction would cost. Between $450,000 and $950,000 for their type of deal, he would reply.
“They would look at me like I had seven heads,” he recalled, and ask: “Can’t you do better than a half-million-dollar swing?”
For law firms, billing software can “instill some efficiency in how legal professionals bill time,” Tiano said, by quickly crunching billing data for waste in time or activities.
The most obvious use for firms is tracking how long timekeepers take for standard tasks. Perhaps it should only take an associate one hour to handle a pro hac vice motion to appear in court, Tiano said. The software flags instances that took more time.
It can also track “workflow inefficiencies,” such as multiple instances of billing for the same repetitive task. That could be a case where a second-year associate is “churning the file because they want to hit 2,400 billable hours for a $75,000 bonus,” Tiano said.
Another issue is making sure line-item descriptions for tasks are “clear and concise,” he said. For example, if a lawyer bills 2.7 hours for something vague like “attention to file,” Tiano said, it’s not good for the firm or the client. The client will push back on that line-item—and so the vague description doesn’t help the firm collect for the lawyer’s time or understand how that time is being spent.
The Resistance
While clients continue to call for more efficiencies from their outside counsel, they too can resist adopting new software tools.
In order to test two different software tools for contract due diligence, Alison Grounds, who heads Troutman Sanders’ innovation initiative as well as its eMerge subsidiary, offered to do the due diligence for free to 20 clients. There were no takers, she said.
Grounds said the clients either couldn’t get the data together or were worried about what she’d do with it. One general counsel even asked her to fill out a 20-page survey on how she planned to use the data—then still said no.
Grounds pressed ahead with data she had on hand—and she will be presenting her findings at a Troutman CLE this summer.
Automated contract review is still new, Grounds said, and it makes up only a tiny percentage of the work her eMerge team does. That work is still overwhelmingly document review to support litigation for both Troutman and outside clients. But it’s where the practice of law is going, she said, adding that clients (and judges) were also resistant to ediscovery at first.
“Clients are getting pressure from their bosses to use technology,” Grounds said.
Ethical Duty to Use AI?
Walters, the CEO of Fastcase and an adjunct professor at Georgetown Law, raised the question of lawyers’ ethical duties to their clients in applying artificial intelligence to legal practice.
Lawyers have duties of competence and diligence to their clients, Walters said. “In a world where you can know the answer, it’s a problem if you don’t know the answer.”
If the information is available and the tools to analyze it exist at a reasonable cost—but lawyers don’t use the technology, he asked, is that an ethics violation?
When clients ask questions like how much a case is worth, how much it will cost and what the likelihood is of prevailing, Walters said, lawyers provide “gut-based answers” based on their prior experience, not “data-based answers.”
But that experience might be based on just 30 to 40 cases, instead of a comprehensive body of data, he said. “Social scientists call decision-making like this ‘hunches.'”
Walters was a young associate at Covington & Burling in 1998 when he was tasked with assisting in due diligence for the $74 billion ExxonMobil merger. He recalled watching an 18-wheeler filled with boxes of documents pull up to the firm’s loading dock—and then another one. It took hundreds of associates and paralegals 18 months to hand-review all the documents, he said.
Now, more than 20 years later, he said, if a client has a warehouse full of documents for discovery it’s reasonable to use ediscovery—and not 200 lawyers—to review them.
The same holds true for newer areas of legal analytics, such as contract review, and automated drafting tools for patent applications or discovery responses.
“The better the data gets, the less reasonable it is not to use it,” Walters said.
Walters left Covington & Burling almost 20 years ago to start Fastcase, an online law library for researching case law, statutes and regulations. It has gained more than 900,000 subscribers since its launch—in part because many state bar associations, including the State Bar of Georgia, offer it as a member benefit.
Fastcase has just bought Docket Alarm, which pulls data from federal case docket sheets via Pacer so users can track cases and outcomes. It allows clients to see how long matters take, who wins or loses and what firms are most effective, Walters said.
“The information is hiding in the docket sheets today,” Walters said. “AI can pull it out and aggregate it.”
What’s more, he said, firms have their own repositories of pleadings and billing data that can be used to help clients assess cost and risk for a matter.
AI brings its own risks, Walters acknowledged. “There are situations where AI will make decisions that people wouldn’t and get it wrong."
He has been thinking about the legal ramifications of self-driving cars, because he will be teaching a class on them at Georgetown Law next semester—due, he said, to student interest.
Autonomous cars might cut accidents in half by reducing human error, but they also introduce new accidents, because they don’t recognize road risks that a human would. They might prevent 40,000 deaths a year but cause 16,000 new deaths, Walters said.
In the same way, the algorithms that power data analytics and machine learning have their own limitations and biases in assessing enormous bodies of legal documents.
That means robots aren’t going to be replacing lawyers, Walters said, in answer to the ubiquitous question about AI applied to law. It’s still up to the humans to figure out how to design and deploy the tools.
