Artificial Intelligence Has Found a Home in Contract Management

Gabrielle Orum Hernandez, Legaltech News

As machine learning and artificial intelligence find traction within legal tech, they've found a fairly natural home in contract management work.

In many ways, contracts provide perfect fodder for innovation and expansion of artificial intelligence (AI) into the legal world. Contracts are often mechanical, highly organized and filled with boilerplate language. They can also be exceedingly long and unwieldy, taking huge amounts of time and resources for attorneys to parse the content of a contract and assess its merits or needs.

"The neat thing about contracts is that they're kind of neatly organized," says Bennett Borden, chief data scientist for Biddle & Reath. "The algorithm breaks those documents up into a bunch of constituent features."

More and more, attorneys at big law firms and legal departments are recognizing the need to adopt technology to help automate parts of their contract management systems, but what role will AI and predictive insight play in the contract management landscape?

How does it work?

Noory Bechor, co-founder and CEO of contract analysis tool LawGeex, was inspired to start his company by his own experience as a practicing attorney. He recalls being frustrated by how little legal thought he applied during his practicing days because so much of his time was consumed by tasks that could have been easily automated. "I kept banging my head against a wall for eight years. How can it be that my work is so manual?"

"When we started this company, we talked about how to solve this problem, and we didn't go right to AI," Bechor says. He notes that LawGeex talked about trying to apply advanced search techniques to contracts to help facilitate legal work, but landed on AI after realizing that machine learning could eliminate the need to search altogether.

"AI is really understanding the same way a lawyer's brain works," Bechor says. The technology can understand, "'What does this text actually mean?' and, 'What is the meaning behind this text?' This, for us, is where AI and machine learning has proven itself the right tool."

Bechor explains that AI insight in the company's tech comes in great part from its ability to compare documents to enormous databases of similar contracts to identify standard clauses and irregularities.

"This technology reads the contract, understands what it says, and reviews the contract in contrast of what it has in the database."

The second step, Drinker Biddle's Borden notes, involves parsing contracts down into sections, clauses and keywords. Contract management AI can then compare individual clauses in the contracts to hundreds of examples to identify different features within contracts.

"They actually have a library of examples of these provisions because they tend to come in types. They're able to highlight the differences but then kind of flag 'this is ususual' or 'this is a burdensome provision' or 'more strict than average,'" Borden says.

Moreover, the databases that AI can draw from are constantly expanding. "It's not about learning from the database of a single business; it's learning from the cumulative world of contracts out there and providing insight. There is a very strong crowdsourcing element built into the system," Bechor adds.

What can it do?

Larry Bridgesmith, professor at Vanderbilt University Law School, sees the beginnings of AI adoption in contract management as a window into future opportunities to improve this work.

"I'm beginning to be very encouraged by its potential state. What I'm seeing is the advancement of technology to be more systematized as opposed to a more ad hoc activity," Bridgesmith says.

Although Bridgesmith thinks that predictive algorithms and automation are still somewhat in their infancy within the legal community, AI could lead attorneys to much faster, more accurate contract management.

"Where I think this is going sooner rather than later is that AI and machine learning are going to be taking a significant amount of time out of that process and improve the output in terms of quality, with far fewer misses along the way," he notes.

Harry Surden, professor at University of Colorado Law School, agrees that contracts are uniquely suited for automation, and that machine learning could siphon off time wasted on simple, repetitive tasks.

"A lot of contract management is ripe for standardization and automation. If you talk to a lot of contracting attorneys, some parts of what they do is dispose of specially tailored contracts. But with these repetitive patterns of contracting, they probably don't need the tailoring that they do today," he says.

Surden cautions, however, that while contract AI is advancing quickly, it still lacks the complexity of human review.

"I'd say the weakness of this technology is that it's not nearly as good as people yet," he explains. "It can do some things — it can identify different parts of contracts and make some good educated guesses. It's more like a tool to help people sift through contracts, but it's nowhere near the level of what an ordinary person could do."

Overall, Bridgesmith thinks that the significant cost-savings will encourage law firms and legal departments to stay on pace with adoption.

"Lowering the price, that's ultimately the driver that's going to be changing delivery," he adds. "Those are ultimately economic forces that cannot be resisted."

Could it replace contract lawyers?

Terms like "artificial intelligence" and "automation" can immediately put attorneys on edge about technology. Understandably, many are skeptical that automation and machines can replace any part of the work of human attorneys, who have developed legal expertise over years of study and practice.

Most agree that automation will not be replacing attorneys anytime soon, but that doesn't mean it won't become a more integral supplement to legal work.

Bechor acknowledges that while automation can get through a great deal of data very quickly, it can't fully tailor a contract on its own. "If you only use AI, the work product you get is pretty standardized, and the lawyers who use this stuff need to be aware of it. This is support for their work rather than a replacement," he says.

"The AI doesn't know what's going on in the world. It didn't talk to a customer or a client; it doesn't know if there are other things to be considered." Bechor says. "There has to be some element of artificial intelligence and some level of human intelligence. There has to be a level of human intelligence invested in the process in order to provide a good work product."

Surden agrees that while automation may reduce attorneys' ability to charge for easily automatable work—things like putting together boilerplate standard contracts and reviewing basic contracts—they will still have plenty of work to do.

"The basic principle is that where lawyers today are acting like computers, tomorrow they will be replaced by computers, but that's only a very small portion of what lawyers do. A much larger portion is doing analysis in environments of legal uncertainty, where you might have an abstract, thorny business question or other legal questions," Surden says.

"That's the real value added of lawyers, and the computers are not really at the point where they're close to being able to do that," he added.

As to whether attorneys could develop unsafe dependencies on automated systems, Bridgesmith says that technology will not create any problems with cutting corners that didn't exist before. "Long before anyone knew what artificial intelligence was, there was an enormous amount of copying and pasting without a great deal of legal analysis. The systems can do as badly, but faster and cheaper than humans do."

What's coming next in automation?

Though experts generally agree that we're not even close to the pinnacle of this kind of technology, folks have different ideas about where machine learning and technology will go in the contract management world.

Bridgesmith says that as technology works its way further into the mainstream of contract review, developers may need to think more deeply about how to verify authenticity of contracts and their signatories.

Bridgesmith explains that adding authentication and identity verification could help enhance the contract's trustworthiness and security. "That's high level tech, but it'll add to the quality," he says.

For his part, Bechor thinks that AI could move into managing contractual obligations. "The next phase of contract management systems for signed contracts is the automated extraction and management of the contract's data, and creating actionable items based on that," he says.

Surden wrote a 2012 article on "computable contracts," contracts written in computer code rather than standard English. He says that as computable contracts start to take hold in the industry, they may be able to bolster some of the trustworthiness and contractual obligation compliance aspects of contract automation that Bechor and Bridgesmith raised.

By coding contracts directly into machine language, Surden explains that computable contracts could reach into signatory databases to see if contractual obligations have been met, notify parties of upcoming obligations or even automate compliance with contractual obligations.

Original