Insights from the Thomson Reuters Legal Geek Takeover: Contract Review and Machine Learning
Thomson Reuters recently hosted a Legal Geek takeover, a half day of virtual sessions on legal innovation, technology and trends. A highlight of the event was the Contract Review: Themes in Machine Learning session hosted by a Thomson Reuters team of Carolyn Bussey, senior user researcher; Steve Fullerton, software product manager; Masoud Makrehchi, director of Research; and Hedy Xu, product manager.
Their session explored machine learning and its role in contract review. Legal Current had an opportunity to catch up with the team after the Legal Geek Takeover, and they shared insights from their session. Below is a recap of the conversation.
Legal Current: How are machine-learning technologies driving efficiency gains for lawyers in the contract review process?
Bussey: When they work well, machine-learning technologies can help minimize the “mindless” work – think organizing documents by type, or finding a certain passage within hundreds of documents – and let the human reviewer get to what they need to read. At the end of the day, machine learning can’t understand legal language and implications, but it can help a human dig into that work. Being able to efficiently gather data points about a contract can also lead to exciting automation opportunities – think automatic generation of standard reports, easier repapering and smart workflow alerts that keep people in the loop when certain information is gathered.
LC: What’s an example of an existing tool already being used, or tested, by law firms and their clients?
Xu: A variety of tools on the market offering machine-learning technology-assisted contract review, including Kira, Leverton, Luminance, ThoughtRiver, Eigen, eBrevia and RAVN. During our Legal Geek Takeover session, we polled our audience and found the majority of participants were unsatisfied with offerings they had tried; 69% said that AI tools only partially met or didn’t meet their expectations.
LC: What’s the potential application of machine learning to further deliver value in the contract review process?
Fullerton: Today, we see lots of opportunities in various transactional workflows. The top two use cases in our Legal Geek Takeover session survey were due diligence reviews and contract management. These two areas are great places for an AI tool because they are frequently very high volume, and often require large teams to complete today.
Also, our Legal Geek audience was very interested in seeing how these tools could enable reviews of documents in other languages! Most current tools are English-language-only. Those that support other languages are less able to interpret legal meaning and simply identify clauses.
LC: What are the biggest challenges in delivering machine-learning-assisted contract analysis?
Makrehchi: It’s still a challenge for machine learning to understand even simple human language – let alone legal documents. Sometimes, very differently written clauses can mean the same thing; other times just a single word can change a meaning in a clause. That kind of meaning and nuance is extremely difficult to map with machine learning techniques.
Both legal concepts and machine learning capabilities are complex, nuanced things. Making sure that we use the right algorithms – and that we’re getting the right results – is at the heart of this project. Our team has included Practical Law editors, data scientists and UX professionals from the start. It’s important for all these teams to advise our product development in order to make a solution that’s accurate, useful and usable.
LC: What’s the one thing legal professionals should know about machine learning and contract review?
Bussey: It can be difficult to get to the bottom of what is “marketing spiel” versus the true capabilities of machine-learning-based contract analysis tools. Many of them require further training and setup to be useful. Our aim is to have a solution that’s easy to understand; we want to make it easier for lawyers to feel more confident that nothing has been missed in – hopefully – much less time.