There’s a cognitive dissonance that arises for those in the legal industry reading the headlines touting AI as a transformative force in law.
The media stories cover both the long and growing list of AI companies promising to revolutionize how law is practiced as well as the long and growing list of partnerships between AI companies and large law firms.
At the same time, most lawyers report that they don’t think AI has upended their law practice. Is there too much hype? Are lawyers holding out for the technology to advance before embarking on adoption? Or are practices lagging behind innovation and what’s available in the market?
Before we begin, we should first be clear on what we mean by AI. The term is used very loosely. It’s often a label attached to software that, while complex and impressive, does not use computing techniques which are part of traditional branches of AI research. For example, a calculator performs a function that humans, albeit on a smaller scale, can perform: arithmetic. Nonetheless, we wouldn’t consider a calculator to be an example of AI. AI must mean something more. This article defines “artificial intelligence” as software that uses algorithms and techniques pioneered through AI research. At this time, the most promising area of research is machine learning, and more specifically deep learning.1
One reason why most lawyers don’t think AI products have had a significant impact on their practice is that AI often works in the background to optimize software we would not traditionally think of as an “AI product”. For example, enterprise search engines may utilize AI to improve data management systems’ search results so that more relevant documents are produced. That’s why iManage acquired RAVN Systems, an AI platform that helps classify, organize and summarize data from large volumes of documents and unstructured data. Another example is e-discovery software, which may also use AI to improve the relevance of documents produced during a document review.
The uptake challenge every software vendor faces when selling to a law firm has also historically been part of the problem. But while the legal industry has been slower than some in its adoption of true AI-based technology, the right software has shown to be quickly adopted by lawyers. For example, e-discovery software revolutionized document review within a few years and received traction almost immediately. Law firms are now, more than ever, taking innovation seriously, having seen the benefits of effective legal tools in the nascent legal tech market.
Part of the adoption problem lies with the software that is available. The most mature and widely adopted true AI software (i.e., not merely AI optimized) is certainly in the document review space, which has sparked competition from providers such as Kira Systems and Luminance. Outside of document review, many AI products available today are either not true AI (as we define it) or have a narrow use case which impedes broader adoption or requires further development.
And so as most lawyers will attest, it is difficult to find true AI software lawyers can build into a regular part of their practice. There are a few reasons for this.
For one, AI needs an incredible amount of data to be effective. A 2018 McKinsey report notes the industry leaders in terms of AI adoption are in the high tech and financial services sectors. That’s because both of those industries produce a massive amount of data. In contrast, legal service providers do not, either individually or in the aggregate, produce nearly as much data. For example, to teach AI software to classify contracts you would need many examples of the same class. While you might have millions of pages worth of contracts, you might only have five silver streaming agreements. As a result, it is hard to “teach” AI software to recognize and classify all types of agreements in your data management system.
AI will be an increasingly important tool in the future and lawyers will be fast out of the gate to integrate these technologies as they emerge.
Adding to the challenge is that the data that does exist in the legal industry is segregated within smaller parties (e.g., law firms, in-house departments, alternative legal service providers etc.). Tech companies do not face this problem because the major contributors in AI, such as Google, Amazon and Facebook, each produce enough data on their own. Still another obstacle is that the little data legal service providers do have is, for the most part, not organized or easily searchable by mathematical functions—in other words, it’s unstructured data. Natural language processing AI is getting better at sorting through unstructured data such as sound files, but further research is needed before it can tackle tougher tasks.
The legal sector is not alone in the limited role AI currently plays. KPMG surveyed 400 executives with ongoing AI projects in their companies. Of the executives surveyed, 51% said it would take three to five years before their AI projects result in significant returns, but just a year ago only 28% held the same view. The number of executives who expected quick results (three years or less) went down from 62% to 47% from last year to this year.
As it stands today, there are a few sophisticated, true AI tools that are mature enough in their technology to make a significant impact on a lawyer’s day-to-day job—and we are only at the outset of the widespread use of AI applications in the legal sector. We believe AI will be an increasingly important tool in the future lawyer’s tool belt, and that lawyers, as diligent adopters of tools that will work for them, will be fast out of the gate to integrate these technologies as they emerge. The first wave of broadly adopted AI solutions will probably be designed to serve a very specific purpose. As AI research progresses, the number of narrow AI solutions will increase until lawyers across different sectors have access to many niche products, which, in the aggregate, will make a real difference in their particular practice.
This is still the most exciting time to be both a lawyer and a tech geek. The landscape of legal software is rich with mature products that can accomplish great things for legal service providers. The most compelling software on the market is most often not AI-based, but is no less impressive. Or it might not be software at all: lots of efficiencies can be realized by simply developing a workflow, identifying inefficiencies and fixing them with better process and communication (for more on process innovation, see “Four steps to an optimized role in contracts for in-house counsel”).
1 There are other historical branches of AI research but none are as promising as machine learning currently is. For example, “expert systems” was a thriving area of search that was a natural fit with law because it worked by attempting to codify logical rules, which are also the building blocks of law.