Running With the Machines

Artificial intelligence in the practice of law

By Sharon D. Nelson and John W. Simek

Artifical Intelligence Graphic

The peaks and troughs of artificial intelligence, or AI, are well documented, and as we are now at a peak, the hype factor gets greater while the reality is lost in the noise. The landscape is changing and large law firms simply cannot afford— for monetary and brand reasons—to be left behind.

Fear of "Robot "Lawyers"
There is no shortage of lawyers who fear AI will replace them. Those who sell AI have come, in the last two years, to realize that it is hard to sell a product that people fear will compete for their morphing. ROSS Intelligence, once called “The Superintelligent Lawyer,” on its website, is now referred to as “an advanced legal research tool that harnesses the power of artificial intelligence to make the research process more efficient.” That is a major change in tone.

Indeed, we are going to have to find a way to coexist with AI. This may take the form of new jobs made possible with AI or new ways of doing our jobs. No matter what the vendor marketing says, it is clear that jobs will be lost—and it is probably a fool’s mission to predict how many.

Understanding AI
Amid so many resources on AI, the Defense Advanced Research Projects Agency, or DARPA, has a good handle on what constitutes AI:

Handcrafted knowledge (many systems have this). These systems can’t really learn and handle uncertainty poorly— they can only enable reasoning over narrowly defined problems. The early self-driving cars were in this category, unable to distinguish a shadow from a rock in desert driving; not knowing where to drive to be safe. Most cybersecurity applications fit here—they can study computer code, compare it to known vulnerabilities and fix it, but that’s all.

• Statistical learning (such as Kira Systems, ROSS, and Lex Machina). These applications are trained on big data. They perceive the natural world, may have facial recognition, and learn from data sets. Their reasoning intelligence and abstracting capacities are still limited. They are best at classifying data and predicting its consequences. They are statically impressive but in individual cases, often unreliable. It took less than 23 hours in March 2016 for Twitter to corrupt Tay, a bot devised by Microsoft for what the company described as an experiment in “conversational understanding.” Microsoft said the more you chat with Tay, the smarter it gets, learning to engage people through conversation. But Tay was bombarded with racist, misogynistic remarks—and Tay began to respond in kind. Microsoft pulled it in less than 24 hours.

• Contextual adaptation (we’re not there yet). These systems will construct explanatory contextual models to explain, for instance, why they made a decision that a cat was a cat. Sounds simple enough, but the reality is very complex. These systems will also reason and learn in a much more human way.

Where Legal AI Is Today
Michael Mills, co-founder and chief strategy officer of Neota Logic, regularly updates a graphic that shows the current state of AI in the legal industry. He identifies players in the following areas of law: e-discovery, contract analytics, prediction, legal research, and expertise automation. There are roughly 40 companies focused in the legal sector that Mills believes qualify as using AI. What can AI actually do? Rich Kathuria, Gowling WLG national director of project management and legal logistics, said:

“AI shows real potential—in the right circumstances and even in the not-so-right circumstances. We used Kira recently for a very large contract analysis project for one of our clients. The project involved reviewing various agreements and documentation to assess the risk associated with various assets. Since Kira did not have built-in models for these types of documents, Kira was not immediately able to extract the required information automatically. But we were able to use the learning capabilities of Kira to teach it to identify the key clauses within the documentation that we were looking for.

Kira learned these well and after the training, it was able to pull out the relevant clauses in various documents. In addition, Kira’s ability to convert the scanned documents into readable text and run comparisons against other similar agreements made the project run much more efficiently.”

That’s a pretty good endorsement. One of the major features of the new generation of AI is the fact that the machines are learning—faster and with more reliability.

There is no doubt among experts that technology assisted review, or TAR, contains some AI. Machine learning, natural language processing, or NLP, and similar techniques are all AI processes that can be used to identify specific document categories and to search for relevant information in documents. While many companies offer TAR, one of the leaders is Catalyst. In October 2016, Catalyst released a peer-reviewed graphic showing how, using TAR 2.0, one reviewer could do the work of 48 reviewers using TAR 1.0—reviewing 723,537 documents in five days.

Contract analytics
JPMorgan Chase is saving money by using software called COIN—short for contract intelligence—to review commercial loan agreements. The software reviews documents in seconds, doing work that once required 360,000 hours of work each year by lawyers and loan officers. The bank said the software has helped reduce loan-servicing mistakes that were often attributable to human error in interpreting 12,000 new contracts per year.

Lex Machina, acquired by LexisNexis in 2015, transforms data from federal court dockets into live charts. Lex Machina’s motion chains enable you to analyze your odds of success for a specific type of motion based on historical data. Similarly, you can see how opposing counsel has performed on similar cases or before your judge.

Legal Research
ROSS Intelligence is the AI platform that first seemed to catch the attention of the legal world. Schooled originally in bankruptcy law, ROSS now works on intellectual property, knowledge management, and contract review systems. ROSS can read more than one million pages of caselaw per second.

Expertise Automation
Certainly one well-known leader in this area is Neota Logic, which offers an AI platform that enables clients to automate their expertise at internet scale through a useful form—as applications embedded in business systems or consulted interactively in a browser. Neota announced the release of Neota Logic System 8.0 in 2017, which included a comprehensive redesign of the proprietary hybrid reasoning engine that is the foundation of the platform.

Calling the Future
We are taking a chance here—and prepared to eat healthy slices of crow pie if we are wrong—but we are pretty sure that the practice of law will morph quickly over the next decade thanks to AI. Not all the changes will be welcomed, but we will have to learn to run with the machines. It’s that or extinction. Time to lace up those athletic shoes. TBJ

This article has been previously published and has been edited and reprinted with permission.

Sharon NelsonJohn SimekSHARON D. NELSON and JOHN W. SIMEK are the president and vice president of Sensei Enterprises, a legal technology, cybersecurity, and digital forensics firm based in Fairfax, Virginia. For more information, go to

{Back to top}

We use cookies to analyze our traffic and enhance functionality. More Information agree