By using website you agree to our use of cookies as described in our cookie policy. Learn More

TDWI Upside - Where Data Means Business

Q&A: Firm Uses Analytics to Review Contracts for Free

LawGeex combines text analytics, machine-learning algorithms, and crowdsourcing to review employment contracts free of charge.

Using LawGeex, a consumer can upload an employment contract and receive an in-depth report at no charge. According to the company, the report details "what's standard, what's questionable, and even what's missing from the contact."

The company says its solution has been used to review tens of thousands of contracts of many different types, including some from tech giants Apple, Google, and Facebook. Companies such as Deloitte and Brandwatch use LawGeex's legal contract review software; the company recently received $2.5 million in venture capital funding. It plans to offer free reviews of additional types of legal contracts soon.

In this interview, LawGeex's CEO Noory Bechor explains how the solution works, how it takes advantage of crowdsourcing, and the challenges of using analytics to find the nuances in thousands of legal documents.

Upside: Using LawGeex, a consumer can submit an employment contract on your website and get back an in-depth report at no charge. How does that work?

Noory Bechor: The underlying premise behind LawGeex is that nobody should sign a contract they don't understand. Yet every year, a significant number of Americans go without legal services mainly because lawyers are simply too expensive. Let's take the employment market, for example. In 2013, one study found that 21 percent of full-time employees in the U.S. planned to change jobs in 2014 -- that's nearly 25 million people. A significant number of those are likely to need their employment contracts reviewed.

When these new employees are switching jobs (or even careers), they are at a competitive disadvantage because they likely do not know what is considered "standard" for their new position and, thus, for their employment contract.

For example, employment contracts often have non-compete clauses for various lengths of time. We believe that prospective employees should know how many other employment contracts have this clause included and exactly what other people in the industry are agreeing to. Knowing this gives the employee bargaining power.

When LawGeex reviews the contract, it is using machine learning, text analytics, and a vast and growing database of over 50,000 contracts to compare each document to thousands of similar ones. Within 24 hours, we provide the user with a contract score (as it scores against similar contracts) and a full-report summary. The report indicates whether various clauses are standard, questionable, or missing.

Can you describe how text analytics, machine-learning algorithms, and crowdsourcing all come together in this solution?

We use text analytics to understand exactly what the contract is saying and translate it into plain English. We use crowdsourcing to increase our database by allowing anyone to upload contracts to LawGeex. Finally, our machine-learning algorithms learn from every new contract processed by LawGeex, making our feedback more accurate. The more contracts we see and review, the better the reports get.

Why has all of this come together at this time? What has made it possible?

This is a great time for a solution like LawGeex. The legal technology market is exploding and new legal service providers are entering the market every day. Also, our solution integrates artificial intelligence; only in the last couple of years has that technology advanced far enough to offer significant advantages.

As a former business lawyer, I understood where the legal market was failing, in no small part because the practice of law is as ancient as civilization and has been reluctant to change until now. This created an opportunity for solutions such as ours.

Some of the technologies and changes that have made this solution possible include:

  • Advancements in machine learning, especially in deep learning.

  • Cheaper, more available processing power. That's important because training new models requires a great deal of computational resources, including faster cores. The ability to rent those resources allows anyone to use ridiculously large supercomputer clusters for short periods of time -- the same task would have previously taken days or even weeks.

  • Legal professionals are being exposed to similar technology in other industries, making them more accepting of technology in their own profession.

What are some of the challenges of using analytics against the kind of unstructured text found in a legal contract?

The biggest challenge has been recognizing and understanding the legalese used in the contracts, and then transforming this into clear explanations for the user and valuable data for our learning algorithms. Most contracts we see today are based on a common template, but over two-thirds of them deviate in various ways. For example, there could be an additional unusual clause or a standard clause that has been removed. It's these deviations that our analytics engine picks up, and this is what we use to create the final report for the consumer.

Another key challenge is prioritizing and organizing the various possible insights that our engine generates, and then picking only the most important ones to present to our users. In other words, although we can say 100 different things about a contract, the challenge is to only say the most important things. We want to avoid burdening users with feedback that isn't a high priority to them.

A lawyer might argue that a machine can't tease out the nuances of a legal contract and find the holes, missing elements, potential legal problems, and so forth. How do you respond to that argument?

To an extent, that is correct and LawGeex does not purport to replace lawyers. It's quite the opposite actually -- after using LawGeex, our users should have much more confidence in deciding if they should see a lawyer. If they do, they can use the lawyer much more efficiently. When it comes to the details of negotiating terms (such as salary, vacation days, and so forth), those aspects LawGeex leaves to the individual.

Our machine-learning platform is getting smarter every day, never gets tired or distracted, and does not forget anything (the same cannot be said for humans). It is extremely accurate at determining whether specific clauses are out of place or should be included. In essence, we serve as a legal information provider -- allowing the user to decide if it is generally safe to proceed or if they really need to consult a lawyer.

You've said that your platform has been used to review tens of thousands of contracts. Can you discuss some of the technology issues that you've uncovered and addressed during that process?

Our greatest challenge was developing the algorithms to recognize, normalize, and translate the legalese in the contracts into plain English, which we provide users in the report summary. This took an incredible amount of effort and ingenuity from our team of engineers. Now that we've successfully developed a platform that can continuously and systematically learn and translate this legalese, we can provide users with the accurate and relevant feedback they need.

Regarding technical issues specifically, legal language is different than plain English -- enough so that traditional natural language processing tools do not apply. A very small change in the text can have significant meaning, sometimes even completely reversing the meaning of a legal clause. Even simple contracts such as non-disclosure agreements might contain hundreds of unique legal concepts.

Why did you choose to start with employment contracts in particular?

As I mentioned earlier, tens of millions of full-time employees look to change jobs every year. That's a significant number of employment contracts that are being written and negotiated, or in many cases not negotiated.

Even more startling is the fact that millennials are now the largest generation in the workforce -- there are an estimated 53.5 million millennials in the workforce. That means that adults between the ages of 18 and 34 now make up one in three American workers, and they are more likely to accept technology and use the LawGeex solution.

You're offering your contract review solution to consumers free of charge. What's the business model?

We generate revenue from our business product. We offer a subscription service that allows businesses to use our platform to review all of their contracts, saving them hundreds of hours and thousands of dollars in legal expenses. Our prices start at $20 a contract, and we believe we are currently the fastest, easiest, and most cost-effective way for small businesses to check their contracts.

We launched the consumer side of the business because we genuinely believed we could help tens of millions of people understand exactly what they are signing and thereby negotiate a better deal.

What's next in the legal arena for this technology in general?

Artificial intelligence (AI) technology is beginning to impact many aspects of the legal space, including litigation, discovery, and contract review. Moving forward, I see AI and machine learning technologies taking over a large amount of the mundane daily work of many legal professionals. If built correctly, these technologies can do the same work better and faster.

As with other industries that have been disrupted by technology, law will still have a place for humans -- lawyers will continue to serve as the main source of knowledge. However, they'll be able to focus their time handling high-end tasks such as helping people and businesses negotiate more complex and highly specific contracts.

TDWI Membership

Accelerate Your Projects,
and Your Career

TDWI Members have access to exclusive research reports, publications, communities and training.

Individual, Student, and Team memberships available.