• By Omer Hayun

What Does AI Contract Work Automation Mean?


It seems that the expression “Artificial Intelligence” (AI) has become a buzzword associated with the marketing efforts of almost every company within the entire legal technology industry in the past few years. As always, there’s a certain gap between how AI technology is perceived, what it actually can do and what are its use cases for the everyday legal work of a lawyer. Spoiler alert: lawyers are not going to be replaced by machines anytime soon. Nonetheless, their work is increasingly becoming more results oriented and focused on the material aspects of the legal work. When it comes to contract work, every lawyer who ever had to draft, negotiate or review contracts can testify that there are many micro tasks that can and should be automated by computer software. In fact, some of that tedious routine work is already being automated by legal tech solutions out there.

Legal professionals handle a lot of different types of contract work. One of the problems they encounter is finding the right solutions for the specific problems they are facing. As mentioned above, almost every legal tech startup is describing its product with the words “Artificial Intelligence”, “AI”, “Machine Learning” or “Contract X Automation”. Some don’t even bother to explain what is it that their products do and just prompt you to request a demo.

Of course, a legal technology product can be multifunctional with technologies capable of addressing more than a single pain point, and a lot of them do so. However, these vague descriptions create a marketing “noise”, causing great confusion among lawyers trying to understand the differences between the offerings of each company.

Breaking down the technologies used for contract work automation and understanding their output will help make sense of the software solutions’ subcategories offered in the market. Combined with examples of specific use cases, understanding these technologies will also help explain what a legal professional should ask legal technology vendors when looking to implement a contract work automation solution.

About Natural Language Processing and Machine Learning

Natural Language Processing (NLP) is the scientific field of computational linguistics concerned with the analytics, processing, understanding, and generation of unstructured text. There are many different NLP tasks that data scientists are researching and developing, from gaining insights at a single-word level to gathering information from full textual documents. In the legal technology space, there are four main NLP tasks involved in the automation of contract work designed to assist lawyers with either drafting or review of contracts, including Named Entity Recognition (NER), Text Classification, Natural Language Understanding (NLU) and Text Generation.

Named Entity Recognition is the task of recognizing proper names, such as people’s names, organizations, locations, dates, percentages, monetary values and etc., within an unstructured text. In the contract analysis context, this could refer to recognizing the parties to an agreement, its effective date, automatic renewal dates, defined terms, or financial terms of a transaction (prices per share values, annual salaries and so on). Text Classification tasks, as their title implies, are about assigning textual segments a certain class or category according to its topic or other user-specified characteristics. The technologies under this category are often used for the classification of contracts and clauses into categories, but also for a more in-depth classification of sentences and phrases into legal concepts within an agreement. For example, most agreements contain a “Governing Law” clause, declaring the laws in light of which an agreement should be interpreted and ruled by. Once a clause is classified as a “Governing Law” clause, the next analytical step is to extract the information regarding the country or state of said law, a task usually performed using Named Entity Recognition techniques. Natural Language Understanding is a subfield of text classification, that deals with more than just classifying textual segments by their topics. It is the process of teaching a computer to understand text and try to determine its meaning. This is by far a much more complicated goal to achieve using a computer program. For instance, while it is fairly easy to identify that an “Assignment” clause exists in an agreement, it’s harder (but not impossible) for a computer to determine whether that clause means that an agreement can or can’t be assigned to third-parties, and if so, to whom it may be assigned and to whom it can’t be. Of course, as the number of potential meanings or options increases, it is more challenging to establish the actual legal meaning of the textual content. While the three former NLP categories deal with different aspects of extracting information from existing text, the fourth category of Natural Language Generation is about the methods used to create textual content using computer algorithms. In the context of contract work, this refers to the generation of new clauses and contracts or making suggestions for drafting revisions for the purposes of contract negotiations.

There are two main approaches to performing these NLP tasks (each resulting in varying levels of accuracy). With the Rules-based approach, language processing systems are governed and designed by pre-defined rules, aiming to apply a rule to every possible scenario and text variation, in order to extract or generate the desired information. No rule – no results. The newer and more advanced approach is called Machine Learning, which is a category of statistical methods used to train computer programs to infer specific outcomes given real-world examples, without pre-defining a set of rules. This is also a “Black-box” approach since the results of applying a trained machine learning model on a set of data cannot be explicitly explained by it. In the legal tech industry, they are both often referred to, in general, as simply “AI” or “Artificial Intelligence”. Although very helpful in improving contract workflows, these technologies are still not advanced enough to replace lawyers in preforming contract work and probably won’t be anytime soon. It is better, therefore, to focus on what these technologies can do for lawyers.

Contract Review Automation

When it comes to contract “review” or “analysis”, there are many examples illustrating the variety of review tasks lawyers handle during their daily work. Although they all require reading agreements, the motivations behind each specific contract review and the actions followed by it vary depending on the task at hand.

For example, a legal intern, working at a law firm’s corporate department can be assigned with a due diligence project in which her only task is to go through volumes of agreements and find specific types of clauses like “Change in Control”. Another example could be of a corporate legal department, responsible for tracking and managing all of the rights and liabilities of the company, that is required to review all active procurement agreements and check whether they are still in force, if they automatically renew or not, and if so – when is their renewal date. In another case, a lawyer representing a company against a potential share purchaser needs to negotiate a share purchase agreement and review a third-party draft sent to him by the other party. He will have to carefully read each clause in the draft, understand the terms of the structured deal and mark up terms that are unfavorable to his client. After finishing his substantial mark-up, he will then ask his paralegal, intern or first-year associate to review his work and make sure the legal terms in the agreement are properly defined and that the cross-references between clauses are not broken.

These examples come to show, that when introduced with an all-in-one general “contract review automation” product, the first question the legal professionals need to ask themselves should be “When will I use this tool?”, meaning, at which point during the contract life cycle will the output of the “automated analysis” will help reduce manual tasks. Therefore, when discussing review automation it’s important to distinguish between pre-execution and post-execution contract work. The main difference being, of course, the purposes of the review and whether it should or shouldn’t be followed by drafting actions from the lawyer. Post-execution contract review usually requires the legal professional to understand and extract certain information from contracts that are already signed and executed in order to perform tasks that are outside the scope of the agreement itself. Pre-execution contract work, however, requires, in most cases, that he will make changes to the contract draft during the contract negotiation process. The second question is whether the output of these products and the complexity level of the actionable information they provide is enough to act upon.

The first two cases described above are text-book examples for post-execution contract review software that can help automate Due Diligence and Contract Management tasks. These types of tasks generally require reading and understanding contracts without making any changes to them. Both of these contract analysis tasks apply Text Classification and Named Entity Recognition technologies to extract data points that enable legal professionals to get a quick overview of the contents of an agreement at a higher level, without reading it thoroughly. The insights gathered for these types of contract review work, however, have limited value when it comes to tasks performed during the pre-execution stage of a contract life cycle, demonstrated by the third and fourth examples above.

Third-party contract review performed as part of a Contract Negotiation process, as described in the third example, requires not only an in-depth understanding of the legal meaning of each element in an agreement, but also an understanding of the relations between them, their coherence and their favorability to the lawyer’s current client. This type of review requires Natural Language Understanding technologies operating at a deeper level. The type of output expected from this kind of technology is that the program will not only point out the risks and problems in a third-party draft, but also provide concrete language to resolve them, which, in itself, should rely on Natural Language Generation methodologies. Effective Contract Negotiation tools are, therefore, designed to streamline both of the substantial review and drafting aspects of the negotiation process. Other than that, the end of each contract draft iteration during the contract negotiation involves another type of contract review for the purposes of Legal Proofreading. Unlike the other review types described above, this analysis has more of a technical nature, since its purpose is not to determine whether the agreement contains the “right” legal concepts, but only to ensure that the internal logic of the agreement is intact.

Contract Drafting Automation

The term Contract Drafting Automation suffers from the same ambiguity. This term

usually triggers lawyers to instinctively think about Contract Generation software, which is, in fact, one of the main categories of drafting automation tools, but not the only one. This type of legal technology normally provides a framework for lawyers to helps them create a questionnaire or pre-defined form which are later used to fill-in the blanks and quickly generate a template-based contract.

Although the distinction between pre-execution and post-execution is irrelevant, since all of the contract drafting work is obviously pre-execution, it’s important to understand whether a “Contract Drafting Automation” tool refers to the drafting of an initial contract draft, or to drafting tasks that are part of the negotiation process, also referred to as Contract Redlining. As previously discussed, the redlining aspect of contract negotiation is complementary to the review tasks involved in it. Since the review and drafting tasks during negotiation are practically inseparable, and Legal Proofreading is also a stage of the pre-execution contract work, some companies in the legal tech space actually refer to all of these types of tools as drafting assistance tools.

The differences between these types of drafting tools are expressed not only in the stage in which they are used , but also in the technologies that power them. On the one hand, Contract Generation tools are mostly workflow automation platforms designed to help legal professionals build sets of rules that will generate all of the legal possibilities for a certain type of agreement. Contract Redlining software, on the other hand, apply textual generative technologies to tailor the language of a contract using substantial drafting suggestions at a deeper and detailed level. Because of that, Contract Generation software can be limited in scope and provide value in simpler types of agreements, while redlining software, if trained on enough datasets, could be useful even for complex transactions.

Final Words

The next time you hear about a contract review, analysis or drafting product, it’s better not to assume you know exactly what it does. The variety of technologies and use-cases that fit the same general categories is huge, so it’s just a matter of asking the right questions to understand whether a product could be a good fit for your organization.

About the Author

Omer Hayun is the Founder and Chief Executive Officer of Bestpractix, a legal technology startup developing a pre-execution contract negotiation platform that provides lawyers with smart drafting recommendations and streamlines their contract drafting and review. He is a former lawyer and a Natural Language Processing expert with a business and an engineering background.

#OmerHayun #NeedtoRead #AllUpdates



© 2017|2020 LegalBusinessWorld™

Australia | Asia | Canada |  Belgium | China | Cyprus | Czech | Denmark | Estonia | Finland | France | Germany | Greece | Hungary | Hong Kong | Iceland | India | Indonesia | Israel |Italy | Japan | Latvia | Lithuania | Luxembourg | Malta | Mexico| Norway | Poland | Portugal | Romania | Russia | Singapore | Slovakia | South America | Spain | Sweden | Switzerland | Taiwan | The Netherlands | Turkey | Ukraine |United Kingdom | United States |Middle East | Africa