All lawyer jokes aside, artificial intelligence (AI) finally is beginning to make inroads in the legal industry. But how deeply will we feel its presence? And which practice areas will it benefit the most?

In the realm of contract drafting, we have already seen the advent of online automated legal services such as Rocket Lawyer, LegalZoom and LawDepot. These services offer the consumer basic legal forms but face legitimate criticism regarding the sufficiency of these agreements in protecting a client against risk, both known and unknown. While online document generators can fill an immediate need in providing a product for a client on a budget, or with a tight deadline to meet, these agreements often are not adequate in terms of anticipating legal pitfalls that might otherwise easily have been recognized by an experienced attorney.

Because of the many shortfalls associated with automated legal documents, sophisticated legal consumers and law firms themselves are beginning to explore the role, if any, that AI can play in drafting contracts and whether this innovation can be relied on to provide adequate protection for clients. Specifically, machine learning and deep learning techniques are being implemented and packaged as software programs for contract drafting.

To understand AI's promise in this area, it is important to understand the difference in these modalities of artificial learning. “Machine learning” utilizes algorithms to recognize and form patterns from large amounts of data to make future predictions based on such patterns and insight. Deep learning is a subset of machine learning which takes it a step further and uses a layered structure of algorithms that mimic the biological neural network in the human brain. Deep learning is the most advanced form of AI and that which most closely resembles how the human brain works.

Through machine learning and deep learning, contract drafting software will be able to generate a bespoke specific legal form for a user based on certain key terms provided by the user. In essence, the software would, based on the ever increasing stockpile of legal documents in its digital armory and its ability to analyze and create patterns from these documents, take the key terms and phrases input by the user and generate its best prediction of what the user needs. This software would aim to either form a full contract or specific contract provisions based on the keywords and phrases provided by the user.

There are, of course, potential limitations that make the utilization of this software in the legal field problematic. Terms and definitions in legal agreements can take on different meanings in different agreements. This can make it difficult for a software program to accurately pick up on patterns if the context behind contract language is too variable.

Experts point out that AI-contract drafting software would face difficulties in adapting to the law which is constantly changing with society's progress. AI software is based on existing data; unless someone is constantly uploading new agreements reflecting changes in the law, the software would continue to pick up on old clauses that may no longer be applicable or provide adequate protections for clients.

The biggest flaw and perhaps a fatal one of AI-contract drafting software, however, is its lack of foresight or planning, which is one the most critical factors for a lawyer to consider when drafting a contract to protect his or her client. Fracois Chollet, an AI researcher at Google and the inventor of the popular deep learning library, Keras, pointed to the major difference between deep learning versus the human mind as deep learning being “terrible at planning” and “only doing straightforward pattern recognition.” Humans are “capable of forming abstract models of a situation,” and abstract thinking often equals legal thinking. When applied to contract drafting, as long as AI software is only capable of recognizing patterns, and lacks the ability to take into account abstract situations, it will be of limited utility and further review by an attorney for accuracy, enforceability, specificity may still be necessary. To that end, we face similar issues as described above with respect to the existing automated legal document platforms.

Perhaps the goal for AI contract drafting software should never be to eliminate the attorney, but to instead to increase a lawyer's efficiency. Regardless of the circuitousness of its path, it will be interesting to see how artificial intelligence progresses in the field of contract drafting.

Lauren Quattromani is a corporate attorney with the law firm of AXS Law Group in Miami. Contact her [email protected].​