Wisconsin Lawyer: Technology Is It Time to Adopt Generative AI into Your Legal Writing?:

Home > News & Publications > Wisconsin Lawyer > Article

July 20, 2023

Technology
Is It Time to Adopt Generative AI into Your Legal Writing?

Advances in generative AI technology have many attorneys questioning how the legal field will survive. The solution? Embrace it.

John Fuller

What AI Lacks

Anthony E. Davis compared humans and AI in the 2020 ABA article, The Future of Law Firms (and Lawyers) in the Age of Artificial Intelligence.¹ He found that AI currently lacks in four traits: judgment, empathy, creativity, and adaptability. Attorneys need to use these traits to write documents to the standard required by the legal field. Attorneys need judgment to properly advise the client and determine which arguments and sources are most persuasive. They need empathy to understand the client’s needs so they can understand which arguments and solutions best fit those needs. They need creativity to find unique winning solutions even when it is not the most logical solution. Finally, they need adaptability to overcome any new and unexpected facts their client’s case throws at them.

John R. Fuller, U.W. 2023, plans to join Perkins Coie as a patent attorney in the San Francisco Bay area.

In other words, for AI to completely replace lawyers, it needs to have the same high level of judgment, empathy, creativity, and adaptability that humans possess. As with any profession, different types of work products require differing amounts of these human-centric traits to meet a given industry standard. To better understand where AI currently stands and where it may be developed, this article divides AI-generated written work product into three levels, with each level having a different answer as to when you should start using AI in your workflows. Each of these levels also requires different levels of the human characteristics of judgment, empathy, creativity, and adaptability to create a quality work product.

Three Levels of AI-generated Work Product

Level One: Level one generally encompasses documents that AI can completely construct with minimal input, such as basic contracts or form documents that can be modified to adapt to different matters. Human input or review might already be either unnecessary or minimal.

Level Two: At level two, AI can write significant portions of the document by analyzing detailed prompts or what the attorney has already written, with briefs and memos being created based on this attorney input. Human input is still critical to creating these documents.

Level Three: At level three, AI can write portions of more complex documents based on what the attorney has already written, such as patents. Any given work product could be seen as a level three work product if the fact pattern is complex or the area of law is unexamined. This level is not fully realized and so it requires significant, if not total, human input.

Understanding the current capabilities of AI and knowledge about your work product’s level will help determine how efficient AI can make your workflows. AI, just like any technology, can create more efficient workflows for more routine work products, so that is the type of work product that will be analyzed at each level. Doing so will give a clearer explanation of why AI has the required amount of judgment, empathy, creativity, and adaptability for one level and not another while also helping provide an answer to when you should adopt this technology.

Level One: Contracts and Form Documents

AI tools, such as Lexis® Smart Forms,² Westlaw Form Builder,³ or Luminance,⁴ currently allow attorneys to generate many contracts and forms in seconds. This is possible because these tools modify a form document written by you or another attorney based on inputs, such as party names or the provisions you want to include. Often the same contract or document can be reused with minor changes to such things as the party names and dates, so these AI tools allow you to make these changes almost instantly, eliminating the tedious task of editing the form document. Currently, these tools are not at the generative AI level seen in ChatGPT, but tools such as Bloomberg Law’s Draft Analyzer,⁵ Spellbook by Rally,⁶ and Casetext Co-Counsel⁷ can analyze your contract and recommend standard language found in other similar contracts. These tools currently require at least some writing by an attorney, but that may change as these tools become more sophisticated.

AI can already automate so much of this process because these documents require a relatively lower level of judgment, empathy, creativity, and adaptability. First, these documents need a lower amount of judgment because standard language is reused, meaning the primary source of judgment is determining which already written provisions are best to use for the given situation. Second, these documents are often relatively standard and tend to not have an emotional aspect that an attorney needs to understand, so they only need a minimal level of empathy. In this same vein, writing different unique, creative language for every contract is usually not wanted or advisable, so a high level of creativity is unnecessary. Finally, there needs to be some adaptability to be able to modify the documents given new facts or changes in the law.

AI can already generate these documents from a base form document, but it will soon reach the point where it has a high enough level of the traits to generate them without any form document. While AI already meets the minimum amount of empathy and creativity required for a baseline form document, it still lacks in judgment and adaptability. AI has not reached the level where it can perfectly judge a client’s facts or adapt to changes in the law to know exactly what language to use in a contract or document. Lexis+ AI⁸ and Thomson Reuters⁹ recently announced new AI tools that aim to change this by generating contracts without a prewritten form document. Thomson Reuters’ tool will combine Westlaw and Microsoft 365 Copilot¹⁰ to generate contracts entirely within Microsoft Word, while Lexis+ AI will generate them within LexisNexis. The standard language of these documents means there is a high likelihood that tools like this will be able to accurately write these documents given no more than a few prompts, including basic facts about the situation and certain client preferences.

In their current state, once the form document is written, these tools can save you countless hours of editing by instantly producing incredibly accurate client-ready documents. Still, as more tools integrate ChatGPT and other large language AI models, these documents will soon be able to be entirely written for you. If your practice involves using contracts and form documents, you should immediately investigate adopting these tools or risk falling behind those who do. Even if your practice doesn’t include drafting basic contracts or form documents, awareness of these established AI tools is critical as they will pave the way for forthcoming AI tools at levels two and three.

Level Two: Briefs and Memos

AI can’t write a perfect brief or memo for you in its current state, but emerging technology may soon make that a reality. Tools you may already have access to, such as Lexis Brief Analysis,¹¹ Westlaw Quick Check,¹² or Bloomberg Law Brief Analyzer,¹³ can already take a prewritten brief and suggest additional relevant authority. This makes your research more efficient by suggesting authority you may have missed or allowing you to update an old brief with new authority. Casetext Compose¹⁴ takes this to the next level using prewritten language to generate motion briefs. A human picks arguments, legal standards, and precedent from a list in Compose and adds facts, and Compose’s Parallel Search automatically finds precedent to generate a tailored motion brief. These tools all require at least some writing to be done by the attorney, but newly announced tools promise to change this.

Unlike the more basic documents found in level one, these documents require a higher level of judgment, empathy, creativity, and adaptability, making them more difficult for AI to generate. First, these documents need a higher level of judgment to know which authority to cite and, especially, which arguments are best. Second, when writing a document like a brief, a significant amount of empathy is required to understand and effectively advocate for clients’ needs. Third, these documents require an increased level of creativity to overcome a bad fact, find a unique winning argument, or look to an analogous area of the law that may perfectly fit a client’s facts. Finally, there needs to be an increased level of adaptability so the document can be modified or rewritten given new or uncommon facts.

AI has yet to reach the point where it has a high enough level of these four human traits to generate these documents without the need for attorney verification. However, the advent of new technology, like ChatGPT, can now harness massive datasets to form large language models and neural networks to attempt to teach an AI tool to mimic the problem-solving and pattern learning abilities of the human brain. This type of AI model combined with the sheer input of data makes it likely that AI will reach the point where it will have the proper amount of judgment and adaptability to be able to write significant portions of these documents.

The more limiting factors, though, are empathy and creativity. Unlike the documents at level one, these documents often contain emotional facts or situations for which the most logical choice is not necessarily the best one for the client. It takes high levels of empathy and creativity to know which legal arguments are both best for the client and also creative enough to win. But recently developed and newly announced programs like Casetext Co-Counsel, Lexis+ AI, and Harvey AI¹⁵ promise to overcome these limitations and start generating huge portions of your briefs and memos. These tools are trained with legal data such as case law and reference material, making them more accurate, but they can still make errors. Casetext Co-Counsel can already generate legal research memos with no need for prewritten language based on user prompts. Lexis+ AI will be generating briefs and legal memos in the near future. These tools still require an attorney to know how to properly prompt the tool by continually reprompting it until it generates the right brief or memo for the situation. Currently, the documents generated by these tools must be verified for accuracy and should be used as starting points and not as final documents.

In their current state, these tools can save you hours of legal research (and the potential embarrassment of missing key precedent), but soon, they will save you even more time by drafting complete briefs and memos to a quality that meets the high standards of the legal field. If your practice involves writing a lot of briefs and memos, the currently available brief analyzers are worth adopting now. More powerful AI tools like Casetext Co-Counsel and Lexis+ AI are worth looking into as they become more widely available because they promise to automate even more of your brief and memo writing. Even if your practice only involves more complicated documents such as those discussed in level three below, advancements in level one and two AI tools will have opened the door for AI tools that are just on the horizon.

Level Three: Complex Documents and Specialized Practice

Many areas of law do not deal with contracts, form documents, or brief writing, but instead require attorneys to spend much of their time creating complicated arguments and documents in niche areas of practice or law. These documents may be complex due to the facts of the case or dealing with an unexamined area of law. A particularly salient example of this type of practice is patent law.

Patents are a specialized niche area of the law where each new patent also routinely deals with complex facts, often starting with a brand-new invention that is supposed to be different from any other invention. This limits what an AI tool can generate but does not entirely prevent it from being able to help with the writing process. Patent-specific AI tools such as Patent Bots¹⁶ act as patent-specific editing tools like spell check. Patent Bots checks a patent to ensure it has the patent-specific syntax required by the United States Patent and Trademark Office (USPTO). More powerful tools such as PatentPal¹⁷ and Powerpatent¹⁸ can take patent claims written by an attorney and rewrite them in plain English in paragraph format to be used in the body of the patent. Patent claims determine what is protectable by the patent, but they need to be fully described in the body of the patent application. By putting the claims in plain English, these AI tools give you a base to expand upon, shortening the writing process. These tools require an attorney to write arguably the most difficult and important part of the patent, the patent claims, and this is likely to be the case for the foreseeable future.

Unlike the documents found at levels one and two, these complex documents require the highest level of judgment, empathy, creativity, and adaptability. First, these documents require exact language. For example, when writing a patent, it is vital that an attorney determine which precise language is proper since using the wrong word could either prevent the client from getting the patent in the first place or force them to go through a potentially costly legal battle. Second, when meeting with a client about a patent, the client may not even know what the most novel aspect is. Hence, there needs to be a high level of empathy to see everything from the client’s perspective to get all the necessary information so that the best patent can be written. Third, like judgment, there needs to be a high level of creativity when writing these documents. Often, for a patent to be approved, creative language must be used or words must be defined differently than their ordinary meaning. Finally, there needs to be a higher level of adaptability so that changes can be made or the claims rewritten so the patent can ultimately be approved.

AI hasn’t reached the point where it has a high enough level of these traits to generate these documents and is unlikely ever to generate them without significant amounts of attorney writing. Unlike levels one and two, where two traits are the primary limiting factors, all four are limiting factors at level three. Writing these documents takes human levels of judgment, empathy, creativity, and adaptability: the judgment and creativity to use the right words to get the patent approved, the empathy to write the patent that is best to protect the client’s invention, and the adaptability to overcome any hurdles at the USPTO preventing the patent from being approved. With current AI tools being able to do little more than reword the attorney’s own words, I believe it is unlikely that AI will reach a point anytime soon that it can write a patent even if an attorney supplies the claims.

In their current state, some of these AI tools can save you significant editing time, with others being able to save some time in your writing process, and in the future, this will likely still be the case. As specialty programs integrate ChatGPT or advance their own AI, lawyers will see greater time savings on their writing process, but it is unlikely that AI will completely generate such documents soon, if ever. If your practice involves writing patents, these patent-specific editing tools are something you should investigate and potentially adopt. Other patent writing tools, though, are something to investigate but are not at the stage where they are a must for your practice. This is likely to change as the technology advances, so it is something to keep an eye on if there are any considerable advancements.

Conclusion

Generative AI will inevitably become a fundamental tool in the legal field, increasing efficiency by automating much of the legal writing process. Continuing legal education, news articles, blogs, conferences, and even social media can help you stay up to date and are great places to find the best AI tools for your practice. AI is not something to fear but to embrace; AI tools will never replace human attorneys, so there is no reason to fight them. Remember that they are just that – tools. No matter what level your work product falls under, there is an AI tool that can help make you more efficient. Whether AI will write the document for you or assist with editing, every attorney should investigate adopting these tools into their workflow.