Choosing the Right AI Tool for Lawyers--Factors You Need to Consider
What is the difference between Specialized vs. General-Purpose Language Models --And which is right for your practice?
As artificial intelligence (AI) continues to transform the legal industry, lawyers, paralegals, and other legal professionals are increasingly turning to AI tools to streamline their work, enhance efficiency, and gain a competitive edge. Among the most popular AI tools are large language models (LLMs), which can assist with tasks such as legal research, document drafting, and even predicting case outcomes. However, with the variety of LLMs available, it can be daunting to decide which tool is best for your purposes. Should you opt for a specialized tool tailored to the legal field or will a general-purpose tool suffice? The answer—in typical lawyer fashion—is: it depends. How will you use the tool? Are you looking for a tool that can take on drafting extensive and complex documents? Are you looking for case outcome predictions? Or are you looking into a tool to automate routine tasks? What is your willingness to learn a new technology? What is your level of knowledge and skill with the law? And finally, what does your budget allow?
This article explores the factors to consider when making this decision and provides guidance on choosing the right LLM/AI tool for your legal needs.1
Do You Need A Specialized AI Tool?
From highly-funded tools such as IronClad, Harvey, and CaseText’s CoCounsel, to smaller start ups such as Leah or LawGeex, there are a number of AI tools that cater specifically to lawyers. There is a lot of buzz surrounding these specialized tools for the legal industry– but is that just hype or are there valid reasons for shelling out the extra money for a specialized tool?
Pros and Cons of Specialized AI Tools
Specialized AI tools are often trained on vast amounts of legal data, such as case law, statutes, and legal documents, enabling them to understand and generate content that adheres to legal language conventions and formatting requirements. But are they worth the extra cost? Some potential benefits of using a specialized AI tool include:
Enhanced Accuracy: Specialized tools may provide more accurate and relevant results for legal research and document drafting. They are usually trained on great amounts of legal cases, statutes, and court rules. They also have another advantage. All LLMs contain weights. The specialized LLMs have weights that are skewed toward legal language and formats. In other words they are tailored to understand legal terminology and context, and use that “understanding” in drafting documents, evaluating claims, and predicting outcomes. When LLM vendors train their models, they repeatedly train, test, and adjust their models. The concept of weights refers to the prioritization of some training content over others. Legal-specific LLMs have been tuned, weighted, and adjusted for the specific task of producing legal content. General purpose LLMs such as ChatGPT have been trained on legal documents, but were not specifically weighted to focus on producing high-quality legal output.
Improved Efficiency: Because of their training, fine-tuning, and the weights they are programed with, AI tools designed specifically for legal tasks can not only help streamline workflows and reduce the time spent on routine tasks, but also focus on higher-level, higher-value work.
Compliance with legal standards: Specialized AI tools may be better equipped to ensure that generated content complies with legal standards and regulations. This specialized training, however, can never take the place of a knowledgeable human critically considering the output you receive.
Some of the most notable cons of these specialized AI tools are:
Higher costs for a subscription.
Higher level of training costs and steeper learning curves compared to general-purpose tools. These tools will likely require a higher investment of time by users.
High costs of maintenance. Regardless of the tool you adopt, it is important to maintain a tight grasp on the outputs, watching for bias, outdated information, and avoiding hallucinations.
Less transparency.
Pros and Cons of General-Purpose AI Tools
General-purpose AI tools like Claude and ChatGPT are designed to be versatile and adaptable to a wide range of tasks and industries. While they may not be specifically tailored to the legal field, they can still offer significant benefits for legal professionals:
Cost-effectiveness: General-purpose tools are usually more affordable than specialized AI tools, making them accessible to a wider range of legal professionals and firms. Not only are subscriptions less expensive, but training costs and upkeep are also less expensive. This allows all sizes and types of firms to take advantage of the benefits of AI.
Flexibility: Tools like Claude and ChatGPT can assist with a variety of tasks, from drafting emails and summarizing documents to conducting general research and answering questions, providing a versatile solution for legal professionals.
Ease of use: General-purpose tools often have user-friendly interfaces and require less training to use effectively, making them accessible to legal professionals with varying levels of technical expertise.2 There is less of a time commitment upfront.
However, general-purpose AI tools have several downsides also:
They may not always provide the same level of accuracy and specificity as specialized tools when it comes to complex legal tasks.
These tools are programmed to be “everything to everyone.” Because of their generalized training, fine tuning, and weights, they might not be the best choice for more involved or specialized drafting legal drafting or analysis. So while they can summarize a deposition or draft a lease, I would not recommend using them to prepare in-depth merger documents, or a very nuanced motion.
Factors to Consider When Choosing an AI Tool
Regardless of whether you choose a specialized AI tool or a more generalized tool, it is imperative that you consider the following:
1. Data privacy and security concerns: Consider the data privacy and security features of the AI tools you are evaluating to ensure that they comply with relevant regulations and ethical standards.
2. Ongoing maintenance: Conduct routine testing to minimize biased output by your AI tools. Also monitor output, and carefully review documents to avoid falling into the trap of hallucinations. More on that in another article.
3. Understand the basis of how your AI model works: If you choose to use AI, and I recommend that you do, you open yourself and your firm up for questioning by clients and courts regarding why you chose a certain tool, what processes you will go through to make sure your use is ethical and valid, and whether you understand the basics on how your tool works. Understanding the basics of how your AI tool works will also give you greater credibility with clients when they ask you why you are using AI, or how you are using AI.
4. Have an AI company policy (and advise your clients to do the same): Having an AI policy is imperative in order to help limit liability. I cannot stress this enough. Using AI tools gives rise to new risks. An AI policy can help manage liability, build trust with clients, minimize bias, help protect data privacy, and lend credence to the position that you are acting reasonably, among other things.
Conclusion
Choosing between a specialized AI tool and a general-purpose tool ultimately depends on the specific needs, resources, and preferences of the legal professional or firm suing it. Carefully consider the factors outlined above for assistance in evaluating the available options. And if you have additional questions or concerns, respond to this Article with a question below, or jump over to our AI Forum for lawyers {LINK}, to discuss this with lawyers in the same or similar situations.
Finally, I would be remiss if I failed to mention that regardless of the type of AI tool you choose, it is essential for legal professionals to stay informed about the latest developments in AI and to use these tools responsibly and ethically, and also to have a basic understanding of the AI tool you choose. By doing so, you can harness the power of AI to drive innovation, efficiency, and success in your practice.
By the way, I hope you enjoy MidJourney’s (an AI image generator tool) attempt to create an illustration for this article as much as I do. I used 3 different image generating tools, with several iterations on each one. Apparently it is very difficult for image generators to create two separate laptops in one scene.
For clarity’s sake, let’s start with a couple of definitions. An LLM or large language model is what devours large amounts of text, and then uses that to power generative artificial intelligence tools. So LLM tools like ChatGPT, Claude, CoCounsel, and others, can be called AI tools (which I frequently do, for ease of reference), but are more specifically called LLM tools. While all LLM tools can be considered AI tools, not all AI tools are LLM tools. LLM tools specifically focus on processing and generating human language using large language models (GenAI), while AI tools encompass a broader range of techniques and applications beyond natural language processing. For example, if you are texting a friend and your phone suggests the next word, that is not generative AI, but it is an example of an AI tool. For purposes of this article, I will use “AI tool” even when I’m technically referring to an “LLM tool” because I think it sounds more understandable to my audience.
If you opt for a more generalized LLM tool, consider spending more time and/or resources in prompt engineering, which might help make up some of the difference in capabilities between a specialized AI tool and a generalized one. Here is an article that shows you how to use prompt engineering effectively.