Widespread use of generative AI by lawyers and legal professionals will occur when AI tools can be applied to specialized and often private data bases.
There has been a lot of hype about ChatGPT of late, but according to various reports including one by Bob Ambrogi, the legal community’s reaction has been somewhat ho-hum. There are some reasons for that.
Use of ChatGBT By Legal Professionals
ChatGBT uses a public database–the internet–to derive its answers. At the risk of oversimplification, ChatGPT works by predicting what word will follow another phrase or word. Hence, using all publicly available information to make this prediction could result in some limited or specialized content being missed or misinterpreted. But this specialized content is often needed to answer legally related inquiries.
The result: the system hallucinates a response that may not be correct. (It’s been postulated that hallucination is the wrong term and that confabulation is a more accurate term). And for legal, incorrect but persuasive answers is a killer. Lawyers and legal professionals are also concerned about confidentiality. The concern is that whatever they put into ChatGPT could become publicly available.
These factors discourage lawyer use, particularly by those knowledgeable about how ChatGPT works. I also think the fact ChatGPT has been overhyped, like other legal tech that didn’t live up to grandiose claims, makes lawyers skeptical.
But widespread use of generative Ai (ChatGPT is generative AI) by lawyers may occur when the tools are applied to more limited and specialized databases. Applying generative AI to limited databases will reduce hallucinations. It will allow the system to focus not on everything but on what is particularly relevant to a query. And if generative AI is applied to a private database, it reduces the risk of a confidentiality breach and lends itself to more specialized uses.
The New Generative AI Tools
These concepts are what’s behind Casetext’s CoCounsel product, for example. CoCounsel applies generative AI to its own data and, by default, keeps all inquiries and responses siloed. By doing so, it can accurately do things lawyers, in particular, need to have done. Things like legal research, and ferreting out key documents. Doing first drafts of pleadings and discovery and even initial deposition preparation.
And recently, Disco, a prominent e-discovery provider, announced a generative AI product called Cecilia. Disco claims that Cecilia can significantly speed the e-discovery process. Cecilia is designed to allow lawyers to ask questions and receive answers with specific citations to supporting evidence drawn only from the user’s private documents.
A typical e-discovery work process, for example, consists of formulating a question about the evidence in a case, crafting bullion searches to try to identify documents related to that question, reviewing the documents obtained to find evidence, and then answering the question based on the evidence.
Generative AI platforms like Cecilia may be able to do much of this. A lawyer or legal professional can ask a question and, within seconds, get an AI-generated direct answer with the relevant documents identified. These answers could be the starting point for further exploration and document review. As Cecilia gets more sophisticated, legal professionals might be able to use it for such things as identifying gaps in production sets. Or finding documents that support a desired finding under jury instructions.
Using specialized data sets will also enable AI tools to do jobs that lawyers need to have done quickly and efficiently
Generative AI tools that use more limited data sets to respond to inquiries get a better result and will also enable AI to efficiently do tasks that lawyers need to have done.
AI tools will also allow the lead lawyer on the case—the lawyer that knows the case, the strategy, and the client the best–to formulate the actual queries. These are the lawyers that know what they are looking for and why and can pose the most effective questions. With all generative AI products, the better and more focused the question, the better the response. As Kevin Smith, Disco’s Chief Product Officer put it when I talked to him about Cecilia, “the more senior attorney knows how to express the question better than most”.
Law firms and in-house legal will also ultimately use generative AI inquiries on their own internal data as well. Many large law firms still need help accessing past efforts, for example. Labeling and characterizing past efforts takes time no one has and can’t be billed. But there is a wealth of knowledge and data within those past efforts. With generative AI, accessing past efforts may be just a question away. And to the extent a firm has a cultural way of doing things (which many firms swear they do), the response will reflect that culture.
AI tools will do alter the landscape much the like how word processing altered how lawyers work and who does what
But make no mistake. These generative AI tools will replace ways of working and work processes much the like how word processing altered how lawyers work and who does what. Before word processing, lawyers verbally provided information to secretaries. The secretaries took down the information and converted it to text. They then gave the results back to the lawyer for handwritten revisions. It was a time consuming and labor intensive process.
Today, most lawyers type their own first drafts and do edits themselves. This fundamentally changed the law office and its dynamics. And sadly, it reduced the need for legal secretaries. Generative AI will do the same thing but perhaps on a broader scale. And it will impact the need not only those who work for lawyers but lawyers themselves.
Great lawyers want great technology
The real question is not whether lawyers should use the public ChatGPT and, if so, how. The question is how generative AI models can be made to use private data and provide lawyers with the real assistance they need. The lawyers, law firms, and vendors that see the potential of these tools will be the real generative winners in the future.
Smith put it best: “Great lawyers want great technology.” These lawyers will drive how generative AI will impact legal.