In the TV Series Suits, Mike Ross possessed a photographic memory and exceptional intelligence that allowed him to absorb vast amounts of legal knowledge and recall it with remarkable accuracy. His quick thinking, analytical skills, ability to connect disparate pieces of information, prepare legal briefs, and pick through contracts, made him a valuable asset to Harvey. Thanks to GenAI, you can have your very own Mike Ross to assist you with your cases!
Roman statesman and lawyer Cicero first said, “The good of the people is the chief law.” Given the rising cost of legal services, Cicero would welcome Generative AI with open arms because it would, first and foremost, serve the larger good by reducing the cost of legal services. One research paper, Better Call GPT, Comparing Large Language Models Against Lawyers, published in January 2024, found that LLMs could reduce the cost of some legal processes by as much as 99.97 percent.
However, cost is merely one of three reasons legal eagles will root for Generative AI. Legal work is frustratingly slow, painstaking, and meticulous. Speed and accuracy are as important as cost. Mike Ross, for the moment, is a fictional character. Generative AI is the answer to all three challenges (effort, speed, and accuracy). And with that, your Honor, we have a watertight case for Generative AI.
The research paper, which used the critically acclaimed legal crime drama Better Call Saul to title itself lightly, showed that LLMs easily outperform junior lawyers and Legal Process Outsourcers (LPO or Contract Reviewers).
The study set out to examine how well LLMs could determine and locate legal issues in contracts. For issue determination, GPT4-1106 and LPO emerged as the top performers with an F-score (a metric used in classification tasks) of 0.87, while junior lawyers achieved an F-score of 0.86. Generative AI did not perform as well in localization accuracy, but it is a matter of time before models are fine-tuned to outperform humans.
Next, the study compared the process of legal contract reviews. Senior lawyers took an average of 43.46 minutes to review a document, LPO took 201 minutes, and junior lawyers took 56.17 minutes. By comparison, GPT-1106 took 4.7 minutes.
Finally, the per-minute cost of GPT-1106 was $0.25, compared to $75.92 for senior lawyers, $74.26 for junior lawyers, and $36.85 for LPO.
The outcomes of the study will leave no room for doubt about the jury’s verdict.
Gartner forecasts that over the next two to five years, LLMs will improve the productivity of legal departments by at least 10 to 20 percent more than any transformative technology. The Stanford Law School (SLS) says that the LegalTech market around research and analysis, document review and drafting, and litigation in the US alone would have been around
An estimated ~$13 billion in 2023. By putting the two together, we know that research and analysis, document review and drafting, and litigation will become the first few areas where Generative AI will have a significant impact. LegalTech will then evolve, says the SLS, to address specialized areas such as patent and IP, immigration, insurance, and regulatory compliance.
Law firms already acquiring Generative AI tools are experimenting with non-billable back-office functions. This is low-value work, and most firms will quickly extend their experimentation to billable work to unlock significant savings while adding value that customers will appreciate. Finally, after the five-year horizon that Gartner provides, firms will begin automating complex legal services and day-to-day legal tasks. A complete transformation of the legal industry will have been completed within the next decade.
The transformation is already underway. John Snow Labs, for example, has transformed how VakilSearch, India’s largest legal service platform, manages document classification, entity recognition, and legal relationship extraction from scanned legal and tax forms using Natural Language Processing (NLP). The implications are astonishing. The VakilSearch team can talk to its library of information, using conversation to extract information and arrive at decisions.
The results of leveraging Generative AI solutions at VakilSearch are significant: The document classification pipeline outperformed the comparable pipelines by 9 percent and attained 96 percent accuracy. Identity card information extraction achieved 87 percent accuracy, while the mail address resolution pipeline achieved 89 percent accuracy. Every process—from identifying companies and their aliases to named entity recognition and their relationships to extracting signers and their roles, etc.—is scalable.
The legal profession has been steadily adopting technology. Some commonly used practices now include accessing information remotely, using technology to help create/complete legal templates, digitizing paper documents for storage and retrieval, and virtual appearances in court.
Also read: 5 Key Elements to Unlock the Power of Generative AI
An entire generation of young lawyers-in-the-making is being weaned on AR, VR, online collaboration, NLP-based applications, and other advanced technologies. When they join the legal mainstream, they will want to continue to use these technologies—blockchain for smart contracts, data-analytics for risk assessment, predictive analytics to anticipate opposition arguments, and make them more efficient and productive. Generative AI will be an intrinsic part of their toolkit, acting as both a cost-saving legal minion and a supreme strategist in the courtroom who is backed by well-reasoned arguments.
Talk To Our Experts to Redefine Your Legal Practice with GenAI
References
https://www.gartner.com/en/legal-compliance/trends/ai-in-legal-industry
https://www.johnsnowlabs.com/legal-nlp/
https://law.stanford.edu/2024/03/07/a-view-into-the-generative-ai-legal-landscape-2024/
