Of Justin Fox

The ability of artificial intelligence to transform the legal profession was evident even before the appearance of the Openai Chatgpt in 2022. However, several lawyers and judges have learned with great embarrassment that chatgpt and other LLMs are not limited to case -law when they formulate legal arguments, often inventing references and excerpts.

“There is an extremely clear use in Law, except that these models, fundamental models, are designed around efficiency, not accuracy,” said Sean Fitzpatrick, Managing Director of Lexisnexis North America, the United Kingdom and Irish services. “In fact, they do not deal with the truth. They deal with chances and then build convincing arguments. “

Lexisnexis, owned by the London Group RELX PLC (formerly Reed Elsevier), hopes that a limited, more “controlled” version of artificial intelligence, based on models of Openai, Anthropic and others, can provide significant productive benefits to its customers without “constructed data”. The same is true of the competitor Thomson Reuters Legal Services, owned by Canadian Thomson Reuters Corp.

Artificial intelligence legal services are not unmistakable. A recent study by Stanford University scientists found that, although the products of the two companies were clearly better in avoiding “constructed answers” than Chatgpt, they were still giving many incorrect or incomplete answers, with Lexis+ AI’s Lexisnexis, Law ai by Thomson Reuters. However, they seem to be successful in the market: Lexisnexis, which has been the latest growing part of RELX over the last five years, has now become the fastest growing, with so -called underlying revenue (without acquisitions, sales and concomitant effects) to increase by 9% in the first half of the first year. Reuters Legal reported an organic increase of 8% in the same period.

In general, legal companies appear as “skies” in artificial intelligence. A recent study by Thomson Reuters has shown that they are far behind the business sector in organizational investment in artificial intelligence. It is therefore tempting for the lawyers to imagine a future in which they will gradually adopt the new technology to become more productive, but there are no fundamental changes in their way of working. This would be good news for lawyers, RELX and Thomson Reuters – and good news for customers, who would get better for their money.

This is reminiscent of the first of the two possible scenarios for the future described by Richard and Daniel Saskid a decade ago in their much -talked -about book The Future of the Professions: How Technology Will Transform the Work of Human Experts. In this “reassuringly familiar” scenario, they wrote, “professionals continue to work as they do since the mid -19th century, but they are largely standardized and systematizing their daily activities.” The second possible future, which Saskid considered more likely, was the one where “the introduction of a wide range of more and more capable systems would displace, in various ways, much of the work of traditional professionals” and “our professions will be gradually dissolved”.

Which of the two applies to lawyers? This will partially depend on whether large linguistic models (LLMS) will ever become reliable enough to completely replace them, a possibility that remains extremely doubtful. But even the automation of repetitive duties has the potential to overturn the finances of the legal profession.

This peculiarity is associated with billing practice per hour. In law firms, “productivity” means how many hours their lawyers invoice. Elsewhere it has to do with how much work is completed at a given time. Lawyers stand out, even compared to other professional services providers, such as accountants, management advisers and technology consultants, about how little their production has increased in recent decades.

If you also calculate equipment and technology expenditure – which in law firms are mainly for Informatics, then the productivity of the US Legal Services industry is currently significantly lower than it was in the 1980s, according to official statistics.

This impressive finding is partly explained by the way technological investments appear in statistics: they are considered a large “price fall”. For example, a dollar spent today on technology is considered almost equivalent to $ 15 in 1988. So even if spending stays fixed at current prices, it seems as if they have grown spectacularly. However, other high -level sectors have managed to increase their productivity, which shows how outdated the model of law firms remains.

And maybe it remains that way for some time. “I hear about the ‘death’ of debt for a long time for a long time, and it has not come,” Fitzpatrick said. In addition, he noted, the immediate impact of artificial intelligence seems to be the increase in hourly wages, as it allows lawyers to complete more work in less time.

In a recent survey of large law firms, Robert J. Kutour of the Center for the Legal Profession of the Harvard Law School has found similar consensus: the increased use of artificial intelligence brings improvements to the quality for which customers are willing to pay higher pay.

Maybe not for long. Bruce McGuen and Janet Stanton, who run the Adam Smith Counseling Company, ESQ, estimate that the reduction in charges will be so great that most law firms will not be able to withstand it. “This will kneel the traditional billing model per hour,” says McGuen. “Pricing with a steady remuneration is the future.” The problem, Stanton adds, is that law firms have a weak financial organization and do not know how to cost their services properly. Thus, the most likely result, according to McGuen, will be “a large liquidation of traditional law firms”.

Will we get rid of lawyers completely? A central theme in Saskid’s book is that Status Quo in Professional Services is not acceptable, mainly because “most people and organizations cannot afford top professionals”. If citizens could simply introduce their legal questions to the chatgpt and receive reliable answers, wouldn’t it be better for society?

Maybe, but this is a very big ‘if’. In a series of studies with various authors, Jonathan H. Choi of the Washington University Law School is testing the performance of LLMS in different legal duties. They are doing very well with simple issues that have clear answers, but they find it difficult with complexity and ambiguity. In his latest study, entitled “Large Language Models Arereliable Judges”, Choe describes how minor differences in the wording of questions lead to huge deviations to the answers to legal issues and how different models also give very different answers. Choi argues that these problems can be inherent in LLMS technology and, therefore, not something that will be corrected soon.

So the lawyers will probably be with us for a while. All that is not clear is exactly how they will be paid.