About | HeinOnline Law Journal Library | HeinOnline Law Journal Library | HeinOnline



5 Stetson Bus. L. Rev. [i] (2025-2026)

handle is hein.journals/setnbisr5 and id is 1 raw text is: 



STETSON BUSINESS LAW REVIEW


             VOLUME 5                               FALL   2025                             ISSUE 1


  AI  &  The  Law


  ARTICLES*


  Rethinking Generative AI's Performance on the Bar Exam and on Law School Exams
                                                                                       D. Benjamin Barros
                                                                                           Alexis M. Moore
                                                                                     Griffin C. Barros-King   1

         Inspired by a study that GPT4 performed well enough on the Uniform Bar Exam  to achieve a passing
         score, we gave GPT4 a property law essay exam and a set of property law multiple choice questions. GPT4
         did not perform well on either one. GPT4 displayed an inability to reliably identify legal issues in the
         essay exam  fact patterns without prompting, and answered  only 54 percent of the multiple-choice
         questions correctly. We argue that GPT4's relative success on the UBE was due to the structure of bar
         exam questions, which provide internal prompts that aided GPT4's performance. We therefore argue that
         GPT4's ability to pass the UBE indicates more about the structure of bar exam questions than it does
         about its ability to do fundamental legal academic tasks, such as issue spotting.

         In late August 2025, we ran the same property essay and multiple-choice questions through the recently
         released GPT5. GPT5 displayed a marked performance improvement  over GPT4. It was able to spot legal
         issues from fact patterns without prompting, though in one instance, it made a fundamental error the
         second time it answered a problem. It answered 94 percent of the multiple-choice questions correctly,
         comparing favorably to the 54 percent that GPT4 answered correctly.


  How  the Billable Hour Can Survive Generative Al
                                                                                           Jonah E. Perlin    42

         The legal profession is rapidly adopting generative artificial intelligence (Al) technology. As a result,
         there is an increasingly common assumption that lawyers will decrease their longstanding reliance on the
         billable hour. The logic underlying this conclusion seems intuitive: if Al reduces the time lawyers spend
         on certain tasks while eliminating others altogether, then the total number of billable hours will fall. That
         decline, in turn, could threaten the financial viability of hourly billing, despite its place as the dominant
         law firm compensation model over the past half century.

         This Article challenges the inevitability of this assumption. It argues instead that the billable hour not
         only has a viable path to survive in the age of Al, it may well thrive as a result.

         The  argument  proceeds in three parts. First, it explains why-despite  decades of criticism-the
         longstanding structural justifications for the hourly billing model are sufficiently sticky to help the hourly
         billing model remain relevant even in an AI-enhanced legal profession. Second, it introduces the CHRGE
         Equation (Compensation = Hours x Rate - Granted Reductions - Expenses), a simple but novel framework
         for understanding how law firms generate revenue under the billable hour model. Because hours worked
         are only one part of this equation, assessing the billable hour's long-term viability requires analyzing how
         generative Al is likely to affect each of the five variables. Third, by looking at how generative Al is likely
         to affect each of the CHRGE variables, the Article shows why generative Al is unlikely to drastically
         diminish billable time across the profession and, even if it does, how generative Al adoption may improve
         the other CHRGE  variables in ways that can offset this decrease in billable hours. To be clear, this Article
         is not a normative defense of the billable hour. Its contribution is instead descriptive and structural. It
         makes the case that, for both institutional and compensation-based reasons, it is far from inevitable that
         generative Al will render time-based billing obsolete.



* The first four articles of this issue were solicited through Stetson Business Law Review's 2025 Symposium, which
highlighted four main practice areas Al influences both academically and practically. We explored Al and its impact on
Employment  Law, Ethics, Intellectual Property, and In-House counsel. Our noteworthy speakers ranged from highly
esteemed professors to successful practicing attorneys. As our keynote, we were honored to have Mr. Barry Jones, AUSA, from
the Department of Justice's National Security and Cybercrime Division.