Among the ELM Amplify 2023 sessions that generated the most interest among attendees was Empowering Legal Operations with Generative AI: Transforming the Future of Efficiency. This discussion, featuring Rose Brandolino, CTO and Client Technology Strategist at Microsoft, along with Vincent Venturella and Jeetu Gupta of ELM Solutions, helped participants understand the basics of generative AI (GAI). The presenters defined GAI and talked about how it can be a useful tool for legal operations professionals. They also discussed the inherent risks and challenges and how these can be addressed. Here’s a sampling of the insights shared during this informative session.
The advent of GAI
Artificial intelligence has been part of our lives for years. Since the creation of AI technology capable of beating humans at complex games more than two decades ago, progress has accelerated rapidly. More recently, out of that accelerated progress, GAI has arisen.
GAI interfaces are simple, with images or text being generated in response to a prompt, while behind the scenes, the software does its best to “guess” what the user wants. It isn’t necessarily likely to choose the “most true” answer, but the most likely answer. These applications are trained on an enormous amount of information, with a cost of up to $5 million to train a single model.
When asked by the presenters, there were many more attendees indicating that they have tried out GAI than those indicating they are actively using it. As it is still a new technology, this isn’t surprising. When users are ready, though, GAI can act as a smart and helpful partner, available 24x7, making your work more efficient and productive.
However, there are risks and challenges associated with using GAI. The most well-known of these is hallucinations. They occur because the AI doesn’t have actual insight into the truth of the output it is creating, rather it is simply trying to predict the next work. This leads to instances like the one cited during the session, in which ChatGPT recommended a legal AI tool that doesn’t actually exist.
General models of GAI – those trained on data of all types – can also suffer from the problem of degrading accuracy. These models can take in so much information that they are influenced even by information that is not correct. The more information they take in, the more errors can appear in their results due to incorrect data in their training. This is clearly an unacceptable risk in legal and many other contexts and is the reason why we are seeing more specific models emerging. These are purpose-built for particular subject areas, such as legal, and can avoid this type of degradation.
Wolters Kluwer is currently working on applications for GAI that will include mitigations for these risks. Among other measures, we have instituted data privacy and transparency measures and are building specific models within a tightly controlled sandbox.
GAI in the legal workplace
Today’s GAI technology can be put to use on both the practice and business sides of law. The speakers agreed that legal operations professionals should expect to see GAI used to save time, create efficiencies, and simplify administrative work. Like the AI applications that came before, it can remove some of the less interesting aspects of legal ops work, allowing people to focus on the more creative aspects of their jobs.
When used well, GAI can supercharge humans, summarizing and contextualizing large amounts of data quickly. However, it cannot replace people wholesale. Some work is still expected to be best executed by people. A Wolters Kluwer survey with Above the Law found:
- More than 80% of respondents agree that generative AI will create “transformative efficiencies” within legal research and other routine tasks.
- 62% believe it will separate successful from unsuccessful law firms within the next five years.
- Only 31% agree that generative AI will transform high-level legal work in job categories such as law firm partner or of counsel.
Generative AI has arrived in the legal function and is positioned to make positive contributions to the work of legal ops professionals. Humans will still be needed to review, guide, and shape the output, however, so it does not represent a replacement for people. Those who leverage the technology well will be able to expedite their work and focus their time and attention on the most value-added tasks for their organizations. In short, learning to effectively use GAI will help legal professionals become even more successful.