At the Thomson Reuters Institute's recent Generative AI & Emerging Technology Forum, the discussion turned to how this innovative technology can help humans on a daily basis
NEW YORK — At times, it may seem like generative artificial intelligence (AI) is a comic book superhero. The technology can swoop in faster than a speeding bullet, magically create content out of thin air, and save a department in distress from oppressive deadlines.
However, Bennett Borden, Partner and Chief Data Scientist at DLA Piper, warns that this superhero still needs some help. “Think more Iron Man than Terminator,” Borden said, adding that generative AI’s real strength is to “combine legal acumen with generative AI to give a new product or service.”
Borden and other experts in the legal and tax industries converged last week on the Thomson Reuters Institute’s Generative AI and Emerging Technology Forum to explain how generative AI’s superpowers translate from the realm of Krypton to solve today’s real-world problems. After all, the panel noted, generative AI needs a villain to fight against, and any application of the technology needs a real goal that it’s trying to solve.
“Why are we doing this?” asked Jeffrey Willinger, Digital Experience Director at advisory and accounting firm Withum. “We’re trying to solve a business problem.”
Another panelist, Peter Geovanes, Chief Innovation & AI Officer at McGuireWoods, called generative AI a “mad scientist” that can supercharge both the practice and business of law. On the legal practice side, generative AI could be used to upload an opposing party’s motions to find inconsistencies in statements or quickly summarize transcripts. On the business of law side, generative AI could scan public news reports and data to predict emerging trends in litigation, such as new regulatory issues in the energy sector for a vehicle manufacturer.
In both cases, however, it’s the human input that is pointing the AI towards a specific outcome that allows the technology to do superhuman things. “I think that changes the whole dynamic of the law firm and the ability to provide value to those clients and prospects,” Geovanes said.
Panelist Arash Nourk, Director of AI Experience Design at Thomson Reuters Labs, agreed and added that generative AI in its current form is not truly “intelligent” — it cannot, for example, conduct critical thinking or form strategy. The technology is only as good as the data it’s trained on, Nourk explained, and the data that an organization collects can also be a reflection of the organization itself.
“As we are training our models based on our firms — or any professional services organization — it becomes a continuation of our brand, of the character of our firms and of the signature that we have,” he said.
Borden added that this does not necessarily always mean the character of the firm today, but can also be a reflection of where the firm wants to go. For example, his team at DLA Piper built horizon-scanning tools that look to pinpoint issues for clients before they arise, allowing the firm’s lawyers to not only aid their clients today with new tools in their belt, but help them identify problems they didn’t even know they had. “It’s what new thing can we create with this, not just how does it help us do what we do today,” he said.
This also works the other way, Geovanes said, noting that working with a client can also provide opportunities that the firm itself cannot see. “That’s where we get these really compelling use cases” for generative AI, which in turn provide a competitive advantage for the firm’s attorneys.
Of course there are risks to generative AI, but the panelists all encouraged experimentation so a firm can learn what use cases make sense for the tool. Willinger noted that “you don’t want to have too many guardrails” that can stifle innovation — but Borden added that generative AI’s output currently is like a third-year associate in a law firm. “It’s nice, but you’d never want to hand it to a client,” he said.
Nourk also explained that humans using generative AI “have to understand how the model came about the answer that it did” when explaining its outputs, and many generative AI tools can now feature modified designs where the AI can be asked how a decision was made.
Indeed, it takes some training and conditioning for humans to figure out how to work best with these technology superheroes. “People are used to giving it a very short prompt, but what we’re finding is that you need to give the AI a backstory and a motivation, like you’re a director,” Geovanes said. “‘I’m the actor, what’s my motivation here?’ You’re angry, you’re desperate!”
But Geovanes noted that this does work. “That really is the difference between getting a world-class answer and something that is very broad.”
As people become more used to giving these prompts, however, generative AI’s powers could ultimately become more crucial and in-grained in the day-to-day lives of legal and tax professionals. And while it may create an arms race as people explore what these new superpowers can do, the panel said, it’s not a race that those in professional services can afford to ignore.
“Firms that don’t lean into this are like dinosaurs the day before the meteor hit,” Borden said.