Agentic AI 101: Decoding the Latest AI Buzzword for Legal Professionals

Will Gaus, chief knowledge management and innovation officer of Troutman Pepper Locke, and Jason Lichter, principal at Troutman eMerge, were quoted in the June 3, 2025 Legaltech News article, “Agentic AI 101: Decoding the Latest AI Buzzword for Legal Professionals.”

“The most succinct definition I’ve seen for agentic AI is an artificial intelligence system that can accomplish a specific goal with limited supervision,” said Jason Lichter, principal at Troutman Pepper Locke’s e-discovery subsidiary Troutman eMerge.

AI agents “can make decisions about what to do next, use various tools that we give them together, things like information, and then iterate on its work until it achieves its desired outcome,” added Will Gaus, Troutman Pepper Locke’s chief knowledge and information officer.

…

“If you’re building an effective agentic AI system or solution, the paralegals and attorneys should have no idea what’s happening behind the scenes,” Gaus said. “The best, the most innovative products seem almost too simple and too easy to use.”

…

“I think that we have to be mindful, especially given the token costs and other charges and the economics of using these models and multi-agent orchestration, that we don’t have to boil the ocean of agents and orchestration in every instance,” Lichter said.

…

“You can ask generative AI, any model, the same question 10 times, and you’re going to get potentially 10 different answers that could potentially be all correct,” Gaus said. “When you extend the model to things like multiple agents with multiple sets of tools and access to all sorts of different types of data sources, it’s tougher to declare ‘yes, this is an effective tool.'”

Still, Lichter noted that the agentic AI also offered the potential for individual users to create workflows adapted to their own preferences and working styles.

“There is the potential for enterprise scale systems and deployments you need to put in all of that upfront legwork and have all the requisite experts involved [for], but for kind of smaller scale agentic AI solutions, there is that DIY element that in certain circumstances may be suitable or may be appropriate.”