How AI can become a ‘thought partner’ to compliance programs

In a recent webinar, industry leaders discussed applications, challenges, and strategies for implementing AI in compliance programs.

The webinar, title ““Harnessing AI and Analytics for Compliance Excellence: Strategies to Navigate 2025’s Regulatory Landscape,” produced by RLDatix, began with Gus Papandrikos, global head of compliance assurance, at Daiichi Sankyo, detailing the company’s initial start in AI adoption, focusing on “low-hanging fruit” such as analyzing public data like open payments and other data sources like Concur.

“We’re mining that information so it gives us visibility and access to payments to HCPs, other customer types of activity of our employees as it relates to their spend habits,” he said. To analyze the information, the company built a tool and developed key performance indicators that, if violated, are indicative of potential wrongdoing, Papandrikos further noted.

The AI tool doesn’t provide definitive answers but acts as an effective guide, he continued. “They’re really smoke signals,” Papandrikos said, emphasizing that the tool is “not going to unequivocally tell you something is right or wrong, but it’s going to focus your attention to a particular activity versus another activity.”

This approach has improved efficiency, he noted. In the past, he said teams spent “equal parts in time finding issues and then remediating the issues.” Now, with AI’s assistance, they are “spending almost 100% of the time remediating potential issues,” allowing for a focus on “higher risk, higher activity.”

Preah Dalton, compliance officer at Merz Therapeutics, described her small team at being at the forefront of AI. She highlighted their initial success by partnering with their data team to quickly analyze existing available data.

Beyond data, one area she said that they are engaging AI in is training. “There are a number of training platforms that are incorporating AI within those solutions where you could say, upload a policy, or upload some subset of information and then tell the tool the kinds of things you wanted to do, like a 15-minute training on this policy,” she said.

Dalton’s team is also proactively engaging with their IT group’s new AI team, becoming an early adopter.

AI as a “thought partner”

The panelists stressed that AI systems require ongoing human guidance to remain effective.

“The tool only knows what you teach it,” Dalton said. “Data is going to continuously evolve. The data that we have today is not the data we were getting five years ago. So there’s going to be constant refinement of tools.”

Audrey DeGuarde, vice president, compliance operations at RLDatix Life Sciences, resonated the concept of AI as a “thought partner,” highlighting its role as an accelerator that speeds up processes without waning the need for human expertise.

“It does not eliminate people,” Papandrikos said, agreeing with DeGuarde. “That’s one thing that, at least in our instance, in the way we designed and built it, it was never intended to replace people, and I don’t think AI will ever replace people because somebody has to check.”

Dalton echoed this sentiment by describing AI as a “great tool” that “could be as powerful as we want it to be, but it’s not like we turn a lever or turn a button on and boom, we’ve got all this great data, and it’s infallible.

“I learned recently that if you’re going to get to failure, you might as well get to the failure fastest, right? Because that’ll allow you to identify the holes within this data,” she added.

According to Papandrikos, “There’s certain things that it can do to eliminate mundane tasks, but you’re still going to need human capital to be able to review and make determinations.”

An important aspect of the human-AI partnership is establishing trust in the tech’s outputs. Papandrikos described the validation processes to ensure their AI systems were reliable before full implementation.

“We took the better part of the year playing with the data,” he explained. “We had a good understanding of what the data should tell us. So we did this over and over again ... we had faith that when we got a report out of it and it said, ‘Here’s your top 10 offenders,’ that they were legitimate top 10 offenders.”

“Dos and don’ts”

The webinar further explored the “dos and don’ts” for organizations embarking on their AI journey.

Papandrikos noted the importance of embracing AI, advising companies to “find a good partner” to guide them through the process. He stressed, “Take your time and think about what your biggest pain points are and what are the biggest challenges that you want to overcome and find solutions for.”

Papandrikos also cautioned against going into AI adoption blind, urging users to “learn about it. Understand it. Understand what requires to be executed well.”

Dalton encourages not to “let intimidation sort of be the barrier to entry,” noting the available resources for learning and growth.

“There’s always going to be someone way more advanced, and there’s going to be people that haven’t even heard the term,” she said. “Just owning where you are today, knowing that you’re progressing further is OK and that should be embraced.”

DepositPhotos/Scrudje