Annesha White and Fredrick Manasseh talk AI risks | AMCP Nexus 2026

AI is helping managed care pharmacy teams work more efficiently. However, when the technology influences whether patients can get their medications, human oversight remains essential, according to two pharmacists who presented this week at AMCP Nexus 2026, the Academy of Managed Care Pharmacy's annual meeting.

Annesha White, Pharm.D., M.S., Ph.D., senior associate dean and professor at the University of North Texas Health College of Pharmacy and Health Professions, and Fredrick Manasseh, Pharm.D., MBA, BCGP, senior vice president of pharmaceutical strategies at USI Insurance Services, co-presented a session on AI in managed care pharmacy at the meeting. In this first of two interviews with Managed Healthcare Executive, they discuss where AI is helping today and the risks payers and employers should watch for when AI tools shape decisions such as prior authorization.

Where AI is helping

White said AI is improving efficiency in areas such as pipeline forecasting and prior authorization review. Because AI is moving so quickly, she said their session stresses governance, human oversight and new workforce skills.

“We definitely want to make sure as healthcare professionals that we're prepared to use it responsibly,” she added.

Manasseh said there is no single preferred AI platform. The key is finding one that fits an organization's needs, with guardrails and a framework to vet whether a tool suits the task.

Risks in prior authorization

Prior authorization involves many steps, and White said AI can help streamline them. Although, when a tool affects access to critical medications, rigorous validation and clinical oversight matter most, she said.

“These tools are helpful, but we have to have these safeguards in place as we use them more and more,” she added.

Manasseh pointed to studies in which decisions made with AI tools were later overturned. A tool's output is only as good as its input, he said, so the patient population behind the data must be well represented.

“Sometimes you have to use clinical judgment as well,” he said.

White noted that their session addressed algorithmic bias and automation bias, adding that as people grow comfortable with a tool, they may check its work less often and let it take the lead, and errors can increase.

“We know that bad data in means bad data out,” she said, adding that organizations need to keep checking whether the data feeding a tool leans toward one population or has gaps.

Effort up front pays off

Asked whether oversight adds extra work, both said a strong governance process can save time in the long run. White said healthcare professionals should be able to explain what an AI tool is doing behind the scenes and confidently stand behind its recommendations.

“The heavier lift is in the beginning,” Manasseh said, adding that ongoing validation and oversight are critical.