All In on AI for Efficiency, But Not For Clinical Judgment| AMCP Nexus 2025

Five years from now, most healthcare organizations will be using artificial intelligence (AI) to decrease costs and provide better outcomes for patients, said panelists at a session of the Academy of Managed Care Pharmacy (AMCP) Nexus. They were optimistic that any savings on the operational side will not result in layoffs but instead be allocated toward having human beings on the front lines reviewing prior authorizations and doing quality checks.

“AI will still be used as a tool to assist us to get more efficient and faster for the services that we need to do for our members and provide,” Pejman Ahmadi, Pharm.D., director, pharmacy program management at Centene, during the panel discussion, told the AMCP Nexus audience. “The operation side will have a limited number of people, where clinicians and pharmacists can specialize in specialty drugs. Humans will still be needed for this to make sure things are right, because we have humans that are going to be taking a drug, not a robot.”

The AMCP Nexus meeting is being held at the Gaylord National Resort and Convention Center in National Harbor, Maryland, outside of Washington.

Ahmadi said Centene, a Medicaid and Medicare managed care company, is currently using a subscription-based Microsoft Copilot across the organization for communications, presentations, and education and training. “It’s the tool of the future,” he said. “We recommend and encourage it as a tool. We just want our teams to use our private accounts and not to use PHI [protected health information].”

Centene has also invested in AI technologies for its call center as the first step for member communication. But Ahmadi said the company is still researching how to use AI for prior authorization. “We’re still developing and seeing what we can do, because there are obviously risks. There are quality checks that need to happen. But we’re excited about what it can be.”

The five largest companies in the AI sector in the U.S. are spending $375 billion this year on developing the technology and an estimated $500 billion next year, said moderator Harry Travis, MBA, president of The Travis Group. “To put that in context, that’s more than we spent for the Human Genome Project and the Apollo program, adjusted for inflation,” he said.

Travis was optimistic that many steps along in the process from prescription to the patient getting the medication can be automated with AI, including benefit verification, prior authorization, copay assistance and final adjudication. “Not to oversimplify it, but it is a set of questions that need to be answered to move the prescription or move the case down the line.”

Dozens of companies are developing tools for each one of these steps. One company, Develop Health, has developed an AI-powered platform to automate benefit verification and prior authorization.

His company's AI tools integrate into electronic health records, said panelist Vinay Panchal, Pharm.D., MBA, head of pharmacy operations and PBM relations at Develop Health. For prior authorizations, “the artificial intelligence scans the whole record and populates the answers that they think are appropriate for the questions asked. It provides the rationale and why it's appropriate where they found it and whether the confidence level around it is high, medium or low.”

He said this allows the medical assistant at the physician’s office more time to do other work, instead of hunting down specific information or having to talk to the doctor about specific lab information.

Ahmadi cautioned that before implementing this technology, Centene wants to compare the outcomes from the AI with what people can do. “AI in general will make you know things faster, as far as being able to get drugs to members quicker, get their notifications quicker and make that experience of interactions with us better,” he said.

But both Ahmadi and Panchal said AI is really good for the exchange of information, but it cannot provide clinical judgment. Both said AI should not be used for peer-to-peer consults with clinicians. “It just knows black or white; it does not know the gray in between,” Ahmadi said. “Humans have to fill that gray in between.

Quality assurance emerged as a concern among people in the audience. Panchal described Develop Health’s validation process, which includes confidence scoring, clinical review of medium-confidence outputs, and continuous model refinement. “It has to be a combination of technology and real humans who understand the business process to be able to get to proper quality,” he said.