At AMCP Nexus 2025, pharmacy and health plan leaders described how artificial intelligence is moving from slideware to shop floor—promising faster, cleaner workflows while raising urgent questions about governance, bias, and clinical judgment. The session, led by longtime pharmacy and PBM executive Harry Travis, BS Pharm, MBA, mixed live audience polling, an artificial intelligence (AI) crash course, and a candid discussion with Pejman Ahmadi, PharmD of Centene and Vinay Panchal, PharmD, MBA, BCPS of Develop Health.

To frame the stakes, Travis traced AI’s breakneck progress and its concentrated development among a handful of frontier labs, noting the technology’s outsized energy demands and economic implications. He urged organizations to think in layers: first, augment the workforce with secure, everyday assistants; next, automate clearly defined workflows under human oversight; and, ultimately, reimagine services and business models that only become possible with AI. The strategic fork, he said, is simple: you can use AI to cut costs and steps—or to create new value.

Ahmadi described Centene’s enterprise approach. The company has rolled out Microsoft Copilot in a private, secured environment and formalized training through “Centene University,” where teams get protected time each week to learn and practice. Today’s use centers on content generation, meeting support, and member/provider self-service through bots, with letters and other communications automated behind the scenes. For prior authorization, Centene is methodically testing vendors and internal builds, comparing agent outputs against human gold standards before ceding any real clinical or coverage decisions. A cross-functional AI council governs build-versus-buy choices, model risk, and security.

Panchal offered a complementary picture from the clinic-facing side. Develop Health integrates with the electronic health record (EHR) to read a patient’s chart, map answers to payer question sets, cite where each answer came from, and display confidence levels along with a signal of likely approval based on similar cases. Meeting transcripts and action items flow automatically into a central hub (Pylon) that also pulls in email and Slack, giving leaders a live view of each customer and eliminating routine “what did we decide?” churn. Far from reducing human touch, Panchal said, these tools “cut the fat” so teams can concentrate on the few decisions that actually move outcomes.

Both panelists drew a bright line around clinical judgment. Bots are well suited to high-volume, rules-based tasks such as benefit verification or gathering evidence, but peer-to-peer reviews should remain clinician-to-clinician. In ambiguous cases—or where the stakes are high—humans must review even if agents perform the first pass. Quality management starts with head-to-head comparisons of human and AI outputs; misses are examined to separate true gaps from helpful simplifications that reduce noise. Bias is treated like any other performance risk: find root causes, monitor results, retrain, and diversify data. Security depends on private tenants, data minimization, and narrowed scopes of use; AI doesn’t excuse weaker controls, it demands stronger ones.

Equity questions centered on whether AI will improve access for underserved populations. Ahmadi and Panchal said the largest benefits come from removing administrative friction across lines of business, which can shorten time-to-therapy and clarify communications for everyone. If anchored in ethics—right drug, right time, right cost—AI can improve experience and quality while curbing waste and freeing resources that matter most in Medicaid-heavy populations.

Culturally, Develop Health’s always-on meeting capture hasn’t chilled candor; transparency improves follow-through, and professionalism sets the tone. Centene does not broadly record meetings but encourages an open-door environment to surface issues quickly. Both leaders stressed the importance of normalizing a learning curve: early drafts may be imperfect, but accuracy climbs with use, better prompts, and structured feedback.

Travis closed on a sober note. The same forces accelerating AI—capital, compute, and competitive pressure—also heighten the need for real governance. Managed care, he argued, must be techno-realist: enthusiastic about productivity and patient benefit, skeptical about risks, and unwavering about alignment with clinical and ethical goals.

Bottom Line

AI in managed care is past the novelty phase. Organizations that start with secure augmentation, graduate to carefully validated automation, and invest in governance will capture early wins without surrendering the clinician’s role—or the member’s trust.

Reference

Travis H, Ahmadi P, Panchal V. Unleashing the Power of AI: Transforming Healthcare and Pharmacy. Presented at: AMCP Nexus 2025; October 27-30; National Harbor, MD.