Managed Care Experts Call for Shift to Workflow-Integrated AI by 2030

Managed care organizations must move beyond early, low-risk artificial intelligence (AI) use to strategically governed, workflow-integrated systems by 2030, experts emphasized during a Wednesday morning session at the Academy of Managed Care Pharmacy (AMCP) 2026 meeting in Nashville, Tennessee.¹

Aligning Innovation With Growing System Complexity by 2030

Smit Patel, PharmD, of UpTheStack and The Economist Impact, opened the conversation, “Unlocking AI’s Potential in Managed Care: An Actionable Roadmap,” by emphasizing the growing complexity of managed care. He noted that employer health plan spending has risen approximately 10% from 2025 to 2026, member cost sharing has roughly doubled compared with 3 years ago, and prior authorization volume and complexity continue to increase. At the same time, CMS has introduced new interoperability and transparency requirements, adding further administrative burden.

Against this backdrop, AI capabilities are expanding rapidly, creating what he described as a “watershed moment,“ where operational and clinical complexity are rising alongside technological innovation. With that, he framed the session’s central question as how to align AI with these growing challenges by 2030.

Patel then outlined 4 domains that must be addressed before scaling AI. The first is strategy, or clearly defining which problems AI is meant to solve and which outcomes matter most. Next is governance, which requires organizations to establish accountability before scaling. As a result, he emphasized the importance of multidisciplinary governance committees that include clinicians, data scientists, and business leaders.

The third domain is workflow integration, with Patel stressing that AI must be embedded into real-world workflows where decisions are made, rather than deployed as isolated pilots. Scaling, he noted, should come last, only after the right strategy, governance, and workflows are in place.

AI Adoption Lags Behind Awareness

Before looking ahead, Candace Zheng, PharmD, MS, BCPS, of Cencora, reviewed 2024 market research from the FormularyDecisions platform on AI adoption among health care decision makers.² She reported that approximately 30% were very aware of AI applications and about 70% were somewhat aware.¹

However, experience lagged behind awareness, with only 53% reporting being somewhat experienced using AI tools. During this time, AI use was largely limited to lower-risk, augmentative tasks, such as basic automation, summarizing research papers, and supporting communication and administrative processes.

“To us, this suggests that many organizations understand the potential of AI but have not fully operationalized it,” Zheng said. “It's a good reminder that awareness alone does not equal readiness.”

She added that about 75% of respondents reported having access to AI tools, indicating that access is no longer the primary barrier. Instead, more pressing challenges include comfort, trust, security, and integration into real decision-making.

Breyanne Bannister, PharmD, MS, of Cencora, expanded on this shift, noting that by 2025, the conversation had moved from AI readiness to whether it can be trusted in high-impact access decisions. Market research showed that about 30% of stakeholders reported using AI in these contexts by 2025, with roughly half using public tools such as ChatGPT and Gemini and the other half using in-house solutions.

“I think this highlights that there are different approaches to being able to integrate AI,” Bannister said. “You really have to look at the pros and cons and understand what question or challenge you're trying to adjust and what really makes sense for your organization, specifically.”

Patel emphasized that organizations must distinguish between low- and high-risk situations for AI use. As mentioned, tools such as ChatGPT, Claude, and Gemini pose relatively low risk and are commonly used for general tasks, like drafting communications or answering simple questions. In contrast, higher-risk applications include clinical trial recruitment, clinical decision support, or prior authorization decisions. In these situations, he explained that AI errors can have significant consequences, as false positives or negatives can affect patient care and coverage decisions.

“So, risk levels grow higher and higher,” Patel said. “As risk levels grow, companies’ oversight is something that should grow.”

Looking ahead to 2026, Bannister noted that AI use is expected to expand into areas such as budget impact assessment, coverage scenario analysis, and broader decision support. However, key barriers remain, including limited trust in AI outputs, training gaps, and poor workflow integration.

“You can see that these are not just technical barriers but also organizational and operational barriers,” Bannister said. “Even the strongest AI tools struggle if users do not trust the outputs that they're getting or if they’re not going to fit into their existing systems and workflows. This really highlights why governance and change management are really important in the AI space.”

Patel described trust as the “million-dollar question” in AI adoption, emphasizing the need for human oversight as AI becomes more embedded in managed care processes.

Sheila M. Thomas, PharmD, of CVS Healthspire Life Sciences Solutions, added that building trust requires early and inclusive engagement, ensuring that all stakeholders, including patients and members, have a voice during AI design and implementation, as human judgment remains essential.

“The patient’s perspective has to be incorporated before the technology is built or deployed, not afterwards,” Thomas said. “For me, that means starting with a simple question: how will a patient experience this decision differently because AI is involved?”

Rounding out part 1 of the discussion, Bannister noted that despite ongoing challenges, decision makers reported seeing the continuous clear value in AI. The most frequently cited benefits included improved information aggregation, increased staff productivity, and reduced drug review time.

“These benefits really matter because they address a lot of the pain points, immediate pressures, and a lot of the managerial workflows,” she said. “While we addressed a lot of challenges and barriers today, and we are seeing some cautiousness when it comes to adoption, the interest in AI still remains strong.”

From AI Pilots to Decision-Grade Systems by 2030

In part 2 of the conversation, the experts looked ahead to 2030, emphasizing the need for organizations to move from ad hoc, lower-risk AI applications to “decision-grade” tools that are validated, continuously monitored for bias and hallucinations, and embedded directly into workflows to influence outcomes.

To support this shift, Patel outlined a 4-step implementation framework: plan, deploy, monitor, and scale. The planning phase focuses on establishing a safe, strategic AI supported by strong governance. It also includes assessing data readiness, integration capabilities, and aligning key stakeholders before the next step: deployment.

During deployment, AI is integrated into real-world workflows to support decision-making without replacing accountability. This phase requires validation by clinical and technical experts, standardizing inputs and outputs, and stakeholder training. Bias and accuracy audits, as well as pilot implementations with defined key performance indicators (KPIs), are also needed before broader expansion.

Next, the monitoring phase ensures ongoing accuracy, equity, and alignment with KPIs through continuous tracking, bias reviews, and governance audits. Doing so enables systems to evolve into audit-ready, learning infrastructures, according to Patel.

Lastly, the scaling phase transforms successful AI pilots into enterprise-wide capabilities, with AI becoming embedded infrastructure that supports closed-loop learning and more autonomous workflows.

Together, these steps align with the experts’ broader AI capability roadmap for managed care pharmacy, spanning 5 levels labeled “none," “initial," “ad hoc," “defined," and “mature.” With a timeline extending to 2030, the roadmap helps organizations assess their current position and identify the steps needed to reach fully integrated, enterprise-level AI.

“‘Defined’ is where more organizations are moving towards in 2026 and 2027, where they are starting to build those playbooks and have AI training [programs], governance committees, and actionable KPIs in place,” Patel concluded.

References

1. Patel S, Thomas S, Zheng C, Bannister B. Unlocking AI’s potential in managed care: an actionable roadmap. Presented at: AMCP 2026; April 13-16, 2026; Nashville, TN.
2. FormularyDecisions. Cencora. Accessed April 15, 2026. https://www.cencora.com/solutions/formularydecisions