AI Across Payers, Providers, and PBMs: Turning Pilots into Progress
- Katie Barone
- 3 days ago
- 4 min read
Originally published by Avalere Health
In recent years, health plans, providers, and pharmacy benefit managers (PBMs) have piloted artificial intelligence (AI) tools with the goals of reducing administrative burden, streamlining decision making, and improving services and care. The next step is to apply findings from pilot initiatives into wide-scale implementations that improve member and provider satisfaction, generate operational or clinical cost savings, and/or promote healthier outcomes, all without risks to compliance, litigation, customer experience, or bias.
Stakeholders have been testing use cases to support member- and provider-facing (front-office) activities and internal (back-office) activities.
Figure 1. Examples of Front- and Back-office Applications
Front-office Applications and Considerations
Member and provider touchpoints have been a common testing space for AI applications.
For health plans, virtual assistants can answer member questions related to benefits, cost-sharing, and network, as well as draft responses for live agents. Plans have deployed AI chatbots to help members navigate care and disease pathways, suggest next-best actions, and send personalized reminders for screenings and medication refills. They are also testing tools to help providers navigate prior authorization (PA) requests, serving as a real-time feedback tool to aid providers in completing submissions. These tools can be helpful, but to move beyond pilot programs they need to demonstrate measurable improvements in customer satisfaction, first-contact resolution, complaint/appeals rates, and average handling time. Front-office applications must sustain these gains across channels and populations, not just in a single queue.
Providers are deploying AI tools to assist in self-scheduling, intake, and basic triage, as well as providing patients with pre-visit instructions and more consistent follow up. Ambient scribe tools are also used to convert patient-provider conversations into structured notes and plain-language after-visit summaries, with the goal of saving providers time and standardizing the medical record. It remains to be seen whether this reduces documentation time; maintains note quality, completeness, and accuracy standards; and is satisfactory to clinicians without shifting burden to reviewers.
PBMs are testing formulary transparency tools and prescriber prompts to prevent avoidable PA and member out-of-pocket cost surprises. They are deploying tools intended to help members navigate clinical programs and improve medication adherence. To justify the investment needed to profitably scale these tools, PBMs should see reduced abandoned prescriptions, fewer PA-related callbacks, improved time-to-therapy, and higher prescriber satisfaction.
Another to consider for all of these member-facing applications across stakeholders is whether these tools and applications are acceptable to patients and members.
Back-office Applications and Considerations
AI can be applied to back-office workflow via process automation, data handling and analysis, and summarization.
Health plans have been using AI tools to abstract data, converting unstructured data into structured data to support various processes including risk adjustment, quality, and claims processing. They are increasingly exploring chatbots to support internal staff navigating policies, products, and processes. Internal tools must show meaningful improvements in productivity, FWA, and payment integrity.
Providers are testing AI tools to support revenue cycle management, eligibility and benefits verification, pre-billing, denial prediction, and appeal drafting. Electronic health records are increasingly integrating tools to summarize visits and convert faxes and images into structured data. For example, providers are using AI to assist in evaluating radiology images or electrocardiograms. Clinical applications such as these must prove outputs that are reliable and accurate. Additionally, operational applications must demonstrate measurable and auditable improvements in metrics including faster PA cycle times with stable overturn and appeal rates, reductions in preventable denials, and faster accounts receivable.
Like health plans, PBMs deploy AI tools for data abstraction and workflow optimization. Further, they are piloting tools to support formulary, network, and rebate analytics to design high-value formularies and networks, with the goal of minimizing unnecessary spending. AI tools need to consistently lower cost to serve without creating downstream obstacles or harm for PBM clients, prescribers, and patients.
Reasons Pilots Fail to Scale
Technology program pilots function best when they use curated data, hand-picked users, and manual backstops. At enterprise scale, heterogeneity, edge cases, and audit requirements expose integration gaps and brittle economics, turning demonstration wins into operational drag unless applications address these issues up front. Pilots typically fail to scale for a variety of reasons.
Figure 2: Reasons AI Healthcare Pilots Do Not Scale Successfully
Operational Outlook and Next Steps
In addition to operational feasibility, the regulatory landscape also influences AI adoption. The state regulatory landscape is quickly evolving, and at times at odds with federal priorities. While multiple states advance AI regulations that increase oversight, the federal government has expressed a desire to reduce red tape and bureaucracy.
Over the next 12–24 months, expect fewer new program pilots and a shift toward structured, stage-gated evaluations of existing pilots. Initiatives progressing beyond pilot stage must demonstrate sustained improvement on a small set of enterprise metrics (e.g., member experience, cycle time, cost to serve) with credible attribution, audit-ready traceability, and integration into systems of record. Until tools provide such evidence, healthcare leaders should maintain a measured posture. Plans, providers, and PBMs should focus on targeted deployments in AI-ready workflows, define clear go/no-go thresholds, and maintain continuous monitoring while avoiding letting demonstrations drive broad rollouts.
At Avalere Health we help health plans, providers, and technology companies translate policy and regulatory activity into strategic and operational action. If you are interested in learning more about how we can be a long-term partner, connect with us here.