How Health Insurers Can Implement AI Tools

This series has covered how health insurers can see real benefit from artificial intelligence (AI) in claims processing, appeals and grievances, member care, risk adjustment, and quality. In our final segment, we look at best practices for health plans to follow when implementing AI. The following strategies and tactics will help payers achieve optimal benefits from their AI initiatives while avoiding potential issues.

Have a “Big Picture” Strategy

Today’s AI solutions can build on each other, with one team of AI agents handing off work to another agentic AI team. Those collaborative relationships can increase the value of both solutions, and that value is optimized when the broader automation is envisioned from the start. The ultimate goal—automating as much of the payer value chain as possible, from enrollment through explanation of benefits (EOBs)—should influence decisions about which tools and solutions will serve the long-term objective.

For example, the intelligence and workflows developed while training AI agents for “pend” or “edit” resolution should be extended across other high-value operational areas. Health plans can repurpose these agents’ capabilities to drive efficiency in prepay and postpay audits, appeals and grievances handling, complex claims research, and claim reprocessing. By reusing foundational learnings—such as clinical and regulatory context interpretation, data cross-validation, and multisystem navigation—payers can accelerate their deployment of AI agents across functional domains, reduce manual intervention, and enhance both compliance and member-provider experiences. This approach not only maximizes return on investment (ROI) investments but also creates a scalable foundation for expanding AI-based automation.

Avoid Ad Hoc and Point Solutions

Implementing AI point solutions or allowing ad hoc initiatives consumes resources without creating a solid foundation for agentic AI expansion. Standalone AI tools can lead to fragmented data silos, making it difficult for organizations to see patterns across functions, such as recurring denied claims that could be overturned on appeal. Siloed systems prevent AI solutions from learning holistically and sharing insights, limiting their value and overall impact. A coordinated approach enables end-to-end visibility of AI tools, allowing them to recognize patterns across the value chain and ensuring that AI-powered workflows build on each other. That leads to operational excellence and stronger returns.

Solve a Small Problem First

While “reducing high administrative costs” is a worthy goal, it’s also very broad. Picking off a clearly delineated pain point, such as a high occurrence of appeals and grievances coming from a single large hospital system, will be easier to tackle. Early success and measurable returns also help build acceptance of and more appetite for AI. A sharper focus at the outset also helps ensure the problem drives tool and solution selection.

For example, consider aggregating the manually processed pend volumes across key categories such as duplicates, benefits, contracts, and pricing for an initial AI agent deployment. These typically represent more than 70% of manual work. Each of these pend types requires the AI agent to access and reconcile information from a broad ecosystem of payer source systems using application programming interfaces (APIs). These include core administration platforms, provider data management, clinical and case management, pricing, contracts, appeals and grievances, and customer relationship management systems. By aligning development with these cross-functional data dependencies, payers can accelerate scalability, maximize the impact of automation, and ensure AI agents deliver measurable outcomes across high-volume pend categories.

Benchmark the Current Process and Resources Allocated to the Situation

Measuring success is essential for building confidence and accelerating adoption in subsequent phases of AI agent deployment. To do this effectively, payers must capture readily available metrics before and after AI implementation. These include incidence rates, full-time equivalent (FTE) effort, productivity gains, cost savings, rework levels, and compliance adherence. Continuous monitoring of these key performance indicators (KPIs) ensures transparency of value delivery, strengthens the business case, and provides leadership with the data needed to scale AI across additional functions with confidence.

Build a Sound Data Strategy

High-quality AI requires high-quality data. Starting with a smaller project makes it simpler to identify required data and makes data cleansing more manageable. Going forward, following industry standards such as Health Level Seven (HL7) and Fast Healthcare Interoperability Resources (FHIR) should ensure data is interoperable internally, as well as with third parties. Payers should prepare to collect, store, and manage audio, visua, and text-based data. Multimodal AI will help draw insights from this variety of information.

Get Employee Buy-In

AI in health care must be overseen by humans in the loop. Employees will require ongoing training to understand how AI tools can help them be more efficient and where AI will need their management and input. In addition, employees have front-line views of process and operations issues. Their input is essential to understanding how operations work in the real world vs on paper to ensure that solutions solve actual issues. Involving operations staff in solution designs also helps build their acceptance of AI tools and agents as valuable assistants.

Develop Strong Governance, Security, and Ethics Protocols

The odds are good that employees already are using AI, even if that’s just to query free chatbots or use the AI tools built into popular business software. Payers must ensure their governance and ethics policies address these applications. In addition, strong governance policies that prohibit the use of uncoordinated, unverified, or non-compliant AI tools are critical. Any of these could undermine data and operational integrity.

Payers also must take care to avoid “black box syndrome.” If an AI solution denies a prior authorization request or appeal, there must be a clear audit trail about what policies and/or evidence guidelines the agent consulted to arrive at that decision.

Finally, payers must ensure AI tools meet compliance and privacy requirements and develop strict security protocols governing access to the tools and their data.

Be Transparent About AI Use

Payers need to continually communicate with their stakeholders about how they are using AI. Specifically, providers and health care systems need to trust that care decisions are made without bias, and members should understand what their health plan’s AI use does for their quality of care, costs, and overall experience.

Capitalize on Third-Party Expertise

Service providers with expertise in the health care industry and experience implementing AI solutions can help payers evaluate, design, and pilot these tools—and then scale them as they demonstrate value. Further, third parties can apply lessons learned across multiple industry installations, where an in-house team necessarily has a more limited view. In addition, technology partners can stay current on fast-moving AI technology advances and regulatory compliance while health plans focus on their members.

Conclusion

The value AI tools deliver will depend on the approach health plans take when implementing them. Health plans with a clear vision for how to expand and build on AI wins that eliminate true pain points will be positioned to grow the value of AI across their business. Those returns should include more efficient operations, improved member and provider experiences and, ultimately, better care.