The AI Reckoning in Healthcare: What Payers Can't Afford to Ignore 

The Problem Nobody Can Ignore

The American Medical Association recently reported that 60% of physicians believe AI tools are causing unnecessary claim denials, and that 80% of prior authorization denials are overturned on review. Major insurers, including UnitedHealthcare, Cigna, and Humana are now facing lawsuits as a result.

These are not small problems. They represent a fundamental trust crisis between payers, providers, and patients. And the way AI is being deployed, or misdeployed, sits right at the center of it. At Improving, we believe the answer isn't to decrease AI use, but to be smarter about how you use it.

Getting Governance Right Before Anything Else

One of the most important things Scott emphasizes is that AI governance is the foundation that everything else is built on.

"It's about observability and AI governance, ensuring the right tools are in place," Scott explains. When building AI models for clinical or administrative decisions, payers need structured governance processes that review data quality, test for bias, and involve the right people, like compliance officers, clinicians, legal counsel, IT, and data scientists working together, not in silos.

Human bias can be baked into AI models without anyone realizing it. The good news is that toolkits now exist specifically to detect these patterns. But you have to build the process to use them consistently.

Beyond bias detection, Improving works with organizations to conduct formal risk analyses, set acceptable risk thresholds with legal teams, and implement human-in-the-loop reviews for any decision that falls in a gray area. This isn't just the right thing to do, but it’s increasingly what regulators demand.

Staying Compliant as Rules Evolve: State by State

Texas, Arizona, and Maryland have already passed laws prohibiting fully automated medical necessity denials. Colorado now mandates annual AI impact assessments, and this is just the beginning. Regulations are expanding rapidly, and they look different depending on where you operate. For many health plans, this creates a real strategic question: do you build now, or wait for the regulatory dust to settle? Scott's answer is clear: build with flexibility.

Rather than pausing development or building for a single regulatory environment, Improving helps health plans design adaptive systems that can respond to new rules automatically. This approach allows payers to stay compliant across multiple states without rebuilding from scratch every time a new law passes, and it opens the door to expanding market reach rather than being limited by regulatory uncertainty.

What CMS Is Requiring, And What It Takes to Deliver

Two major CMS initiatives are reshaping what payers must be able to do technically, and the deadlines are approaching.

The WISER Model is a six-year Medicare pilot program where AI is used by tech companies to review prior authorizations. For payers and vendors alike, this signals that CMS is actively testing AI-driven review at scale. Improving is watching this closely, because a successful pilot likely means broader adoption, and organizations that are already operating mature AI workflows will be far better positioned.

Interoperability Requirements by 2026 mandate that payers achieve 80% real-time prior authorization decisions using FHIR standards, with specific denial reasons made public. This is a significant technical lift. Most payer organizations today operate on batch-processing systems where data is collected and processed in large periodic intervals rather than in real time.

Making the shift to real-time processing requires more than upgraded software. It requires a fundamental rethinking of how data flows through an organization, supported by powerful rules engines and decision frameworks that can operate at speed without sacrificing accuracy or compliance. Improving brings deep technical expertise in exactly this kind of infrastructure transformation, helping payers modernize their systems while keeping operations running throughout the transition.

You Own the AI You License

One point Scott makes that every healthcare leader should internalize: CMS has made clear that health plans are accountable for the AI tools they purchase from vendors. "We bought it from a vendor" is not a defense.

This changes the vendor relationship significantly. You can't just evaluate a vendor's product. You need to understand how it makes decisions, what data it was trained on, whether it has been tested for bias, and how it will be monitored over time.

Improving leads health plans in building the internal capability to evaluate, monitor, and document the AI tools in their portfolio, whether those tools were built in-house or licensed from a third party. This includes ongoing bias checks, model drift management, and maintaining the documentation you'll need if regulators come asking.

AI as a Competitive Advantage, Not Just a Cost-Cutter

There's a temptation to frame AI purely as a way to cut operational costs, and it actually does deliver there. Predictive fraud detection, waste management, and claims processing automation all show meaningful efficiency gains. For organizations operating on thin margins, that matters. But Scott sees the bigger picture: AI is becoming a competitive moat.

When a health plan uses AI well, it generates better data. Better data leads to better outcomes for members. Better outcomes improve the member experience and strengthen retention. The organizations that invest in AI infrastructure thoughtfully today are building advantages that will be very difficult for competitors to close later.

Protecting It All: Cybersecurity Can't Be an Afterthought

Healthcare payers hold some of the most sensitive data in existence, protected health information (PHI) at massive scale. With AI expanding data access and processing, the security stakes rise accordingly.

Improving recommends a zero-trust security architecture for payer environments, which is a model where no user, device, or system is automatically trusted, and every access request is verified. Paired with AI-powered behavioral analytics and micro-segmentation, this approach provides a strong defense against both external attacks and internal vulnerabilities. Continuous, real-time vendor security monitoring rounds out a mature security posture.

Built for Compliance From Day One

As an AWS Advanced Consulting Partner, Improving helps healthcare organizations design and operate cloud‑native environments on AWS that are architected for HIPAA compliance from the ground up. Rather than retrofitting controls after systems are live, this approach embeds security, governance, and compliance directly into the platform’s foundation. The result is an AI‑ready infrastructure that can scale, adapt to evolving regulations, and stand up to regulatory scrutiny without slowing innovation.

Where to Start

The complexity of AI in the payer space can feel overwhelming. Regulations shifting, vendors multiplying, CMS deadlines approaching. It's a lot to manage while also running a health plan.

Improving's approach is to meet organizations where they are. Some payers are already operating sophisticated AI programs and need help with governance maturity or interoperability readiness. Others are earlier in the journey and benefit from a structured assessment of where they stand and what to prioritize first.

What's consistent across every engagement is our belief that getting AI right in healthcare isn't optional. For payers, it's becoming the foundation of sustainable, compliant, and competitive operations. Ready to talk about where your organization stands? Let's start the conversation.