Unreliable Data Undermines Health Plan AI Investments

Health systems and insurance plans are facing a fundamental contradiction. While most organizations acknowledge that their provider data is unreliable, they continue to invest heavily in artificial intelligence tools that depend on that same data. A new report highlights that nearly all health plans encounter data errors at least once a month, yet the majority lack a single, verified source of truth to correct them.

This gap between ambition and infrastructure is creating operational friction. The errors manifest as duplicate records, outdated locations, and practitioners listed as active who are no longer in the network. These inaccuracies lead to out-of-network billing surprises for patients and increased administrative costs for employers, undermining the efficiency gains that AI is supposed to provide.

Data Quality Remains Poor Despite Efforts

Research commissioned by Verato and conducted by Sage Growth Partners reveals the scale of the issue. Surveying leaders from major health systems and plans, the study found that 92 percent of health systems and 98 percent of health plans encounter data inaccuracies regularly. Despite this, only 28 percent of health systems and 36 percent of health plans have a fully implemented single source of truth for provider data.

The core problem is structural rather than technical. Healthcare organizations often build isolated tools for different departments, meaning a fix in one system does not automatically update others. This siloed approach requires constant manual review to keep data current across directories, electronic health records, and claims systems, leading to administrative waste and persistent errors.

AI Projects Depend On Broken Foundations

The stakes are high because AI is the top technology priority for most organizations in the sector. Seventy-eight percent of health systems and 70 percent of health plans rank AI as their primary investment area for the next few years. However, these investments rely on accurate data for revenue cycle management and claims adjudication, areas where fewer than half of respondents say they currently use data effectively.

When AI models are fed flawed input, they inherit and amplify those errors. As Jason Bihun of Verato noted, network growth and claims processing run on this same data. If the foundation is wrong, every downstream process, including new AI workflows, will produce incorrect results, potentially derailing strategic goals and increasing operational costs.

Structural Barriers Prevent Data Consistency

The persistence of these issues stems from the fragmented nature of healthcare data. Provider information is scattered across multiple platforms, including CRM systems and ERP tools. Without a centralized, automated flow of truth, organizations are left with point solutions that fragment again as soon as data leaves the specific system where it was fixed.

This lack of cohesion means that even as technology advances, the underlying data hygiene remains a manual and error-prone process. The trade-off is clear: organizations may gain speed in processing through AI, but they risk losing accuracy and trust in the process due to unreliable foundational data.