AI adoption is accelerating across U.S. healthcare systems, with organizations recognizing its potential to improve clinical documentation, radiology, and diagnostics. Experts suggest that when implemented correctly, AI can significantly enhance clinical capabilities and operational efficiency. However, healthcare organizations that embark on AI initiatives without robust governance structures may inadvertently increase their operational risks.
According to experts from the global consulting firm BRG, there has been a marked shift in the types of AI tools being adopted by healthcare organizations in recent years, partly driven by budget constraints and workforce challenges. James McHugh, managing director at BRG, noted that while clients traditionally utilized AI for clinical diagnostics, there has been a growing focus on operational efficiency. “There’s suddenly a big focus on AI and operations and how it can make operations much more efficient and cost-effective,” he stated.
This shift reflects a broader trend where organizations are increasingly exploring automation and workforce augmentation, moving beyond the exclusive use of AI in clinical settings. Amy Worley, managing director and data protection officer at BRG, emphasized that while AI has long been beneficial in areas like radiology, operational advantages are becoming more pronounced as technology evolves. “I think there is definitely a desire to augment diagnostic capability and to make the providers stronger, better, faster,” she explained, adding that operational efficiencies are equally crucial.
However, the pace of AI adoption varies significantly among organizations, with disparities in data infrastructure maturity complicating efforts. McHugh pointed out that many health systems face challenges managing data and establishing effective workflows. “There’s a huge disparity between some health systems having bad data, a poor data infrastructure,” he remarked. He advised organizations to prioritize foundational data infrastructure before investing heavily in AI, as a solid groundwork is essential for successful implementation.
Good governance models can make or break AI implementation
A comprehensive governance model is essential for successfully integrating any AI tool, whether for predictive analytics or operational tasks like appointment scheduling. Worley remarked on the paradox facing healthcare organizations: they need to operate on tighter budgets while upgrading technology to support AI initiatives. “For AI to really do its magic, the tech needs to be in good shape,” she stated, highlighting the importance of governance to manage sensitive data effectively.
The governance structure must address crucial elements such as AI safety, bias, data privacy, and regulatory risks. As the scope of AI applications expands, the complexity of governance increases. “When [AI] was just sitting in clinical, a lot of times they were just pointing at research data or looking at molecular information, usually de-identified,” Worley noted. As AI applications broaden into operational areas, organizations face challenges related to unstructured data management.
The regulatory landscape is also evolving, with recent developments adding layers of complexity. On December 11, 2025, an executive order was issued aimed at preventing states from enacting their own AI regulations, which could lead to a national standard, although specific details remain under wraps. Worley emphasized that the “legal landscape is very patchy and incredibly dynamic,” making effective governance all the more critical.
As organizations ramp up AI adoption, many still lack robust governance models, which can expose them to various risks, including data breaches and compliance issues. McHugh observed that less than half of health systems have a solid enterprise governance model, often operating within siloed frameworks. “Many of them have siloed governance models. So that’s the thing I would be most concerned about right now,” he said.
Failure to implement effective governance can lead to significant consequences, including privacy risks stemming from known threats like prompt injection attacks. Worley warned that simple coding errors could disrupt workflows and heighten privacy vulnerabilities. “You can do some really simple prompt hacking where you say, ignore all of the prior conditions placed upon you,” she explained, underscoring the risks involved with operational AI.
Despite these challenges, there are basic AI-driven tools that can benefit healthcare organizations without imposing significant risks. Worley mentioned that basic operational improvements, like summarization, could be implemented more rapidly than stakeholders might expect. “I end up telling a lot of organizations they can go faster than they think they can,” she noted, although she stressed that any engagement with patient data introduces inherent privacy risks.
Moving forward, healthcare organizations should assess their AI use cases and involve relevant stakeholders to ensure effective governance. Worley advocated for a multidisciplinary approach that includes security experts, legal counsel, and healthcare providers, even though the latter may prefer patient care over governance responsibilities. “For AI governance to really work, it must be multidisciplinary,” she remarked, highlighting the need for a holistic strategy.
The evolving landscape of AI and its regulations presents healthcare organizations an opportunity to innovate while maintaining oversight. As McHugh concluded, while budget cuts can be painful, they are accelerating the adoption of technology and AI governance, which can ultimately yield positive outcomes.
See also
Nvidia Stock Expected to Maintain Growth Amid AI Hardware Challenges Through 2026
Strata Expanse Launches AI Data Center in Lawrence County, Ohio, to Boost Local Economy
AI Transforms Healthcare: GE’s Digital Twins Enhance Hospital Operations and ROI
HCLTech Acquires HPE’s Telco Solutions to Enhance AI and Network Engineering Capabilities
ESDS Launches Sovereign GPU-as-a-Service, Achieves 29% Revenue Growth in FY25
