From Adoption to Control: The New Mandate for Health System Boards in the Age of AI

*:last-child]:mb-0 [&>*:last-child]:pb-0 [&>p]:font-normal text-left [&>.prm-button-bar]:justify-start [&_.portable-text-link]:font-semibold [&_.portable-text-link]:underline [&_.prm-button-bar_a]:no-underline [&_li]:text-light-gray [&_p]:text-light-gray">

Key Takeaways:

• Artificial intelligence (AI) is a structural force reshaping how health systems allocate capital, manage risk and define performance, requiring boards to move from adoption-focused conversation to conversations about control.
• Treating AI as a collection of point solutions creates fragmentation and risk. Boards must ensure AI is governed a coordinated, enterprise-wide portfolio with continuous oversight
• Digital literacy at the board level now means the ability to interrogate AI-driven strategy, demand meaningful reporting and ensure governance models keep pace with a dynamic, evolving technology.

Artificial intelligence (AI) is rapidly becoming central to a health system's strategy.

And yet, in many boardrooms, the conversation hasn't fully caught up. Discussions still center on vendors, features and whether a given tool "works." Those are reasonable starting points. But they don't equip boards to govern a technology that is fundamentally reshaping how healthcare organizations operate and compete.

Digital literacy at the board level isn't about understanding AI itself. It's about understanding what AI changes.

AI is a structural force reshaping how capital is allocated, how risk is introduced and managed, and how operational performance is defined.
Many boards are underprepared for oversight at that level.

The AI Control Problem in Healthcare

Across the healthcare industry, organizations are investing in, piloting and deploying AI tools at a rapid pace. Health systems are experimenting across clinical, operational and administrative domains. As of late 2025, approximately 71 percent of U.S. hospitals had integrated some form of AI into daily operations, up from 66 percent in 2023.

What's missing isn't activity, but coherence.

In most health systems, AI implementation efforts are spread across departments, driven by individual leaders and evaluated in isolation. The result is a growing layer of complexity that few leaders are managing effectively.

A digitally literate board does not need to understand how algorithms are trained or models are tuned. But it does need to recognize that AI behaves less like traditional IT and more like a dynamic portfolio of assets - each with its own risk profile and return potential.

When organizations evaluate one AI tool at a time, they inevitably accumulate tool sprawl, redundancy, misalignment and waste. As a result, they deploy point solutions stitched together like Frankenstein's monster rather than creating a thoughtful AI ecosystem.

When AI is viewed as a portfolio, different strategic questions emerge:

• Where is the organization over- or under-invested?
• Which capabilities are foundational versus experimental?
• Where are there redundancies, and where can additional synergies be found?
• How do technology investments ladder up to enterprise priorities and future-facing strategy?

Governance as Strategy

This shift toward a portfolio mindset also reframes governance. In many organizations, governance is still treated as a compliance function, a necessary but often reactive layer applied after technology decisions are made (if it's conducted at all). In fact, although AI deployments continue to increase, most health systems (80 percent) still lack internal governance standards guiding future adoption.

This is a strategic miss.

AI introduces categories of risk that are not easily pre-empted or contained. These range from liability tied to clinical decision-support tools to more subtle but equally significant risks associated with data quality, bias and unapproved use.

AI is inherently dynamic, making these risks even more challenging. Models evolve as they ingest new data, and their performance can change over time in ways that are not always predictable or visible.

That reality exposes a gap in how many boards think about oversight. It is no longer sufficient to approve a technology investment and assume its performance will remain stable. Governance, in the context of AI, becomes an ongoing discipline that requires clear accountability, continuous monitoring and an explicit understanding of where the organization is willing to accept risk and where it is not.

Boards don't need to design governance models, but they do need to ensure they exist and that they're built to accommodate and flex with a dynamic, evolving technology.

Rethinking ROI Models for a Dynamic System

The rapid pace of change in the world of AI also complicates one of the most familiar concepts in the boardroom: return on investment. Most current AI ROI frameworks are designed for static technologies, where inputs and outputs can be measured in linear ways. But AI's impact is often intertwined with workflow changes, staffing adjustments and broader operational shifts. Gains that appear to be driven by AI may, in reality, be the product of multiple factors. At the same time, the costs associated with AIare frequently undercounted. This is particularly true for costs tied to supervision, integration and change management.

For boards, this creates a subtle but important risk. Without a more rigorous approach to value measurement, organizations can convince themselves that an investment is working when the evidence is, at best, incomplete. Digital literacy, in this context, means recognizing that traditional financial and operational measurement models are no longer sufficient, which, in turn, should lead toward adoption of more nuanced, longitudinal approaches to performance evaluation.

Boards should push leadership to answer key questions:

• How are we isolating the true impact of AI?
• Is performance improving, stable or degrading over time?
• What hidden costs (e.g., supervision, workflow disruption) are we not accounting for?

If leadership can't answer those questions, the organization isn't measuring value - it's assuming it exists.

Policy and Markets Shaping Awareness

Beyond the enterprise, boards must begin to directly engage with broader forces shaping AI adoption. Policy and market dynamics are becoming central factors influencing the future of AI. Clinical adoption, for example, is already constrained by uncertainty around liability. Physicians are understandably hesitant to rely on tools when the boundaries of accountability are unclear. At the same time, structural issues, such as the degree of interoperability allowed by electronic medical record platforms, are shaping which innovations can scale and which cannot.

Boards that treat these dynamics as externalities risk being reactive. Boards that understand them as strategic variables have an opportunity to shape the environment in which they operate-whether through advocacy, partnerships or more deliberate positioning within the market.

How Boards Must Structurally Prepare for the AI-Enabled Future

It all points to a more fundamental challenge that is less comfortable to address. Many boards are not structurally equipped to manage in a digital-first environment. Not because they aren't capable, but because the skill mix hasn't caught up to the moment.

The traditional composition of healthcare boards emphasizes clinical expertise, financial oversight and community representation. But these perspectives reflect the priorities of a different era. They remain essential, but they are no longer sufficient on their own.

Upholding the Right to Question

If AI and digital capabilities are central to strategy, then the ability to interrogate those topics cannot sit at the margins of board expertise. That does not mean every board member needs to become a technologist. It does mean the board must collectively have enough depth to:

• Challenge assumptions
• Interpret signals
• Guide decisions in a landscape that is increasingly defined by technology

The good news? This gap can be closed with intentionality and urgency. Instead of high-level overviews of AI concepts, boards need to be exposed to the real decisions their organizations are making: where capital is being deployed, which use cases are being prioritized, what is working and what is not. Case-based discussions grounded in the organization's own experience are far more effective in building the kind of fluency that enables meaningful oversight.

Giving Thought to Who Is in the Room

Who is in the room is equally important. Boards that lack digital expertise must find ways to bring it in, whether through new members, formal advisory structures or engagements or deeper collaboration with internal leaders who are closest to the work. Relying on vendors to fill that gap is incomplete at best and risky at worst. Independent perspectives, particularly from those who have implemented AI in complex, real-world environments, are critical in separating signal from noise.

Normalizing and Demanding AI Reporting

Boards must normalize AI reporting, making it an ongoing area of strategic scrutiny. This includes regularly examining the organization's AI portfolio, understanding how those tools are performing over time and pressure-testing the oversight around them. It also means being willing to ask uncomfortable questions about where things could fail and the consequences.

Increasingly, many organizations are recognizing that this gap cannot be solved internally alone. Rather than navigating AI adoption independently, they are pooling insights through participation in strategic collaboratives, sharing use cases, pressure-testing vendors, benchmarking performance and, importantly, aligning on governance and risk frameworks that can stand up to clinical and regulatory scrutiny.

These emerging collaboratives create space not only for education but also for applied learning. Here, boards and executive teams can engage directly with real-world implementations, compare outcomes and build a more grounded understanding of what works. Collaboratives also offer a unified voice in shaping policy, particularly as questions of accountability, interoperability and clinical liability continue to evolve.

Final Thoughts

AI is often described as a transformative technology, but its real impact lies not just in what it enables but in what it exposes. This includes gaps in governance, weaknesses in operating models and, increasingly, limitations in how boards engage with strategy itself.

The organizations that realize lasting value from AI will not be those that move most quickly. They will be those that move most deliberately and with a clear understanding of how to align technology, operations and oversight.

Digital literacy, in this context, means understanding how AI reshapes the decisions that matter most. Until that shift happens, boards will continue to have conversations about technology - when what they really need are conversations about control.

For more:

Discover how Premier helps health system leaders govern AI deliberately and competitively through hands-on advisory powered by data, technology and real-world operating experience.