From Pilots to Performance: A New Playbook for AI in Medical Groups

AI has quickly become a standing agenda item in strategy discussions, capital planning conversations and, increasingly, investor and rating agency expectations. While it’s widely understood that AI will reshape healthcare operations and the practice of medicine, measurable, sustained impact on physician enterprise economics — access, productivity and margin — remains uneven at best. In many organizations, the ROI is still largely theoretical.

This is not a technology problem. It is an execution problem.

The organizations beginning to separate themselves are not those with the most advanced tools. They are the ones recognizing a more fundamental truth: AI is not a technology initiative to be deployed. It is an operating model to be built.

The Strategic Misstep: Treating AI as a Capital Purchase

At the executive level, AI is often evaluated through a familiar lens. A need is identified, vendors are assessed, capital is allocated and implementation follows.

That approach is comfortable but ineffective.

AI does not behave like traditional enterprise technology. It is not static once deployed, and its value is not solely tied to the tool itself. Instead, its performance is also dependent on the environment into which it is introduced — the workflows it touches, the data it consumes, the people who use it and the discipline with which it is governed over time.

As a result, many organizations find themselves in a familiar position months after deployment: The technology is functioning as designed, but the expected financial or operational impact has not materialized. Often, this problem stems from the absence of an operating model designed to capture value.

A More Useful Lens: Capability Before Technology

Boards and executive teams can benefit from reframing the core question. Instead of asking, “Which AI investments should we prioritize?”, leaders should ask, “What capabilities must exist for AI investments to translate into measurable performance?”

Across health systems, four interdependent capabilities consistently determine whether AI produces results or disappointment. They are not discrete initiatives, nor can they be developed sequentially. They function as a system.

1. System design that maximizes capacity gains.

The first is the ability to redesign clinical and operational workflows in a way that captures the capacity AI creates. By nature, AI introduces efficiency, reducing the time required for documentation, coding, scheduling or decision-making. Value emerges only when that recovered time is intentionally redeployed to improve patient access, expand panel capacity, improve care coordination or reduce reliance on premium labor. Without deliberate redesign, organizations simply make existing processes faster without changing outcomes.

2. Clear and continuous governance.

The second is governance — though not in the traditional sense of periodic oversight or approval. AI requires a continuous operating discipline. Tools must be evaluated, deployed, monitored and, when necessary, replaced in a structured and repeatable way. Leading organizations have moved beyond static committees and instead manage AI as a living portfolio, anchored to outcomes and actively steered based on performance. This shift from episodic decision-making to ongoing orchestration is subtle but decisive.

3. Clean, consistent data.

The third is data readiness. AI systems do not correct underlying data inconsistencies; they magnify them. In physician enterprises where documentation practices vary, scheduling templates are inconsistently maintained, and data flows across multiple systems, AI performance becomes uneven and unpredictable. Trust erodes quickly, and adoption rates fade.

4. A people strategy built on intentional ownership.

The fourth is people strategy. AI does not manage itself, and it does not integrate seamlessly into clinical environments without intentional ownership. The absence of a clearly defined role bridging clinical workflows and technical capability — often referred to as a Clinical AI Product Owner — is one of the most consistent failure points. Without that connective function, tools are implemented but not optimized, and initial gains plateau before translating into enterprise value.

Individually, each capability matters. Collectively, they determine whether AI becomes a source of competitive advantage or an accumulation of underperforming investments.

The Market Dynamic Leaders Cannot Ignore

The pace and volatility of the AI vendor landscape only compounds AI’s execution challenge.

Providers are navigating a fragmented market with dozens of vendors in each major category including clinical documentation, revenue cycle, decision support and patient access. Many of these companies are early in their maturity, with limited long-term evidence and uncertain financial durability. At the same time, major electronic health record (EHR) platforms are accelerating their own AI roadmaps, creating both opportunity and ambiguity around when to adopt third-party solutions versus waiting for native capabilities.

This dynamic often leads to strategic hesitation. While the instinct to wait for clarity —particularly given the scale of existing EHR investments — is understandable, waiting does not eliminate risk. It simply shifts it.

Organizations that delay building the capabilities required to manage AI will find themselves equally unprepared when those capabilities arrive through their EHR or any other channel. The challenge is not access to AI. It is the ability to operationalize it effectively.

Anchoring strategy to outcomes, rather than tools, is a more durable approach. When AI initiatives are tied to specific, measurable objectives — improving access, reducing denials or optimizing cost of care — new technologies, whether from startups or EHR vendors, become inputs to evaluate rather than disruptions to manage. This orientation reduces volatility and preserves strategic continuity in a rapidly changing market.

Where Measurable Value Is Emerging

Despite the broader variability in performance, several AI applications have matured to the point where measurable returns are being realized within physician enterprises.

• Ambient documentation solutions are reducing administrative burden on clinicians and, when paired with intentional workflow redesign, can expand access and improve retention:
• Revenue cycle applications are demonstrating clear financial impact through coding optimization and denial prevention. 
• Clinical decision support tools are beginning to influence ordering behavior in ways that reduce unnecessary utilization and improve cost transparency while maintaining discretion at the point of care.
• AI-enabled voice agents are reshaping patient access functions, albeit with significant dependency on underlying data and workflow discipline.

Recognizing that these use cases are not simply point solutions is critical. They are catalysts that should prompt the enterprise to reshape how care is delivered, how data is managed and how decisions are governed. 

The Financial Reality: Moving Beyond Hypothetical ROI

As AI adoption accelerates, organizations are increasingly being asked to project meaningful financial returns. In many cases, those projections rely heavily on assumed labor efficiencies, including reductions in full-time equivalents.

While AI can and will change labor models over time, projecting workforce reductions before adoption is stabilized and workflows have been redesigned introduces significant risk. Doing so can undermine clinician and staff engagement, distort implementation priorities and ultimately delay the realization of value.

More durable AI ROI models focus initially on:

• Expanding access and throughput. 
• Reducing avoidable cost and waste. 
• Improving revenue integrity. 
• Enhancing retention of high-value clinical talent. 

Labor model transformation follows only after these gains are proven and sustained.

For the C-suite, the implication is clear: Scrutinize not just the magnitude of projected returns but also the sequence in which they are expected to occur.

How Premier Supports Health Systems in This Transition

Given the challenge of successful AI adoption, coupled with the organizational transformation required to support ROI at scale across a health system and its employed physician enterprise, structured advisory support becomes critical.

Premier supports health systems through a combination of Physician Enterprise advisory services and dedicated AI and technology advisory capabilities, designed to help executive teams move from fragmented AI activity to a coherent, performance-driven strategy.

That work typically begins with a clear-eyed assessment of readiness across the physician enterprise, evaluating workflow design, governance maturity, data infrastructure and organizational roles. From there, Premier works alongside leadership teams to define outcome-driven priorities and establish a disciplined approach to evaluating and selecting AI solutions in a rapidly evolving market.

The work that follows selection is equally important. Implementation is structured around measurable outcomes, with deliberate attention to workflow redesign, physician engagement and change management — areas where many organizations generally underestimate the effort required. Baseline performance is established using benchmarking data from Premier’s Performance Insights Value Optimization Tool (PIVOT) to create metrics and KPIs that matter, allowing for objective measurement of impact over time.

As initiatives move into production, Premier supports the development of a repeatable governance model to manage AI as an ongoing portfolio rather than a series of isolated projects. This includes performance monitoring, optimization and, when appropriate, transition to alternative solutions as the market evolves.

In parallel, organizations gain access to deep expertise spanning revenue cycle optimization, clinical decision support and physician enterprise performance. They can also tap into dedicated AI and technology specialists who ensure AI initiatives are governed, measured and aligned to enterprise priorities.

The objective is straightforward: to help health systems convert AI from a strategic concept into a reliable driver of operational and financial performance.

What C-Suite Executives Should Be Asking Now

As AI continues to move from experimentation to expectation, leaders are responsible for ensuring the right tools are acquired and the right utilization models are employed.

That starts with a shift in the questions being asked:

• Are our physician enterprise workflows designed to capture the capacity AI will create?
• Do we have a governance model that actively manages AI performance over time?
• Is our data infrastructure capable of supporting consistent, reliable AI outputs?
• Who, specifically, owns the intersection of clinical operations and AI within our organization?
• Are our investment decisions anchored to outcomes — or to vendor roadmaps? 

The answers to these questions will determine which health systems translate AI investment into measurable performance — and which do not.

AI will, without question, reshape the economics of the physician enterprise. But it will not do so evenly.

The differentiator will be the ability to move beyond viewing AI as a set of tools to be acquired and begin building the operating model required to make those tools matter.

Because in the end, the impact of AI will not be measured in pilots launched or vendors selected. It will be measured in access, margin and ROI.

Ready to build an AI operating model that delivers measurable value for your medical group? Watch our on-demand webinar, “AI in Medical Group Operations: Separating Hype from ROI.”