Artificial intelligence (AI) is poised to help solve healthcare’s most intractable challenges—and some health systems are starting to see strong results. But many health systems risk getting stuck in pilot purgatory without capturing significant value from their AI investments.
As AI applications have rapidly advanced, 9 in 10 health system executives say they are prioritizing AI capabilities over the next 5 years. But health systems are exploring AI opportunities at varying speeds. In the next 3 to 5 years, some health systems will achieve transformative outcomes, while others risk stalling in perpetual pilot mode. In fact, even outside of healthcare, only 5% of AI investments are currently producing calculable impact on the bottom line.
Among healthcare provider organizations, those that are scaling AI from pilots to enterprise adoption are starting to capture significant value. The secret to their success transcends the technology. It’s not that they are adopting different AI solutions from those stuck in pilot mode. Rather, it’s their holistic approach. These organizations have focused their adoption efforts on readying their staff for change and redesigning the processes AI will automate or augment.
Transformative AI adoption doesn’t happen overnight. Three distinct patterns of adoption are emerging, depending on the velocity and scope of health systems’ efforts. We discuss the crucial practices health systems must embrace to realize the promise of AI.
Health systems follow three patterns of AI adoption
Health systems generally fall into one of three categories that reflect how they are adopting AI:
-
Followers will decrease to 10% in 5 years: These organizations are taking a “wait-and-see” approach, typically because of limited resources or uncertainty about where to start.
Follower health systems will represent a smaller and smaller share of the overall adoption curve over the coming years as more organizations find opportunities to make low-risk, modest investments in AI, predominantly through their existing IT platforms.
In our 2025 Digital Transformation Survey, most health systems not yet piloting AI technologies indicated that they are actively planning to test and implement AI in the next 5 years, signaling an anticipated shift in adoption from being on the sidelines to pilot and potential scale. For instance, among the 35% of all respondents that are not yet piloting AI for clinical decision support, nearly all (31%) indicated that they are actively planning toward piloting.
-
Experimenters will increase to 60%: These health systems are actively piloting AI solutions, often propelled by pressure from vendors or internal leaders championing specific solutions. These organizations commonly jump to exploring technology options, versus starting with their enterprise strategic priorities. They may implement point solutions like ambient scribing to solve niche challenges, but these pilots by themselves are more likely to solve a single pain point than an enterprise strategic goal.
For instance, 48% of health system executives indicated in our 2025 Digital Transformation Survey that they are actively piloting AI-assisted clinical decision support, and 45% are piloting AI-supported triage and care navigation. While these point solutions may show promise in creating value if they scale, it’s equally important to align these solutions with broader enterprise initiatives for quality and patient experience.
Over the coming years, the experimenter segment will grow as many health systems pilot multiple AI solutions. However, a significant portion will remain in pilot purgatory. They will be unable to scale beyond initial projects because of challenges like poor integration, lack of change management, or insufficient alignment with strategy. Only a subset will successfully transition to the transformer segment.
-
Transformers will increase to 30%: These health systems are taking a strategy-first approach, weaving AI into their core activities to address priorities like access, cost, quality, and patient experience. Rather than focusing on technology alone, they align AI with their overarching objectives to drive enterprise value.
Over the coming years, this segment of adopters is expected to grow and increasingly stand apart from the experimenter segment. Transformers will replicate successful AI integration experience across multiple strategic initiatives, creating measurable value. As transformers embed AI into strategic initiatives, they will realize transformative outcomes.
Six practices will advance health systems from piloting to transformation
Health systems can take proactive measures to avoid pilot purgatory. These six practices will help health systems realize the transformative value of AI:
1. Align AI projects with organizational strategy, defining success as solving enterprise priorities rather than as having adopted technology.
Many organizations define their AI successes by technology adoption rates (e.g., provider licenses for ambient scribes or patient interactions with an AI-enabled chatbot). While such measures might capture the speed and scale of AI uptake, they fall short of determining the strategic value derived from those technologies.
Instead, health systems in the transformer segment are measuring how AI is advancing enterprise priorities, such as access (e.g., net-new patients seen or decreases in next available appointment) or quality (e.g., reduction in preventable falls, readmissions, or patient decompensation) that will generate return on those AI investments.
2. Integrate AI into existing workflows and strategic initiatives.
As part of a strategy-first approach to AI adoption, health systems in the transformer segment are embedding the technology into care ecosystems and processes that are designed to address core priorities.
For example, some health systems are redesigning how they match inpatients to acute care beds. Rather than simply placing patients in the first available bed, they are leveraging AI tools that can quickly analyze several datasets to optimally place patients based on the anticipated length of stay (LOS), specific care needs, the clinical team best aligned to them, and the care team’s projected capacity during that LOS. This maximizes the experience and outcomes for each patient, while creating a more efficient acute care delivery model for the organization.
Such integrations require seamless interoperability between AI and core IT platforms like the electronic health record (EHR) and adjacent datasets to ensure relevant and timely flow of information to fuel the AI tool and inform clinical and operational decision-making. They also require staff input and buy-in on the appropriate role and use of the technology to carry out their core functions.
3. Invest in data infrastructure and readiness before launching AI projects.
Health systems achieving transformative value from AI are assessing their data infrastructure and staff readiness for adoption before procuring and deploying the technology.
Prerequisites to audit include the sufficiency, accuracy, integrity, and format of data the AI will use. Health systems also must evaluate whether the data includes or is susceptible to bias that the AI tools could perpetuate. Creating a smooth adoption path also depends on ensuring the staff’s ability to effectively use and provide feedback on the AI tools under consideration.
For instance, numerous health systems are considering integrating agentic AI tools to assist with online patient scheduling. The potential benefits include a responsive, personalized patient experience, a lower call volume for the contact center, and reduced risk that patients will switch providers based on scheduling ease. To reap those benefits, the organization first must ensure standardized adoption of visit templates and availability of scheduling visits online. Health systems that make this available across clinic sites and open to a wide range of patient cohorts can create better access from this emerging opportunity.
4. Establish robust change management processes that support adoption and scaling.
Informed by the staff readiness assessment described above, long-term change management efforts are essential to ensure consistent adoption and optimal utilization.
A common pitfall is to provide initial training at launch and assume that staff no longer need adoption support. Instead, successful health systems will monitor adoption and proactively solicit feedback when gaps emerge.
For instance, tracking adoption trends among sites of care or end user characteristics might uncover untapped opportunities, such as addressing gaps in digital literacy, or concerns, like privacy protections for specific use cases. Health systems can then quickly address workflow or technology pain points that otherwise limit uptake and impact of the tool.
5. Create a performance management infrastructure for AI.
As the technology scales, organizations need to ensure sufficient levels of AI performance management. Constant audit and refinement of AI tool outputs are key to minimizing the risk of hallucinations while augmenting the end users’ workflows and cultivating their trust.
IT teams supporting AI adoption also need to protect the organization against unique AI vulnerabilities. A key component is calibrating AI use against approved applications. For instance, an organization may deploy a large language model (LLM) tool to transcribe clinical encounters and suggest potential billing codes, but appropriate use guardrails ensure staff review and confirm those codes before submitting to payers. Other vulnerabilities to safeguard against include emerging cybersecurity threats and risks of bias or inequity.
While health systems’ staff don’t traditionally maintain these skill sets, they are critical performance management capabilities for scaled AI transformation. Health systems can proactively cultivate and support these skill sets by redirecting the staff capacity gains from AI-enabled workflow efficiencies toward reskilling the workforce to support ongoing AI adoption and scale.
6. Define clear criteria for procuring platform-based tools versus point solutions.
Many organizations will continue to leverage their platform-based AI tools as the primary launching point for AI adoption. But health systems that take a strategy-first approach may consider solutions beyond what’s available in those platforms.
For instance, while many EHR platforms are starting to offer their own ambient scribe technology, they have not yet released capabilities beyond medical record transcription. Health systems that have adopted such EHR platforms may consider non-platform solutions that address incremental functionality like automated referral scheduling, order processing, or billing to capture value beyond the ambient scribe transcription.
Rather than exclusively calibrating to platform vendors’ roadmaps, health systems are increasingly building a comprehensive digital ecosystem designed to address enterprise priorities. This approach is helping them capture near-term value from existing and proven AI technology.
Adopt AI with the goal to transform
AI offers novel and transformative opportunities to address healthcare’s most intractable challenges. Effectively tapping into this potential requires health systems to establish the vision and expectation that the technology must closely align with enterprise priorities and tightly integrate into initiatives that are best positioned to scale and create measurable impact. Health systems that do so can avoid the pitfalls of pilot purgatory and achieve meaningful, enterprise-wide AI-enabled transformation.
