Healthcare providers embrace AI tools while systems lag in readiness and governance

U.S. healthcare providers are rapidly warming to generative artificial intelligence, but a widening gap between enthusiasm and system readiness is raising concerns about safety, training, and ethical oversight, according to a new study published in Healthcare.

The study, titled "Healthcare Providers' Perspectives on Generative Artificial Intelligence (GenAI) Adoption, Adaptation, Assimilation, and Use in the United States," provides a detailed snapshot of how clinicians perceive, use, and evaluate generative AI in real-world practice.

Based on a nationwide cross-sectional survey of U.S. healthcare professionals, the research shows that generative AI has moved beyond early experimentation into routine clinical workflows, particularly in administrative and documentation tasks. It also reveals critical weaknesses in training, governance, and infrastructure that could shape how safely and effectively the technology is integrated into patient care.

Strong clinician support meets limited institutional readiness

The study finds overwhelming support among clinicians for the use of generative AI in healthcare, with nearly nine in ten respondents viewing the technology as useful in current patient care and an even higher proportion expecting its usefulness to grow in the future. This level of acceptance signals a clear shift in professional attitudes, reflecting how quickly AI tools have entered clinical environments.

However, the research highlights a stark mismatch between this optimism and the reality of implementation. Fewer than half of surveyed providers reported receiving any formal training in AI, and only a small minority indicated that their organizations had begun adopting or integrating AI systems into workflows. This gap suggests that while clinicians are increasingly willing to engage with AI, healthcare systems themselves are lagging in preparing the workforce and infrastructure needed for safe deployment.

The findings point to a fragmented adoption landscape, where individual clinicians may experiment with or use AI tools independently, often without standardized guidance or institutional oversight. This creates uneven levels of competence and raises concerns about consistency in patient care.

The study also shows that organizational adoption is often driven by top-level leadership rather than coordinated system-wide strategies. At the same time, a significant portion of respondents reported uncertainty about who is leading AI adoption efforts within their organizations, underscoring a lack of clarity in governance structures.

Despite these limitations, willingness to support broader AI integration remains high. More than two-thirds of respondents expressed readiness to embrace AI in clinical settings, provided appropriate safeguards are in place. This indicates that resistance is not rooted in rejection of the technology itself, but rather in concerns about how it is introduced and managed.

AI reshapes clinical workflows with a focus on efficiency

Generative AI is already influencing how healthcare providers perform daily tasks, with its strongest impact seen in administrative and documentation processes. The study identifies report writing, medical documentation, and clerical tasks as the most common areas of AI use, reflecting a clear trend toward leveraging AI to reduce time-consuming administrative burdens.

Clinicians reported that AI tools are particularly valuable in improving time management and streamlining documentation, which are often cited as major contributors to workload stress and burnout. By automating repetitive tasks, AI enables providers to allocate more time to patient-facing activities, potentially improving care quality and professional satisfaction.

In addition to documentation, AI is also being used in research, diagnosis, and aspects of patient care, although to a lesser extent. Its role in diagnostic support and clinical decision-making is growing, but remains more cautiously adopted due to concerns about reliability, transparency, and accountability.

The study highlights that AI's perceived benefits extend beyond efficiency gains. Many respondents believe that AI can reduce errors and improve the accuracy of clinical documentation, which has direct implications for patient safety. Faster turnaround times for administrative processes and improved data handling are also seen as key advantages.

However, the integration of AI into clinical workflows is not without complexity. The research underscores that while AI can enhance productivity, it also introduces new dependencies and risks. For instance, overreliance on AI-generated outputs may lead to reduced critical engagement by clinicians, especially if systems are perceived as highly reliable.

Another emerging issue is the uneven distribution of AI use across different clinical roles and specialties. Providers with higher levels of education, particularly those with doctoral qualifications, were more likely to identify advanced use cases such as diagnosis and report writing, suggesting that familiarity and expertise influence how AI is utilized.

Despite these advancements, the study reveals that many healthcare providers remain cautious about expanding AI's role beyond administrative support. This reflects an underlying tension between recognizing AI's potential and maintaining confidence in human-led clinical judgment.

Ethical risks, training gaps, and governance challenges slow adoption

While enthusiasm for generative AI is strong, the study identifies a range of barriers that could slow or complicate its integration into healthcare systems. The most significant obstacle is limited knowledge of AI among clinicians, which remains the leading barrier to adoption.

Closely linked to this is the lack of structured training programs. Without formal education and hands-on experience, providers may struggle to use AI tools effectively or understand their limitations. This not only affects individual performance but also increases the risk of errors and misinterpretation in clinical settings.

Fear of job displacement also emerges as a notable concern, reflecting broader anxieties about the impact of automation on healthcare roles. Although the study suggests that AI is more likely to augment rather than replace clinicians, these fears persist and may influence attitudes toward adoption.

Ethical and operational risks represent another major challenge. Concerns about data privacy and surveillance are among the most prominent issues identified by respondents, followed by security risks, misinformation, and lack of regulatory frameworks. These concerns highlight the sensitive nature of healthcare data and the need for robust safeguards to protect patient information.

The study also points to the risks associated with algorithmic bias, lack of transparency, and the so-called black box problem, where AI systems produce outputs without clear explanations. These issues can undermine trust and make it difficult for clinicians to fully rely on AI-generated recommendations.

Importantly, the research points out that the absence of human oversight is seen as a critical risk. A majority of respondents identified the need for human-in-the-loop systems to ensure that AI supports rather than replaces clinical decision-making. This aligns with broader calls for responsible AI governance that prioritizes accountability and safety.

To address these challenges, the study outlines several key strategies. These include expanding training programs, increasing clinician involvement in AI design and development, improving data protection measures, and enhancing transparency in how AI systems operate. The findings suggest that successful integration will depend not only on technological innovation but also on organizational and cultural change.

Another notable insight is the need to rethink how efficiency gains from AI are used. While AI has the potential to increase productivity, most clinicians are not willing to accept higher patient loads as a result. Instead, the study suggests that these gains should be reinvested in improving work-life balance, reducing burnout, and strengthening patient-provider relationships.