In our contemporary world, the healthcare environment is undergoing continuous transformation, which results in both incremental improvements and disruptive innovations (1). These changes are reshaping clinical workflows, altering provider responsibilities, and redefining patient and caregiver experiences (1, 2). One of the most significant technological advancements driving this transformation is the widespread adoption of Electronic Health Records (EHRs). An EHR is defined as a system for collecting and storing patient health data in a digital environment (3). This data typically includes demographics, diagnoses, medications, vital signs, laboratory results, medical history, immunization records, and radiology reports (4, 5). The implementation of EHRs has shown to improve healthcare service quality by reducing medication errors, promoting adherence to clinical guidelines, and enhancing operational efficiency (6, 7).

Despite the many benefits promised by EHR systems, their implementation has also introduced substantial challenges that necessitate critical evaluation. One of the main issue among these is the increased workload associated with EHR use, as healthcare professionals are required to spend significant time documenting, reviewing, and summarizing patient data (8–11). This administrative burden is not only time-consuming but has been identified as a key contributor to healthcare professional burnout (12), a psychological syndrome that develops in response to prolonged occupational stress (13). Burnout has been linked to adverse outcomes for physician well-being, increased workforce attrition, and compromised patient care, including reduced care quality and safety (2, 12, 14, 15). Burnout can lead to serious consequences for healthcare professionals, particularly physicians, among whom suicide rates are twice as high as in the general population (16, 17).

High workload is a well-recognized occupational stressor directly affecting care quality and patient outcomes. Prior studies have shown that administrative tasks, particularly those related to EHR use, substantially increase staff workload and time pressure (18). For example, in ambulatory care settings, physicians spend approximately 49% of their time on EHRs and desk work, compared to just 33% on direct clinical interactions with patients and staff (1, 19). This imbalance highlights the extent to which digital documentation demands interfere with patient care. Furthermore, recent surveys show that up to 70% of clinicians experience stress linked to the use of health information technology, with EHR-related stress emerging as an independent predictor of burnout (20). In this regard, AI-driven solutions are increasingly being explored as a means to alleviate the EHR-associated burnout.

Over the past decade, Artificial Intelligence (AI) has evolved from a theoretical concept into a practical and increasingly integrated component of modern healthcare systems (21). The rapid and exponential growth of AI applications introduces an important opportunity to address long-standing challenges in clinical practice, including clinician burnout and increased workload (21). AI encompasses a range of advanced computational methods such as Natural Language Processing (NLP), deep learning, intelligent robotics, and context-aware computing (22). Unlike traditional analytics, which operate on predefined rules, AI systems possess the ability to learn from historical data, adapt over time, and simulate human cognitive functions (22, 23).

In healthcare, AI has already demonstrated utility in various domains, including automating administrative workflows, enhancing diagnostic accuracy, supporting clinical decision-making, designing personalized treatment plans, and guiding robotic-assisted surgeries (23–26). When integrated with EHRs, AI technologies can facilitate data entry, retrieve relevant clinical information, and even transcribe patient-clinician interactions in real time (9, 27). These capabilities hold substantial promise for reducing the time health care professionals spend on manual documentation, thereby improving workflow efficiency and potentially mitigating the risk of burnout. While AI’s clinical applications continue to expand, its role in alleviating EHR-related burnout remains an underexplored yet promising area of investigation that necessitate further study (28).

The articles included in this systematic review were published between 2019 and 2025. Most originated from the United States (n = 6), with two additional studies conducted in Canada. The majority targeted healthcare professionals, primarily physicians, with sample sizes ranging from 10 to 162 participants. Across the eight studies, four categories of AI-based interventions were identified: Ambient AI scribes (n = 4), Clinical Decision Support Systems (CDSS; n = 1), Large Language Models (LLMs; n = 2), and Natural Language Processing (NLP) summarization tools (n = 1). These interventions were evaluated using a range of study designs, including prospective quality improvement studies, observational studies, mixed-methods evaluations, pre-post survey designs, comparative simulation-based analyses, and analytical cross-sectional studies.

To synthesize the findings thematically, the studies were categorized not only by technology type but also by the specific EHR-related challenges they addressed. Two main subproblems emerged as contributors to healthcare professional burnout: documentation burden and inbox management burden. Documentation burden was defined by excessive time spent on clinical notes, inefficient workflow integration, and reduced time for patient interaction. Six studies addressed this issue, including those focused on ambient AI scribes, a CDSS tool, and one LLM-based intervention. The second subproblem, inbox management burden, stemmed from high volumes of patient portal messages, poor triage systems, and a lack of time-saving tools. Two studies in the review used NLP and LLM technologies to address this challenge. Figure 2 presents a visual classification of the included studies according to these problem domains and AI intervention types.

This systematic review identified two primary subdomains contributing to EHR-related burnout among healthcare professionals: documentation burden and inbox management burden. These subdomains represent important targets for AI-based interventions, which this study explores as potential solutions to alleviate burnout. Prior studies that focused on quantifying and categorizing the sources of EHR-related workload further support these findings. For instance, Arndt et al. conducted a time-motion study using EHR event log data and observed that primary care physicians spent nearly 6 h per workday interacting with the EHR, with documentation and inbox tasks accounting for 44.2 and 23.7% of EHR time, respectively (8). Their taxonomy outlined 15 EHR-related activities and emphasized documentation, chart review, and inbox management as the most time-consuming (8). Similarly, a study reviewed the literature on physician burnout and classified EHR-related contributors into several categories, including documentation and clerical burdens, cognitive overload, electronic messaging volume, and complex usability issues (18).

Research shows that EHR documentation burden is a major contributor to healthcare professional frustration and burnout (12). Many healthcare professionals are unable to complete documentation tasks during clinic hours. As a result, they use personal time to finish their work (9, 41). This increased workload contributes to reduced job satisfaction (14), less time for patient interaction (2), and may negatively impact the quality of care provided (12). To address this issue, this review identified four AI interventions, including ambient AI scribe technologies, CDSS, and LLM and NPL based technologies. Among the AI-based interventions identified in this review, ambient AI scribe technologies emerged as a promising approach to alleviating EHR documentation burden. These tools, such as DAX Copilot, Abridge, ScribeMD, and others, advantage real-time speech recognition and NLP to transcribe clinician-patient conversations simply, automating the generation of clinical notes. Across multiple studies, these systems demonstrated evident benefits including reduced after-hours documentation, improved workflow efficiency, enhanced clinician satisfaction, and even modest reductions in self-reported burnout. For example, Shah et al. and Owens et al. found improvements in usability and task load (33, 34), while Albrecht et al. reported increased satisfaction and reduced after-hours charting (35).

While ambient AI scribe technologies offer promising solutions to alleviate documentation burden, recent findings caution against early adoption of AI-generated notes in clinical practice without rigorous evaluation. A study employing standardized simulated patient-provider interactions found that clinical notes generated by ChatGPT-4 failed to meet acceptable quality standards for clinical use (2). Notably, the study revealed substantial variability in error types, including both errors of omission and commission, which were not confined to specific sections of the SOAP note (2). Moreover, inconsistencies in error patterns across identical case replicates suggest that the model introduces an element of unpredictability, which makes it difficult for healthcare professionals to anticipate or correct errors reliably (2). This unpredictability presents a risk to clinical oversight and safety. It demonstrates the need for rigorous quality assurance before integrating AI technologies into routine documentation workflows (2). Integration into existing clinical workflows also presents logistical and technical difficulties, and the rapid pace of technological evolution may overtake institutions’ ability to adapt effectively (2, 9, 27). Moreover, most current studies relied on short-term, subjective measures (e.g., surveys), limiting their generalizability and robustness. The lack of patient perspectives and validated outcome metrics further constrains the understanding of their true impact.

While the majority of studies in this review focus on physicians, it is critical to recognize that nurses are also significantly impacted by EHR-related documentation burden. Nurses represent one of the largest group of EHR end-users in healthcare and routinely document 600–800 data points per 12-h shift, equating to roughly one data point per minute (14, 42, 43). Despite this, their experiences are underrepresented in burnout research and AI intervention design. A study by Gesner et al. provides strong evidence that documentation burden in nursing is positively correlated with emotional exhaustion and depersonalization, two key dimensions of clinician burnout syndrome (14). Moreover, the study identifies poor EHR usability as a significant contributor to documentation burden and calls for improved system design tailored to nursing workflows (14). The authors emphasize that while AI-based interventions such as clinical decision support and voice recognition tools hold promise, nurse-centered design and evaluation are essential to ensure these technologies effectively reduce burden without introducing new risks or inefficiencies (14). These findings highlights the importance of expanding the scope of AI research and implementation to include nurses, not only to alleviate burnout more equitably across professions, but also to improve documentation efficiency and the overall quality of patient care.

In examining the use of CDSS to address EHR documentation burden, Laing et al. reported valuable improvements in documentation efficiency and clinical decision-making (37). However, their study did not explicitly assess the impact of CDSS tools on clinician burnout, which leaves a gap in understanding the psychological and workplace satisfaction dimensions associated with these technologies. In parallel, LLMs have also been explored as tools for generating clinical documentation. Barak-Corren et al. evaluated the use of LLMs, specifically ChatGPT, in emergency medicine settings and found substantial gains in documentation efficiency and clinician productivity (38). Nevertheless, the study also highlighted critical concerns related to the reliability and clinical validity of AI-generated content. This highlights the ongoing need for clinician oversight to ensure both patient safety and documentation accuracy (38).

Inbox management emerged as the second major subdomain contributing to clinician burnout in this study. The volume of patient portal messages has grown dramatically in recent years, a trend significantly accelerated by the COVID-19 pandemic (39, 40). This increase in asynchronous communication has placed increasing pressure on healthcare systems and providers. Consequently, inbox management burden contributes to cognitive overload and emotional exhaustion. Studies indicate that the rising volume of patient messages can undermine clinician well-being, identifying message triage and response as key stressors (9, 39, 40). In addition to the increased volume, poor EHR inbox design, inefficient workflows, and inadequate prioritization mechanisms further increase the burden, leading to issues such as notification fatigue and missed critical information (9).

To address the growing burden of inbox management, research suggests that both technological and organizational strategies may be effective (9, 44, 45). Team-based care models, which distribute messaging responsibilities across clinical staff, have shown potential in reducing clinician burnout (9, 44). In addition, improving the design and usability of EHR inbox systems, particularly by involving frontline clinicians in the redesign process, can enhance workflow efficiency and reduce cognitive overload (9, 46). Early studies indicate that combining these approaches with quality improvement methods may lead to more sustainable and clinician-friendly communication systems (9, 39, 40). However, some studies have found that traditional strategies for managing inbox burden are no longer sufficient, which highlights the need for scalable and innovative solutions to support clinical communication without compromising safety or empathy (39, 40).

This review identified AI-based approaches that focus on alleviating inbox management burden. These include the use of LLMs and NLP tools. For example, researchers at Stanford Health Care implemented a GPT-4 system to generate draft replies to patient messages, which aimed to reduce clinician workload and cognitive strain (39). Similarly, Yang et al. developed a BERT-based NLP model that automatically prioritizes high-acuity patient messages with the aim of reducing the risk of missed critical information and improving triage efficiency (40). While AI tools show promise in reducing inbox management burden, they also raise important concerns. Research points out the risk of reduced personalization in AI-generated responses, which may negatively affect patient-clinician relationships and the overall quality of care (39). Similarly, another study reported challenges regarding the completeness and clinical accuracy of AI-generated triage outputs (40). Their findings indicate the need for a hybrid approach that combines automated systems with human oversight to ensure clinical safety and data reliability (40).

Moreover, several additional concerns related to the integration of AI remain to be addressed. The use of AI in healthcare raises critical ethical and practical challenges that warrant careful consideration. A primary concern is patient data privacy, as questions persist about the extent to which AI systems can reliably safeguard sensitive health information (4, 36). This underscores the importance of data governance, particularly when AI tools are developed, owned, or managed by private entities. In such cases, it is essential to establish clear guidelines regarding who can access, use, or potentially monetize patient data (2, 9). Another pressing issue is the phenomenon of AI hallucinations, wherein systems generate inaccurate or misleading outputs (36, 38). In clinical settings, such errors could seriously compromise patient safety (36, 38). Addressing these concerns is essential to ensure that AI tools are implemented ethically, safely, and effectively within healthcare workflows.

This systematic review demonstrates that AI technologies integrated into EHR systems hold considerable potential to alleviate burnout among healthcare professionals. Across the studies reviewed, AI interventions, particularly ambient AI scribes, CDSS, and LLM/NLP technologies, showed meaningful improvements in documentation efficiency, reduction in after-hours workload, and enhanced task management. These technologies address two primary contributors to EHR-related burnout: documentation burden and inbox management. However, while the findings suggest positive impacts on clinician workflow and, in some cases, self-reported burnout, the evidence also highlights the need for caution. Concerns regarding the accuracy, reliability, and personalization of AI-generated outputs highlights the importance of continued clinician oversight. Future research should prioritize longitudinal studies, diverse clinical settings, and robust outcome measures to assess the sustained impact and safety of AI-based interventions. Overall, AI offers a promising way to support healthcare professional well-being when thoughtfully integrated into EHR workflows.