AI Transforms the OR as Surgeons Navigate Complex Challenges

The landscape of modern medicine is continuously shaped by technological advancements, and artificial intelligence (AI) is emerging as a transformative force in surgical practice.

Building upon the increasingly high-tech nature of healthcare, AI promises to revolutionize how surgeons operate, document, and conduct research, ultimately enhancing patient care and streamlining professional workflows. This article provides an overview of current and emerging use cases for AI in surgery, offers practical examples of implemented tools, addresses crucial ethical situations that surgeons will need to navigate, and focuses on three primary categories of AI applications (i.e., ambient AI, prediction tools, and writing and research solutions).

Ambient AI: Automating the Clinical Encounter

One of the most rapidly evolving and widely adopted applications of AI in healthcare is ambient AI. These tools are designed to operate seamlessly in the background of clinical interactions, serving as virtual assistants or digital scribes. Their primary function is to enhance the efficiency and quality of patient encounters by automating many routine administrative tasks.

Ambient AI leverages advanced speech recognition capabilities to listen to physician-patient conversations and integrate directly with electronic health records (EHRs) to facilitate automated clinical documentation.

For instance, a system like Dragon Ambient eXperience (DAX) CoPilot combines ambient listening technology, Dragon Medical One speech recognition, and generative AI. Tools like DAX CoPilot (and similar solutions in the ambient AI healthcare space) can transcribe and organize an entire patient encounter, creating a templated note within the EHR by intelligently populating relevant sections with details from the conversation.

This capability represents a significant shift from traditional manual documentation, allowing clinicians to focus more intently on the patient rather than on concurrent note-taking. The increased attentiveness to patient needs is a direct result of AI managing routine tasks, lightening the cognitive load for the surgeon.

Studies have shown that burdensome documentation is a major cause of physician burnout.¹ Ambient AI offers a powerful solution to the long hours spent writing detailed notes, entering data into EHRs, and completing coding for insurance—all tasks that divert attention from direct patient care and other professional responsibilities. By automating the documentation process, ambient AI can significantly reduce this administrative burden and reduce a primary driver of burnout.

Finally, AI tools have the potential for producing higher-quality documentation.² Maintaining accurate and comprehensive patient records is a fundamental, yet time-consuming task in healthcare. Ambient AI tools can automate the creation of detailed, accurate clinical notes by passively listening to conversations or analyzing the clinician's actions in real time. This can lead to more thorough and consistent records, which are vital for patient safety and continuity of care.

Several tools are available on the market for ambient AI. These include the previously mentioned DAX CoPilot, Augmedix, and commercial solutions like the Limitless wearable pendant. Additionally, some institutions, such as The Permanente Medical Group, have developed home-grown solutions.

A critical consideration when evaluating these tools are data privacy and compliance with elevated privacy standards for health-related data, such as those set forth in the Health Insurance Portability and Accountability Act (HIPAA). Commercial options that do not have built-in EHR integration (i.e., Limitless Pendant) are generally not advisable for use in patient encounters, because most of these systems upload and store conversations in cloud-based platforms, leading to loss of control of the data and potential public disclosure of sensitive patient information and breaches of confidentiality. Surgeons must be acutely aware of these data security implications to protect patient privacy and adhere to professional standards.

Prediction Tools: Foresight for Enhanced Surgical Practice

Building upon the foundational applications of ambient AI, prediction tools represent another rapidly evolving domain for AI in surgery. These tools currently are an immensely active area of research and are designed to leverage vast datasets to forecast various outcomes, optimize resource allocation, and provide crucial decision support for surgeons and their teams.

Prediction tools can be broadly categorized into three main areas: risk prediction, resource utilization, and clinical decision support. While many of these tools are still experimental, those that have been implemented clinically have been met with mixed success but have shown immense promise.

Risk Prediction

One of the most intuitive applications of AI in surgery is the prediction of surgical risk. By analyzing comprehensive patient data, AI models can identify individuals at higher risk for complications, allowing for more informed preoperative planning and patient counseling. A notable example is the POTTER (Predictive Optimal Trees in Emergency Surgery Risk) calculator,³ which uses a machine learning backbone to predict the risk associated with emergency general surgery.

By employing data from the ACS National Surgical Quality Improvement Program (NSQIP^®) database as training data, the POTTER algorithm outperformed the American Society of Anesthesiologists classification and the ACS NSQIP calculator for predicting both morbidity and mortality.³ Originally, these findings were published in 2018, yet despite these promising performance statistics, POTTER and other similar tools have not yet been widely adopted across clinical practice, highlighting a recurring challenge in AI implementation.

Resource Use

AI-driven prediction tools also are proving invaluable in optimizing resource efficiency and allocation within surgical departments. These applications often operate somewhat "behind the curtain," streamlining operational aspects without direct patient interaction, which may contribute to their greater success in implementation compared to risk prediction and clinical decision support tools.

For instance, a research group in Italy has published extensively on using AI models to improve OR management.⁴ Similarly, a group in Israel has developed a model specifically designed to predict the duration of surgical cases, facilitating more optimal OR scheduling.⁵ These types of AI-based resource management models are increasingly being used at institutions worldwide, indicating a growing recognition of their practical benefits in enhancing efficiency and reducing operational bottlenecks.

Clinical Decision Support

The concept of clinical decision support tools has been present in healthcare for some time, preceding the development of AI, with earlier implementations often taking the form of pop-up alerts within EHRs. Many surgeons may recall experiences with these early clinical decision support tools, such as alerts suggesting evaluation for sepsis. However, these initial versions often faced significant clinical push-back. Reasons for this resistance included predictions that were not sufficiently accurate, a lack of contextual understanding in the algorithms, and the creation of additional clicks and distractions within the EHR, rather than streamlining workflows.⁶

However, newer generations of clinical decision support tools are being developed to overcome these limitations and often overlap with the risk prediction and resource utilization categories. For example, a research group in New York State has developed an algorithm capable of predicting the trauma activation level for pediatric trauma patients upon their arrival at the hospital, based on prehospital information.⁷ This algorithm serves as a direct clinical decision support tool for nurses who perform this critical triage task at many institutions.

Challenges

Despite these advancements and the clear potential of such tools, their widespread implementation remains a significant challenge. A primary hurdle is the absence of standardized pathways for integrating these new technologies into most healthcare institutions.

Unanswered questions persist regarding who is responsible for building or integrating these new tools into the EHR or a dedicated application; who is accountable for the ongoing maintenance and performance monitoring of the background AI model; and how institutions can effectively assess for potential biases within these algorithms,⁸ ensuring equitable and safe care for all patient populations. Addressing these fundamental questions will be critical to unlocking the full potential of prediction tools and embedding them seamlessly into surgical practice.

Writing and Research Tools

The landscape of scholarly work is increasingly benefiting from AI applications, with a burgeoning array of solutions designed to streamline workflows in research and writing, ranging from summarizing articles to assisting with complex systematic reviews and meta-analyses.

Article Reviews and Summaries

For tasks such as article reviews and summarization, popular AI tools like Notebook LM and ChatGPT are readily available.

These platforms allow users to upload a document and generate a summary or a critical review of its content. While seemingly convenient, their use for scientific review comes with significant caveats, particularly concerning privacy and security.

A major dilemma can surface due to the fact that most journals currently do not allow the use of AI tools when performing scientific reviews of articles.⁹ This restriction stems from at least one fundamental concern: when an article is uploaded to an AI service like OpenAI (which powers ChatGPT), the user effectively loses control of that information, which can lead to potential public disclosure of unpublished, sensitive data.

The risk is that an author's manuscript data, intended to remain confidential during the peer-review process, could inadvertently be used to train the AI model in the future. Such scenarios have already sparked numerous lawsuits concerning the unauthorized use of data for training large language models, a reality that underscores why surgeons should avoid such breaches of confidentiality when using these tools.

Systematic Reviews

The tools available for performing complex research tasks like systematic reviews are becoming increasingly sophisticated. Specific examples include SWIFT-Review, Rayyan, and ResearchRabbit with Rayyan being one of the most well-developed of these tools.

An illustrative example of the potential of systemic review solutions comes from Abigail Loszko, MD, who recently conducted simultaneous systematic reviews, one using traditional methods and another using only AI tools.¹⁰ While the AI method was not quite as thorough or well-developed as the traditional approach conducted by a research librarian, the study demonstrated significant time savings using AI tools. Furthermore, the AI solutions identified a reasonable number of additional papers that were not found using traditional search methods, highlighting the potential of AI tools to augment rather than simply replace established research practices.

Meta-Analyses

Moving to even more complex research tasks, AI tools also are emerging for performing meta-analyses, such as Elicit, Grapha, and DataSquirrel. Meta-analyses, however, are significantly "trickier" as they require AI tools to extract, pool, and analyze data from multiple sources, introducing a higher degree of complexity and potential for error. As a result, most of these tools are not yet ready for widespread use.

Given the considerable amount of manual data oversight that still needs to be performed to produce trustworthy results, it is currently safer to conduct these analyses manually. Nevertheless, the rapid pace of development suggests it will not be long before these tools are reliably incorporated into everyday surgical research workflows.

The applications of AI in surgery, from ambient AI to prediction tools and sophisticated research aids, are just a few examples of the transformative potential of this technology. Despite certain ethical considerations and current stipulations from journals that may require disclosure of AI use or prohibit it for scientific reviews, these tools are rapidly evolving. AI will soon become as ubiquitous in our professional lives as are the spelling and grammar checkers we rely on daily, with most common word processing tools already incorporating predictive text features built on AI frameworks.

AI tools represent the future of surgical practice and research. Becoming familiar with them undoubtedly will make certain aspects of your professional life easier. It is paramount for surgeons to remain acutely aware of the pitfalls, potential mistakes, and critical ethical considerations that accompany their use.

Thoughtful adoption, coupled with an understanding of their limitations and responsibilities, will be key to harnessing AI's full potential in enhancing patient care and streamlining surgical workflows.

Disclaimer

The mention of specific company names, products, or technologies in this article is for informational purposes only and does not constitute endorsement by the ACS.

Dr. Nicole Wilson is a pediatric surgeon at Oklahoma Children’s Hospital/The University of Oklahoma Health Sciences Center in Oklahoma City. She also runs a research laboratory focused on the intersection of engineering with surgical diseases and outcomes.

References

1. Budd J. Burnout related to electronic health record use in primary care. J Prim Care Community Health. January-December 2023;14:21501319231166921.
2. Balloch J, Sridharan S, Oldham G, Wray J, et al. Use of an ambient artificial intelligence tool to improve quality of clinical documentation. Future Healthc J. 2024;11(3):100157.
3. Bertsimas D, Dunn J, Velmahos GC, Kaafarani HMA. Surgical risk is not linear: Derivation and validation of a novel, user-friendly, and machine-learning-based Predictive OpTimal Trees in Emergency Surgery Risk (POTTER) calculator. Ann Surg. 2018;268(4):574–583.
4. Bellini V, Russo M, Domenichetti T, Panizzi M, et al. Artificial intelligence in operating room management. J Med Syst. 2024;48(1):19.
5. Azriel D, Rinott Y, Tal O, Abbou B, et al. Surgery duration prediction using multi-task feature selection. IEEE J Biomed Health Inform. 2024;28(7):4216–4223.
6. Harrison AM, Gajic O, Pickering BW, Herasevich V. Development and implementation of sepsis alert systems. Clin Chest Med. 2016;37(2):219–229.
7. Liu C, Chacon M, Crawford L, Polydore H, et al. Machine learning improves the accuracy of trauma team activation level assignments in pediatric patients. J Pediatr Surg. 2024;59(1):74-79.
8. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447–453.
9. Naddaf M. AI is transforming peer review—and many scientists are worried. Nature. 2025;639(8056):852–854.
10. Loszko A. Artificial intelligence in the systematic review process: Key takeaways from a comparative study on traditional versus ai-enhanced approaches. Oral Presentation. Presented at: New York Chapter American College of Surgeons 37th Surgical Symposium. May 2025. Lake George, NY.