The Evolution of AI in Anesthesia: What is Left for Anesthetists?

The practice of anesthesia has been transforming since its conception in the 19 th century. What began with ether-soaked rags as the first attempt at anesthetic delivery has become a science of precision. Thanks to breakthroughs such as vaporizers, intravenous induction agents, neuromuscular blockade, and target-controlled infusions (TCI),(1) today’s anesthesiologists dose more accurately, monitor more closely, and can tailor care to each patient. The practice continues to evolve and is never static. With artificial intelligence (AI) becoming increasingly embedded in modern life, it is no surprise that AI applications are being explored in perioperative care. AI could help anesthesiologists enhance safety, reduce physician cognitive workload, and individualize anesthetic delivery.

What AI Can Do Now

One of the most promising applications of AI in anesthesia is the closed-loop delivery system. These systems use real-time physiologic inputs—EEG patterns, blood pressure, and heart rate variability—to guide anesthetic drug infusion automatically.(2,3) Their aim is to maintain stable anesthetic depth with minimal fluctuation. Preliminary studies suggest such systems may reduce time to emergence and improve intraoperative consistency.(4,5) Though these systems are not yet common in daily practice, they show that automation can assist anesthesiologists in managing complex physiologic states.

One of the most promising applications of AI in anesthesia is the closed-loop delivery system. These systems use real-time physiologic inputs—EEG patterns, blood pressure, and heart rate variability—to guide anesthetic drug infusion automatically.(2,3) Their aim is to maintain stable anesthetic depth with minimal fluctuation. Preliminary studies suggest such systems may reduce time to emergence and improve intraoperative consistency.(4,5) Though these systems are not yet common in daily practice, they show that automation can assist anesthesiologists in managing complex physiologic states.

Another AI integration is predictive analytics. Algorithms trained on perioperative data can help forecast risks such as post-induction hypotension, delayed emergence, or even postoperative delirium.(6) These insights may allow clinicians to anticipate complications earlier, adjust plans proactively, and make more informed decisions. Additionally, AI tools that highlight evolving trends during surgery could reduce information overload and streamline intraoperative monitoring.(7,8)

A notable example of Canada’s involvement in anesthetic innovation is McSleepy, a fully automated anesthesia system developed at McGill University.(9) It monitors hypnosis via EEG, analgesia via blood pressure, and neuromuscular blockade, adjusting drug delivery accordingly through conventional infusion pumps.(9) While McSleepy has not seen widespread adoption, it demonstrates that AI-guided systems can work in real-time alongside anesthesiologists.

Limitations

Although AI in anesthesia continues to advance, it does not come without its limitations. These systems rely heavily on data quality; when exposed to biased or incomplete inputs, their outputs can be misleading.(10,11) Many AI models operate as “black boxes”—producing results without offering insight into their reasoning. In a high stake setting such as the OR, this lack of transparency can hinder trust and accountability. Patient perceptions also matter. Surveys suggest many individuals feel safer with a human provider at the helm, especially during vulnerable procedures such as surgery.(12) Moreover, privacy concerns are real: systems that constantly analyze physiological data must meet high standards for cybersecurity and data protection. These challenges reinforce the need for AI to be integrated with heavy consideration—with ethical oversight, informed consent, and clinician-led interpretation. To minimize these drawbacks, AI tools should complement—not replace—clinician expertise.

What Is Left for the Anesthetist?

With the rise of automation, many students wonder if AI will eventually replace their specialty of interest, including anesthesia. The short answer is not completely. The longer answer is AI systems can support drug titration, monitor trends, and even anticipate complications. However, they cannot replicate the core attributes that define clinical practice: nuanced judgment, adaptability in unpredictable situations, and human connection. Like what a preceptor tells you on a rotation: anyone can be taught knowledge, but there are certain human attributes that cannot be taught. Anesthesiologists do not simply manage physiology—they interpret the patient’s story, respond to surgical nuances, make ethical decisions in real-time, and communicate with teams and families during high-stakes moments, which takes incredible amounts of empathy and humanity. These are responsibilities no algorithm can fully assume. Even the most advanced AI tools must be supervised, adjusted, and interpreted by trained professionals. For example, a closed-loop system might maintain a target anesthetic depth—but deciding whether to alter that target based on surgical stimulation, patient comorbidities, or unexpected events remains a human decision.

For medical students, this means learning to work with technology, not against it, andstarting as soon as possible. A trainee who is well-versed in modern technology will be moresuccessful when technology inevitably becomes even more prominent in healthcare. However, ofcourse, developing clinical reasoning, communication skills, and ethical awareness is just asimportant as understanding how to use new tools. AI may help with monitoring more efficientlyor making predictions faster—but the responsibility, accountability, and art of medicine will still rest in our hands.

While difficult to predict where the specialty is headed, the one certainty is that the practice of anesthesia will inevitably continue to change. Anesthetists—and future anesthetists—should be at the forefront, ensuring each change is driven by an unwavering focus on patient safety and well-being.

REFERENCES

1. Kenny GN, White M. A portable target-controlled propofol infusion system. J Clin Monit Comput. 1992;8(3):179-182. doi:10.1007/BF01145171. Available from: https://link.springer.com/article/10.1007/BF01145171.

2. West N, van Heusden K, Görges M, et al. Design and Evaluation of a Closed-Loop Anesthesia System With Robust Control and Safety System. Anesth Analg. 2018;127(4):883-894. doi:10.1213/ANE.0000000000002663. Available from: https://pubmed.ncbi.nlm.nih.gov/29210791/.

3. Liu D. A Model-Based Architecture for Automatic Anesthesia Infusion. arXiv. Published online December 28, 2021. doi:10.48550/arXiv.2112.14144. Available from: https://arxiv.org/abs/2112.14144.

4. Cai X, Wang X, Zhu Y, Yao Y, Chen J. Advances in automated anesthesia: a comprehensive review. Anesthesiol Perioper Sci. 2025;3(1):3. doi:10.1007/s44254-024-00085-z. Available from: https://link.springer.com/article/10.1007/s44254-024-00085-z.

5. Xu C, Zhu Y, Wu L, et al. Evaluating the effect of an artificial intelligence system on the anesthesia quality control during gastrointestinal endoscopy with sedation: a randomized controlled trial. BMC Anesthesiol. 2022;22(1):313. doi:10.1186/s12871-022-01796-1. Available from: https://bmcanesthesiol.biomedcentral.com/articles/10.1186/s12871-022- 01796-1.

6. Kendale S, Kulkarni P, Rosenberg AD, Wang J. Supervised machine-learning predictive analytics for prediction of postinduction hypotension. Anesthesiology. 2018;129(4):675-688. doi:10.1097/ALN.0000000000002374. Available from: https://pubmed.ncbi.nlm.nih.gov/30074930/.

7. Naaz S, Asghar A. Artificial intelligence, nano-technology and genomic medicine: The future of anaesthesia. J Anaesthesiol Clin Pharmacol. 2022;38(1):11-17. doi:10.4103/joacp.JOACP_139_20. Available from: https://pubmed.ncbi.nlm.nih.gov/35706631/

8. Pham FM. Artificial Intelligence-Supported Systems in Anesthesiology and Its Standpoint to Date—A Review. Open Journal of Anesthesiology. 2023;13(7):140-168. doi:10.4236/ojanes.2023.137014. Available from: https://www.scirp.org/journal/paperinformation?paperid=126300.

9. Hemmerling TM. World first: researchers develop completely automated anesthesia system. McGill University Newsroom. Published May 1, 2008. Available from: https://www.mcgill.ca/newsroom/channels/news/world-first-researchers-develop- completely-automated-anesthesia-system-100263.

10. Sarker IH. Machine Learning: Algorithms, Real-World Applications and Research Directions. SN Comput Sci. 2021;2(3):160. doi:10.1007/s42979-021-00592-x. Available from: https://link.springer.com/article/10.1007/s42979-021-00592-x

11. Bellini V, Petroni A, Palumbo G, Bignami E. Data quality and blockchain technology. Anaesth Crit Care Pain Med. 2019;38(5):521-522. doi:10.1016/j.accpm.2018.12.015. Available from: https://pubmed.ncbi.nlm.nih.gov/30633992/

12. Riedl R, Hogeterp SA, Reuter M. Do patients prefer a human doctor, artificial intelligence, or a blend, and is this preference dependent on medical discipline? Empirical evidence and implications for medical practice. Front Psychol. 2024;15:1422177. doi:10.3389/fpsyg.2024.1422177. Available from: https://www.frontiersin.org/articles/10.3389/fpsyg.2024.1422177/full