Will AI Replace Physicians?
- Nisha Mehta, MD
- 7 days ago
- 9 min read
Updated: 3 days ago
As AI capabilities expand by the second and reports of AI outperforming physicians on standardized exams come out, many doctors on our physician communities are scared that AI will replace them or cause them to lose their jobs. While we do think that AI will continue to play a large role in healthcare innovation, we are skeptical that AI can fully replace physicians. In fact, our hope is that AI will be able to support healthcare workers and streamline efficiency such that physicians can focus in more on the heart of what they do, rather than obviate the need for physicians. Below, we’ll cover the reasons why we feel AI can’t replace physicians, ways that AI may actually help physicians, provide guidance on how to view AI, as well as give steps physicians can take to protect their careers as the clinical applications of artificial intelligence increase daily.
Disclosure/Disclaimer: This page contains information about our sponsors and/or affiliate links, which support us monetarily at no cost to you, and often provide you with perks, so we hope it's win-win. These should be viewed as introductions rather than formal recommendations. Our content is for generalized educational purposes. While we try to ensure it is accurate and updated, we cannot guarantee it. We are not formal financial, legal, or tax professionals and do not provide individualized advice specific to your situation. You should consult these as appropriate and/or do your own due diligence before making decisions based on this page. To learn more, visit our disclaimers and disclosures.
Article Navigation
Can physicians be replaced by artificial intelligence (AI)?
We believe the short answer is no, but of course, we need to justify that statement given the growing number of people claiming that AI will make many physician jobs obsolete. The fact is that being a physician is about more than just having book knowledge - it’s about experience, passion, empathy, and the ability to apply knowledge to unique situations.
Couldn’t you argue that AI can quickly have more knowledge and ‘experience’ than a physician?
Yes, to play devil’s advocate, while physicians see thousands or tens of thousands of cases in their training and career, technology allows us to train artificial intelligence models on literally millions of cases in a miniscule fraction of that time.Â
However, the unique thing about medicine is that it’s not just a science, it’s also an art. How you apply the knowledge to each individual case matters. When physicians make decisions, they’re judging what is best for that particular patient’s medical issue as well as other factors such as patient preferences, family situation, support structures, cultural norms and intricacies, socioeconomic challenges, and more. They also need to communicate those findings, options, and plans in a way that the patient and their loved ones understand, and help them through their decision making process. Bias in the way that data is chosen to train models can also exacerbate health disparities or promote certain studies or datasets that may not apply to a particular patient (arguably, this can also happen with human interactions).
Perhaps more importantly, it’s hard to replicate the ‘spidey’ sense that a physician develops after years of training that allows them to know when to dig deeper when something doesn’t add up, or to read body language from a patient or their loved ones when putting together a treatment plan. Yes, wearable technology, algorithms looking for deviations, and more products are constantly being created that may alert the medical team to changes, and even suggest obscure diseases that may not be top of mind or on the differential for physicians. These are great, but tools to augment clinical decision support, rather than be the only metric.
AI does not currently have the depth to do all these things. Physical exams, spidey senses, and an overall synthesis of the situation is still hard to completely replicate by algorithmic medicine. Is it possible that it will in the future? Potentially, but we think there’s still several obstacles to AI replacing physicians completely.
What are some other reasons AI can’t replace doctors, or obstacles to medical care completely being dictated by AI?
Healthcare is human
How many people want to talk to a computer or an app when something is seriously wrong? It’s hard to imagine that the compassion and trust people place in their physicians and clinicians can be completely replicated in a way that would provide comfort when something is truly wrong. Imagine being told that you have cancer or that your family member had passed away by an app on your phone. Imagine trusting a robot to do your dangerous surgery. Empathy and trust are key parts of the doctor-patient relationship that are hard to replicate.
Who will take responsibility for malpractice and when something goes wrong
If you’re cynical, you’ll also wonder who is going to take responsibility when something goes wrong with a patient, particularly if AI got things wrong. Malpractice lawsuits are a part of everyday life for physicians, and the litigative nature of the current healthcare landscape often delivers large verdicts in favor of patients or patient families. Who will patients blame if the AI messes up? Will they go after a multi-billion dollar AI company? The person who wrote the algorithm? The hospital system that uses the AI?Â
Plaintiff's lawyers would have a field day going after some of these bigger pockets, so the cynic in us feels they will always have physicians signing off on care recommendations and algorithms.
Inability to do procedures
Always performing surgery or procedures in the same way would inevitably run into issues with devastating consequences. Anatomical variation and aberrant anatomy, adhesions or scarring, complications, and other unique factors often require a surgeon or other physicians to change approaches and make quick decisions in the operating room or procedural setting. It's hard to imagine a situation in which AI could safely replace humans no matter how advanced technology gets.
The variability of human biology and experiences
As alluded to in the last point, every human has unique genetics, unique anatomy and physiology, and unique life experiences and circumstances, which mean that the same treatment algorithm rarely applies across patients when there are more complex cases. Sure, AI could treat a simple case of the flu, but it would be much harder to imagine AI being able to spit out an answer to a complex case discussed in tumor board which factors in path results, a patient’s genetics, their family history, their preferences for treatment, where they live and what medical care they have access to, what has already been tried and with what results, and what experimental trials are currently out there that they may be candidates for.Â
Ethical considerations and moral conduct
Complicated situations arise in medicine all the time. End of life decisions, properly getting patient consent and making sure they truly understand risk and benefits, and other moral dilemmas in medicine come up constantly. Very few of these are black and white, and the practice of medicine involves navigating the grey areas in a way that protects patients, their rights, and their dignity.Â
Patient privacy concerns
Can you imagine feeding your legal dilemmas into an AI system that was constantly storing and learning from your answers. There’s a reason people trust their lawyers to protect them. Similarly, many people may be very uncomfortable with the idea of providing their most sensitive information in a way that could be discovered by others. Physicians deal with some of the most personal information about people, and many patients would not tell their information to a big box company that was recording everything with privacy policies that they can’t understand.
Practical barriers to implementation and integration to access multiple sources of data
The healthcare ecosystem is made up of so many different products that don’t always play nicely with each other (different EHRs, order systems, legacy patient information and charting, pharmacy records, remote patient monitoring, etc.). Having an AI product that could access all of these records across different systems in the country to get an accurate view of the patient’s history and incorporate it to appropriately make decisions seems incredibly difficult, not to mention expensive for smaller health systems or practices.
Need for innovative ideas to advance the field
In addition to providing medical care, physicians advance the field based on their observations of needs and problems within the healthcare system and patient care. While tools can be built to address issues, we are not yet at a point where AI can come up with the ideas and run a clinical trial without help. Physicians who understand the human body as well as the practicalities of patients will always be needed to come up with innovative ideas.
So, what are ways that AI is likely to influence the field (or may change physician jobs)?
Regardless of the points above, there is no question that artificial intelligence will significantly impact the medical field. This may be by tools that improve efficiency, support clinical decision making, help with early detection, pre-screen patients, help with billing and coding, or even provide preliminary management plans for patients.Â
You could make the argument that by improving efficiency and doing some of the preliminary work, physicians may be able to see more patients, thus eliminating the need for so many physicians. However, with the Baby Boom generation retiring, patients living longer with chronic disease, and a general physician shortage that is being exacerbated by physician burnout and changing physician demographics, the demand for physicians may remain larger than the supply.Â
Should doctors be scared of AI, and what ways should we be looking to it to help us?
AI is coming, whether physicians like it or not, so the better approach is to figure out how it can help you and your patients. So much of what is being built can help you to get paid better, improve efficiency, and help you make better, more informed, decisions for your patients, thus improving patient outcomes.Â
Some great examples include:
OpenEvidence is the leading AI-powered medical information platform for physicians. It is currently used in over 7000 hospitals and care centers, and over 150 countries, serving millions of conversations per week. It is free and unlimited for physicians, and built from the ground up to give trustable, verifiable information. Every response is grounded in the high quality medical literature, spanning everything from clinical trials to meta-reviews to guidelines. Sign up today through our affiliate link.
Abridge, founded by one of our physician members, offers members 50% off AI scribing services, bringing the price down to only $99 dollars per month. You can try this free for one month without inputting payment information through our affiliate link with code PSG50, after which you will receive the 50% off price.
How can physicians protect their roles in the face of more AI tools looking to decrease the reliance of the healthcare system on physician labor?
Emphasize the doctor patient relationship
We suggest getting back to the basics. The heart of healthcare is the doctor-patient relationship. As long as patients get more from an interaction from you, whether it be empathy, drawing in their history and personal circumstances, or a good bedside manner that explains options and diagnoses clearly to them, keep doing the things that a robot cannot replicate. Build trust with your patients such that they value your opinions and experience, and feel more comfortable with talking to you than a robot.
Incorporate procedures into your practice
This is one thing that we are a long way off from replacing, as referenced above in the example of how surgeons need to approach each case differently. These are skills that are very hard to replace by an automated process.
Provide the most helpful information that you can
Certain specialties are going to adapt AI at a faster pace than others - for example, radiology will likely be impacted by AI generated models that can recognize patterns and interpret images. What will distinguish the radiologists interpretation from the AI generated report will be the ability to put the findings into context with the other history from the patients.
Develop areas of niche expertise
Having very specific expertise in a disease state that is more of a zebra, a specific procedure, a specific patient profile, etc, will keep you positioned as a key opinion leader in that field. While AI may also be able to see the same number of cases, your ability to incorporate your experience in something that the AI model may not have thousands of corollaries to will set you apart.
Function as a leader of the healthcare team
Continually show your patients and administrative staff how you provide value by bringing together all of the pieces of information and services to come up with individualized care plans unique for every situation.
Conclusion
AI is, and certainly will continue, changing the healthcare ecosystem at an alarming rate. However, instead of being fearful of it, physicians should remember what makes them unique and irreplaceable, as well as the obstacles and weaknesses of AI that will preclude it from taking their jobs. The better approach is to find ways to have AI help you to be the most effective and efficient physician as possible, and allow AI to take things off your plate so that you can focus on the heart of healthcare, the doctor-patient relationship.
Additional health tech resources for physicians
Explore related PSG content:
You can also sign up for our entrepreneurship and health innovation series for alerts on free events.