How will AI change primary care?
The doctor-patient role has been evolving throughout history of society and AI will be the biggest change yet. What will we still need PCPs for?
I grew up with a primary care doctor that was seeing three generations of my family. It was not unusual for the doctor to ask about my sister and grandfather during the same visit. I was very lucky in that sense, because my doctor had polygenic risk data by treating my parents and grandparents before DNA sequencing was even a thing. My grandparents generation grew up with a village doctor that only visited when someone was dying and now, I don’t have a primary care provider anymore since I have grown to fully rely on AI. Similarly, over the past century, the doctor-patient relationship has continually evolved, reshaped by societal shifts, technological advancements, and changing patient expectations.
From the early 1900s to the 1930s, the family doctor was typically seen at home, paid in cash or even barter. The village doctor that my grandparents grew up with wielded immense authority, their advice rarely questioned. The Flexner Report of 1910 ushered in an era of formalized medical training in North America and cemented physician authority, reinforcing this patriarchal trust.
By the 1940s and 1950s, primary care shifted from homes into hospitals and offices. Employer-based insurance became widespread due to wage controls during World War II, institutionalizing medical care. Though doctors still retained authority, the relationship grew more transactional.
The 1960s saw the birth of “primary care” as a specialty, with Medicare and Medicaid expanding healthcare access. Family medicine emerged in 1969, rekindling continuity and lifelong patient-doctor bonds within a formalized setting.
In the 1970s and 1980s, managed care reshaped primary care again. PCPs became gatekeepers controlling costs and specialist referrals, sometimes straining patient trust. Meanwhile, patient-rights movements and informed-consent laws empowered patients, gradually shifting toward a shared decision-making model.
My most recent health plan was a HMO which meant I needed a referral for every specialist visit. Since I had freshly moved to New York City, I was randomly assigned a PCP by my new insurance plan. I wanted to do a sleep study and for that I needed a referral from my PCP. When I tried calling the office where my doctor was supposed to work, I was told that the next appointment is in three months. I could not be bothered to wait three months to get a referral so I paid for the study out of pocket. For most Americans, this is not an option and they will postpone or ignore necessary specialist visits.
By the 1990s, the internet introduced patients to an overwhelming amount of medical information. “Dr. Google” transformed the PCP’s role from unquestioned authority to interpreter and guide, responding to patients who arrived with printouts and questions. Google became the top-of-funnel solution people tried before needing to go to a doctor. There was definitely also a cost factor, since Google is free and people gladly tried to solve their problems for free before needing to want and pay for a PCP visit. Surprisingly, people searching their symptoms online did not lead to a decreased number of in-person visits, but it did lead to more informed discussions during those visits.
Today, ChatGPT can assist patients with routine tasks remarkably well. Have a late-night question about a common rash or cold symptoms? ChatGPT answers instantly, clearly, and without the inconvenience or expense of an office visit. In fact, a tool like ChatGPT is often better than a human at providing rapid, reliable, standardized responses to straightforward queries: 24/7, without fatigue or wait times. Furthermore, AI often has memory now, which means that it can follow your symptoms for days/weeks/months and give you a much better diagnosis than a PCP who only sees a snapshot of your health during a visit. Both methods however need good data in and the user needs to know which prompts to use and what data to include to get accurate symptoms. With more data, better reasoning models and improving IQ for foundational models, ChatGPT is overtaking a lot of traditional PCP responsibilities.
Yet, today's PCPs still have important responsibilities: preventive care like screenings, vaccinations, and lifestyle counseling; diagnosing and treating common illnesses; coordinating care among specialists; educating patients and advocating for their needs; and building long-term relationships based on trust and understanding.
Many of these tasks are already being replaced or significantly enhanced by AI. Routine tasks such as refill requests, reviewing normal lab results, scheduling appointments, and initial symptom assessments are now being handled faster and more accurately by algorithms. AI-powered scribes, like Nabla, Abridge and Nuance’s DAX, have freed physicians from hours spent typing into electronic health records, allowing them to reengage directly with patients.
Moving forward, AI will and should take on an even larger role, especially in preventive care and chronic disease management. AI-driven health platforms can continuously analyze data from wearable devices and health apps, providing personalized insights and timely interventions to prevent illness. AI's ability to track complex chronic conditions in real-time, predict flare-ups, and adjust treatment recommendations proactively can substantially improve outcomes and quality of life. The adoption will likely happen in two phases: first AI creating treatment plans that doctors need to approve and over time taking over simpler workflows completely. Just like in B2B SaaS - first you have the copilots and then come the agents.
This technological advancement will significantly expand access to high-quality healthcare, particularly benefiting rural and underserved areas where PCP shortages are common. Patients previously limited by geography or healthcare availability will have constant, expert-level guidance at their fingertips.
However, even the most advanced AI cannot replace certain human roles. Accountability, intuition, ethical decision-making, and building trust through genuine personal relationships will always remain uniquely human capabilities. Human doctors that have more longitudinal data about their patients can sense hidden anxieties, understand unspoken fears, and provide nuanced emotional support and reassurance during critical health moments. It’s easier for human doctors to rely on their gut feeling and push back to patients when they feel they are not being fully honest. It will take some time for AI to develop such instincts.
For purely data-driven clinical decisions, AI is already superior to doctors. I have been helping a friend’s parent understand how their chemotherapy is going and whether the sessions are improving their situation. For such a simple data crunching exercise, I feed the research model all data and I get a very detailed report how the biomarkers are changing and how the treatment is working. The report summary has always been very similar to the doctor’s comments, but the depth of the report would have taken a human a full day to compose. AI is already better than any human doctor, but because health is so sensitive, we still need a human in the loop for many clinical decisions.
The same applies to having the last call for clinically more dangerous procedures like high-risk prescription drugs. With enough real-life training data about clinical outcomes, AI can start to prescribes drugs with a great safety profile, but we want to always have the clinicians in the loop for more dangerous drugs and diagnosis. Where to draw the line legally will be a tricky challenge to solve.
As we embrace AI's potential, more and more control and decision making will move to the patient. They will always have the most data, be most invested and soon have the best models in the world to guide them in their health journeys. This does not mean that the human connection should be completely forgotten. Most Americans lose the connection to their PCPs after leaving their parents’ homes. You can keep your doctor if you don’t mind doing virtual visits , but physically moving around and changing jobs makes it challenging to keep the same PCP and develop a relationship with them. Right now we are getting the worse of both world, where we don’t have a connection to our family doctor anymore (thus loosing the data that doctors have about our parents and grandparents), nor can we use the data from our wearables and advanced diagnostics like polygenic risks from DNA sequencing. The primary care will need to be more mobile and data-led to give modern families a more holistic coverage.
Even if you are ok with sticking to your childhood doctor using televisits, it’s a lot harder to keep your doctor if your new employer insurance plan does not have them in the network. Good solutions here could be either people using direct primary care providers and pay out of pocket for their subscription. That’s kind of what OneMedical and formerly Forward Health provided - same PCP experience no matter where you lived. Most of their customers paid out of pocket for the membership and visits. It will be interesting to see the direct primary care evolve as AI will take over lots of their responsibilities. A likely scenario here is that the AI-led tools will become financially a lot more accessible and the out-of-pocket costs for the rest of the PCP interactions will be manageable to a larger part of the population.
My dream is that AI will enable the best of both worlds for primary care: more consumer-led healthcare, better access to primary care and a stronger bond between the PCP and patient going forward. Our laws will need to evolve to enable AIs to take some of PCP responsibility and free them from a lot of unnecessary work. Right now, we are focused on the copilots in healthtech and soon we can start expanding access to better primary care. Every American having access to a concierge like primary care service does not seem like a distant dream anymore. It’s an exciting time to be a consumer and builder in healthcare!