31 October 2025
Dear Interested Readers,
A Potpourri: SNAP, Dental Health, and AI In Primary Care
My original intent this week was to write about the disaster that many of my neighbors and millions of Americans are experiencing with the cessation of SNAP payments today. The issues are complicated, and the response to the loss of support will vary from state to state, and in my state, from town to town. This morning, I participated with 31 other community members in a strategy session that involved the two nonprofits I support (as a board member, donor, and volunteer), as well as my church’s leadership (I serve as the moderator of the church), leaders from other area churches, the local hospital, our regional VNA, and other civic organizations. In all, there were 32 concerned participants on the call. The New York Times has reported this morning that a federal judge in Massachusetts is considering whether it is illeagal for the Trump administration to withhold funding from SNAP when there are some emergency funds available. We will see.
The outcome of our meeting was that there are no new sources of local or state funds that can be dispersed to individuals. We will focus our collective efforts and philanthropy on the food banks in our area. I expect that organizations and churches with a mission to provide emergency funds to individuals and families in need will do their best to assist families and individuals in need with emergency cash and gift cards.
There is no way to replace in a timely fashion most of the financial losses that our most vulnerable neighbors will experience. In my volunteer work, I have two “client families” that are already food bank customers and also depend on SNAP supports of over $700 a month to get by. A working member of one of those families is a federal employee whose pay before the shutdown was inadequate to support a large family. I am very worried about how they will weather this storm. I hope that these short-term losses will ultimately help repair some of the long-term damages to the federal “safety net” programs, which are sure to occur in the aftermath of the passage of the “One Big Beautiful Bill (OBBB)” and Congress’s failure to renew support for the ACA’s insurance exchanges.
Earlier this week, while I was contemplating how to address all of this complexity, I had a dental appointment. As I have rolled into my ninth decade, I have had my out-of-pocket dental expenses soar to more than $10,000 a year. It is easy to understand why so many Americans have dental problems. Dental work is expensive, which means that many dental problems go unaddressed until teeth are lost. Medicare doesn’t cover dental work. Medicaid coverage is limited because many providers don’t accept the low reimbursement that Medicaid offers. My wife and I had dental insurance, but we canceled it because we calculated that the premiums were more than the benefits we received. It is no surprise to me that most of the SNAP and Medicaid beneficiaries I know have unaddressed chronic dental problems or no teeth at all. Dental health is a crucial component of our overall well-being, and our deficient system of care does not effectively address it.
My dental problems stem from the combination of a congenital dental defect, teeth lost playing football, and the deleterious long-term impact on my gums from wearing a dental plate for over ten years that I could leave in the locker room during practice and games sixty years ago. Several teeth had to be removed last October before my total knee surgery was done in early December. This last year has been a tedious one of multiple appointments during which I had bone grafts and the eventual installation of the infrastructure of implants. I had been looking forward to Tuesday for a long time because Tuesday was finally the day when my periodontist decided I was ready to be “sized up” for my new teeth. In the past, when other teeth were replaced, this step involved the dentist making a plaster impression of my teeth and selecting colors, much like choosing a paint color when you decide to repaint your living room. A close inspection of my variously acquired caps and previous implants reveals that there is very little uniformity to the color of my teeth. The only consistent factor is that they are all stained by coffee.
Over the last twenty years, as I have had more interactions with dentists, I have developed the sense that dentists have been ahead of medicine in terms of introducing practical technology into everyday practice. Fifteen years ago, my dentist in Brookline, Massachusetts, who was also my patient, had an EMR that seemed to me to be equal to or better than the Epic edition I was using at the time. The technology in the office of the periodontist in Concord, New Hampshire, who has been managing my slow dental descent for the last six years, is impressive. It is clear from the charges for these services that his significant investment in that technology must be recovered.
In the division of dental labor, the periodontist installs the bone grafts and implant posts, but my local general dentist is responsible for the teeth that will fill the gap in my smile that has driven my wife crazy for the last 13 months. I have refused to use the temporary “flipper” that I was given except during a few social occasions, even though it was a perfect cosmetic solution. I just didn’t like it.
To my surprise, my dentist did not proceed as I expected. There was no plaster. He turned on his computer, which was attached to a wand. He passed the wand over my gums and teeth while I watched a fantastic representation of the interior of my mouth and my teeth appear in living color on his screen. That was it. My dentist is in his mid-fifties, so it occurred to me that this technology was not developed while he was in dental school; he had to learn how to incorporate it into his skills while he was managing his busy practice. Access is a big issue in dentistry, just as it is in primary care. Acquiring a new skill requires time out of the office. The time needed for training must further compromise access to him for appointments. Like most PCPs, he has an access problem. It often takes months to secure an appointment.
As I was questioning my dentist about this new technology and how he had learned to use it, I was simultaneously impressed by his openness to innovation and thinking about how inefficiently we incorporate new technologies into the general practice of medicine. Moving to an effective EMR required training and was an uphill climb for many older physicians and staff. There was justified resistance from those who couldn’t type. Many who resited saw the EMR as part of a larger plot by medical administrators to turn them into professional slaves who would forever be asked to increase the pace of revenue generation. I know that some excellent doctors retired rather than accept the role of a medical “imputter.” Hopefully, the challenge of adapting to new technologies and innovations will never end. We need innovation, but innovation always has a learning curve. Perhaps the biggest technological challenge to healthcare over the next ten years, which will surely meet significant resistance, will be the new era of AI-assisted practice.
I was thinking about the issues related to innovation while driving home from the dentist’s office. When I arrived home, I began searching the Internet for information about the introduction of AI to medical practice. I discovered a recent podcast, “Can AI Solve Primary Care?“ — NOS Episode 3.2, from the series produced by the New England Journal of Medicine entitled “Not Otherwise Specified.” The series is hosted by Lisa Rosenbaum, MD, of the Smith Center for Outcomes Research at the Beth Israel Deaconess Medical Center in Boston. She is also the “national correspondent” at the NEJM. On the AI podcast, she interviews Stephen Lin, M.D., Chief of General Medicine at Stanford and a leader in the research efforts to understand the issues arising from the introduction of AI support in primary care. It is an enjoyable 45-minute conversation that I encourage you to hear.
The conversation begins with an interesting question from Dr. Lin:
Is human, human plus AI, or AI alone better at making decisions?
You may be reluctant to accept the answer, which is “AI alone.” Dr. Lin helps us a little by continuing:
It’s paradigm-shifting. It’s frustrating. It provokes anger and resistance. And of course, medicine is not just diagnosing things. We all know that. But it does make you think: our worry and resistance about adopting AI, as well-intentioned as they are, are we actually harming patients if we are inferior to the new tools that we are developing? It’s very thought-provoking.
The issues discussed in the podcast address the challenges that make it difficult to be a primary care physician, as well as the question raised by some non-PCP physicians: whether we really need primary care. Some suggest that an AI-equipped nurse could fill the role of a primary care doctor. For those of us who are no longer practicing, or who never did practice, a big surprise is that AI is already a transformative reality in practice. I used my ChatBot to find out about the AI tools that the latest editions of Epic offer and what will be available in the near future. I was surprised by the Chatbot’s answer. Please note that the names Penny, Emmie, and Art are the names that Epic has applied to financial functions, patient portal functions, and practice support functions like an AI scribe.
Here’s a table of major AI-related tools from Epic Systems Corporation, their target users, functionalities, and release status — useful for assessing relevance for your organization.
| AI Agent / Feature | Target User | Key Functionality | Status / Availability |
| Penny | Revenue-cycle staff, coders, billing operations | Suggests coding, drafts appeal letters, automates claims follow-up | Currently available/pilot for selected clients. Healthcare IT Leaders+1 |
| Emmie | Patients (via MyChart) | Chatbot for lab results explanation, appointment suggestions, screening prompts | Planned for release later in year; full features by late 2026. Archyde+1 |
| Art | Clinicians | Ambient scribe, visit summarization, clinical insights, order suggestions | Development underway; note-taking full capability expected early 2026. Archyde |
| AI Charting / Native Scribe | Clinicians | Automatically generate or assist documentation directly in Epic, reducing manual entry | Announcement made; “limited use” expected around early-2026. Digital Health |
| **Cosmos AI / CoMET | Researchers, data scientists, clinicians | Large-scale predictive models using 300 M+ patients’ data; research-insights, look-alikes, rare-disease growth charts | Active research stage; contributions and early module releases underway in late 2025. Healthcare IT Today |
During the conversation, Dr. Lin provided in-depth descriptions of the role of AI in various aspects of clinical practice, ranging from routine office management to the complexity of summarizing medical histories and drawing diagnostic conclusions with therapeutic suggestions. He sees an AI Bot as capable of many of the onerous tasks of practice. Currently, it is being widely used to augment charting, which reduces the enormous burden of clinical documentation, and in some places is beginning to do “In Box” management and much of the work associated with “patient portals.” As Dr. Lin explains:
Steven Lin: One particular issue is just the burden of clinical documentation. Writing charts is a really big time suck for not just family medicine, but really all specialties. And there’s plenty of studies showing how for every 1 hour we’re actually spending in front of patients, another 2 hours are spent in front of the computer. And a large portion of that is just writing notes — which is important but we’re doing that in our pajama time, once we get home from work, when we have to work with our children on their homework, we’re writing notes and not being present. A lot of doctors are writing notes in the room and not looking at their patients, unable to pay eye contact or be there physically and mentally and be present for your patients. And so that’s a really, really big problem that’s sort of universal.
One of the most well-established tools now that I think many health systems have adopted or are about to adopt is AI scribes…
That statement opened one of the most interesting parts of the conversation because Dr. Rosenbaum countered with:
So if you were to offer me a scribe … I would say no thank you. And the reason for that is because writing is how I think. And so one concern I always have when we start talking about offloading this task to AI scribes is the act of synthesis.
The potential for AI to erode clinical acumen and skills was a topic discussed by Dr. Eric Topol of the Scripps Clinic in a recent podcast. Dr. Topol has also written extensively in support of the future of AI in medicine. You can hear Dr. Topol in a podcast from the Vox podcast “Explain it to Me,” entitled “Paging Dr. ChatBot.” Dr. Rosenblum fears that her clinical acumen, her ability to synthesize the complex issues, is a function of the exercise of writing her note. Dr. Topol agrees that there is a risk of losing some clinical skills. Worrying that innovations will erode clinical skills is not new in healthcare. I was better with my stethoscope than younger physicians who relied on Echos and other new diagnostic technologies. I was worse with my stethoscope than older physicians who didn’t have the ability of more advanced imaging and cath data to tell them what to look for and follow at the bedside. During my training years at the Brigham, I heard that Dr. Merrill Sosman, the longtime pioneer radiologist at the Brigham who passed away in 1959, predicted that radiology would render the stethoscope a historical relic. Clinical skills have changed before.
What is more concerning than a loss of bedside skills is the ability to integrate the patient’s personality, ability to communicate, social history, past medical history, current symptoms, clinical findings, and lab tests into a diagnostic formulation and therapeutic plan. AI may eventually be able to do this, but Dr. Rosenbaum is justified in being skeptical in the moment. Dr. Topol is a realist who is thinking about the inevitable changes.
Dr. Lin had much more to say. Just as there are AI tools that can sift through hundreds of unread emails and highlight what needs your attention, AI can also clear the clutter in an inbox and generate answers to patients’ questions about lab tests and requests to reorder medications, but the AI’s responses need to be reviewed. Just as there are now people who spend hours with a ChatBot talking about what concerns them, it is conceivable that your patient may get more emotional support from your Bot than from you when you finally get to their patient portal question in the wee hours of the evening after a long day at the office.
Dr. Rosenbaum’s concerns did not flummox Dr. Lin. Like the reality that my dentist had no education on how to use his magic wand in dental school, today’s doctors did not receive any instruction on how to practice with AI tools. Modifying medical education to shape it toward the current and future realities of AI is a work in progress. Errors will be made as learning occurs. Dr. Lin’s response to Dr. Rosenbaum’s concerns about the damage to critical thinking and patient relationships was essentially, “We will see.”
Steven Lin: I love that question, because it is certainly something that can be an unintended consequence… And now that this is a little bit more widespread and really has moved out of the theoretical realm into daily practice, folks who are interested in, say, continuing medical education, but also the training of our next generation of physicians have raised that as a legitimate concern: writing as a tool for clinical decision making, refining your differential. And so the short answer is we don’t really know how this is going to affect all of that… And so as AI gets more sophisticated, a lot of the analysis, a lot of the differential diagnosis will become increasingly dependent on AI — in one future. And what is the role of the human clinician then?
He answers his own question:
…And so it’s to wrap analysis in human judgment and compassion, tailored to the individual needs of the patient in front of you. That’s one answer. But then what does that mean for how do we select med students? How do we train med students and residents?
Brave new world! Nobody really knows. I am personally of the mindset that there’s a core corpus of skills around critical thinking, analysis that we should not give up to artificial intelligence — especially thinking about a lot of the limitations of artificial intelligence, hallucinations of large language models being one of them. …I think the role of the human clinician will change. And it becomes a little bit more, OK, here’s what the AI says could be going on. Here’s what my human instincts tell me. And then these are the things that are right for my patient and not just a generic human patient.
You may think that sounds dangerously naive. Studying what is happening is not the same as conducting a randomized clinical trial. If Epic, and I assume other EMRs, are rapidly integrating AI tools into their products, then these issues will be more akin to learning to live with global warming than the process of approving a medical device or new medication.
Dr. Lin reports that most, but not all, of the Stanford PCPs like their AI tools. One recurring concern that both Dr. Rosenbaum and Dr. Topol raised was whether the time saved using AI would be exploited by evil CEOs who seek to cram more patients onto a clinician’s schedule. Used with care, AI has been shown to reduce burnout, and it may also have the potential to mitigate moral injury. However, there are no guarantees, and we can expect that in some places, or perhaps in most places, it could be used to further reduce the joy of practice and further exacerbate all that currently undermines practice and patient care.
Dr. Lin does not see AI as ready now to perform chart reviews, but he expects it will be in the future. What is even more surprising is that he is unsure whether primary care will survive with or without AI.
Steven Lin: … I don’t think AI alone is enough to save primary care. That really is a very complex decision that we have to make as a physician community, but also policymakers will have to really understand where do they want to invest the money for the healthcare in this country and right now, it’s clearly not primary care. We need to change that.
After that startling comment, Dr. Lin wanted to return to discussing all the things that AI could do if primary care were to survive.
Steven Lin: Yeah. Well, you can envision in one scenario where you have AI chatbots that are doing fairly in-depth triage. So they’re able to have a conversation with the patients through text via the patient portal and ask some of those red-flag questions that would lead down one pathway versus another. So to differentiate between a straightforward, “Yeah. This sounds OK. We’ll just go ahead and send you to ophthalmology, for example,” versus, “No, no, this doesn’t sound like … There’s some red herrings going on. Something is fishy. We need some sort of human-level judgment, or we need to ask some more questions to tease out what this is.” So that technology today exists. It’s not currently built in any form that I’m aware of that is doing that. And I would say that’s a fairly risky application that needs to be thought of very, very carefully before we deploy something like that.
He calls that application using AI as the front door. He sees it as a servant, not the master, yet. He also sees that every institution will decide for itself when and how to use AI. He sees AI as best used for repetitive, time-consuming, boring tasks.
Basically having AI as the front door. I think right now it still feels like you need a lot of human supervision for that. Or you can just basically decide, “This is what I’m OK with AI doing, and everything else I’m going to leave to the human workforce.” And so at some point, each institution needs to decide where they draw the line in terms of their level of risk that they’re willing to accept. AI is really good at repetitive, low-risk, administrative-burden-reduction tasks, and I think it’s very reasonable to start there. As soon as you start to let it encroach upon not only clinical decision making, but then also let it be the front door to a patient experience, you have to really thoughtfully ask yourself, “Is this the right thing to do, and what are the consequences of us missing something?” And so every institution has to sort of make that decision.
I have skipped a lot of this fascinating conversation. I hope that you will listen. Toward the end of the conversation, Dr. Lin introduces an idea from one of his colleagues, who is also an internationally famous novelist, Dr. Abraham Verghese:
Already, there’s a tendency to think about the entire world existing within your one screen of Epic. And my colleague Abraham Verghese at Stanford, for example, has written a lot about the “iPatient,” how patients are now kind of digital avatars based on what the information is on the screen. And the entire world cannot be just what we can see on their Epic screen for the patients…the counterargument is that humans are not that great at practicing safe medicine at baseline, and that we make a lot of mistakes. To err is human.
We are back to where we started. Based on current research, what/who is most likely to make– the right diagnosis from a simulated patient case: human, human plus AI, or AI alone? And the answer is … AI alone.
There is more, much more, that I will leave for you to hear if you have an interest. You can be assured that I will periodically return to both AI and primary care in future letters to you. After much thought, I think the introduction of AI to medical practice is probably more complex than using a computer program coupled with a wand as an innovation in making false teeth, but I would guess that for many complex reasons there was less resistance to the dental innovation as well as much less impact on the future of dentistry, when dentists use a new technology than there will be for the potential of AI to transform how we practice medicine. What is also true is that AI may “steamroll” practice as we know it. Dr. Lin expresses the problem:
Steven Lin: I’m really worried that — and we talked about this, I alluded to this before…that the pace of this technology has completely outstripped our ability to study it and teach it. I worry about both parts, because I’m both a researcher as well as an educationalist. And we’ve already talked a lot about sort of the research aspect of things. But from the educational perspective, we are basically, at this point, introducing a whole suite of new AI tools to a workforce that is not ready for it. And then our learners are not necessarily getting education, any sort of structured education. Fewer than 10% of medical schools and residencies have any sort of AI in their curriculum whatsoever.
…Where I’m most excited is in the low-risk applications of AI that will really help unburden physicians from the administrative grind that is crippling them and their practices, leading to burnout, leading them to leave medicine altogether, and leading them to have moral injury because they’re not able to practice the type of medicine that they want to practice and that their patients deserve of them.
…I’m an optimist.
Dr. Lin may call himself an optimist. I prefer to say that I am “hopeful.” Optimism almost implies certainty. Hopeful works for me, because hope, more than optimism, considers that there is the possibility of disappointment. I am hopeful that AI won’t doom primary care. I hope that AI will be the new tool that brings us closer to our patients. The thoughtful application of AI by every system of care is not a “slam-dunk.” Time will tell.
We Are Moving Quickly Toward Winter
Today’s header is a picture of my favorite tree, the elderly and infirm Japanese maple in my front yard. I took the photo on Tuesday, which was probably the next-to-last day of what I like to call fall. It was both bright and chilly, but there were still bright colors to be seen. By Thursday, it was colder, darker, and windy, and the leaves were quickly turning brown and falling down.
I am readjusting my gaze from the end of fall and Halloween toward Thanksgiving and the December holidays, and the not-too-distant New Year. I hope that wherever you are this weekend, there will be a little fall left to savor even as you turn your clocks back.
It is unlikely, but each morning I open the New York Times hoping to read that the shutdown is over, the ACA supports are extended, the One Big Onerous Bill will not cut Medicaid, and everyone’s SNAP card is loaded for a trip to the grocery store.
Be well,
Gene
