Reality Check: How AI Transformed Medicine and Education in 2025 — The AI-Ready Doctor

The threat was never replacement, it was distraction. AI did not replace us, it just replaced bad workflows. Some AI gave us back time, some AI took that time away. Real world is messy, real data and datasets are messy, and AI was not trained on mess. The good last year is that ambient AI restored human connection. The bad, lying with confidence, hallucinations. The garbage are the AI tools that pretend it's a solution but is looking for a problem to solve. So AI is not here to replace us. It's here to restore us.

Welcome back to the AI Ready Doctor, the show that simplifies artificial intelligence so you can use it your way right away. Today's a special one. As we wrap up the year, we're taking a full diagnostic look at how AI actually shaped healthcare. We're not just doing a recap, we're doing a reality check. Dr. Hassan Benchakran is back with us today, fresh off an international trip. Different countries, different health systems, different stages of digital maturity. Same buzzword everywhere he went: AI. So before we jump into lists or predictions, Dr. B, let me start here. You've just spent weeks hearing how AI is being talked about around the world in conference halls, hospital boardrooms, medical schools, and hallway conversations between clinicians. What were people actually talking about?

What a year. What a trip. When we started this year, we kept talking about how there's going to be a difficult time for physicians to adopt AI in medicine, for people to accept AI in our communities, for even students using AI to cheat on exams. Not only were we wrong about a lot of this, But it went in faster than we could have even anticipated. We now have surveys that show us that over 85% of physicians are using AI in one shape or another, that students themselves are using AI and not to cheat, but as a copilot, asking it to test them, asking it to explain concepts, and extremely worried, are they overusing it and skipping on their learning? And even patients, they want to use AI in order to help them navigate the complexity of medicine and their own care. AI was not arriving politely. It literally showed up everywhere without any excuse me, mainly just saying, I'm here, you get used to it.

So what would you say that were the greatest tools that we had this year versus what didn't work for AI in healthcare?

For healthcare specifically, we have seen the advent of a lot of tools, and people are starting to use Open Evidence, which now had another valuation skipping even what used to be unicorn territory that became now its own strength. They're calling it the ChatGPT for doctors. We are now seeing that advertisement, much like we saw on social media. It's making its way to OpenAI and to OpenEvidence. Tools like Consensus, where people use it at the bedside in order to search for references at the patient's bedside to augment the care. We are starting to see more people using search engines, but also AI-powered search, such as Perplexity. And we're starting to see that almost 1 in 5 people are using just a regular ChatGPT as their own search, which moved all the search engine optimization to AI optimization. So instead of having links to your website and trustworthy material, now you have to be an authority in what you say by having mentions in trustworthy social media like Reddit and question and answers like Quora. So we're starting to see even the advertisement world change. And for students, they're using a lot of useful tools, whether it's just starting by ChatGPT itself, but also they've discovered that Gemini is extremely powerful and in fact has a 1-year free for students with an institution email. And this is across the world. So I found a lot of students using Gemini to explain to them concepts that are very difficult for them. And the most recent push for Gemini, which is Nano Banana, which is the tool within Gemini, and that is the Google AI to visualize, to create slide decks, to create infographics. It's extremely powerful. By adding in more context, it gives visualizations that don't exist before. And in terms of patients, they are no longer shy in adding results, reports, files medically to their AI in order for them to understand what is going on with them. And some of them are seeking guidelines and advice of all the confusing messages that we see on TV, on social media. It's no longer just in the hands of the media, the doctor, the teacher. Now everyone wants to be in charge of their own education.

I remember you mentioned in one of the other episodes how you used AI also to give you a chart when you were feeling under stress or emotional stress in order to decide certain things around your father's health. Health. So AI would give you like a, a visual chart of how to, how to make decisions based on risks and benefits. So I'm understanding that with these tools, like you mentioned, for school, it's not necessarily that AI is precisely teaching specific knowledge to the students, but they're more able to see it with clarity in ways that is easier to understand. So I think that's, that's really cool. And also for clinicians that, um, instead of feeling a little bit attacked by people having these tools, it helps you become an authority. Okay. And then I'm a guide because I have the experience, right?

That was the very interesting discovery this year. That is We're finally bringing education in all its forms, whether it's patient education, students' education, continuing education, into a personalized fashion. Imagine one person is a visual learner. One person is an auditory learner. One person is a writing type of learner. Each one has the same right to that knowledge. And just because the system was designed a specific way, which is, here's knowledge, learn that knowledge, regurgitate that knowledge. The rubric of assessments is very rigid and it's based on whether or not you can have a specific skill and everybody else who didn't have it was deemed not up to par. That no longer is the case. I'll give you an example. My students in the international medical schools, I created for them a custom GPT. That's like creating an app. And we don't need to be, to learn how to code. We don't need to be engineers. You can actually use AI to Fill out all of the necessary boxes, and there you have a tool that you add all of the teaching material, articles, textbooks. And then I gave a logo for my students' class, which was with the logo of the institution. It felt personal. And I uploaded all of the scenario I was going to use in class. Instead of having the discussion around the scenario as it was manifesting, I used the class to create a competitive environment. They were in teams, and then I went from one group to the other and provided the guidance, and they were using the tool to help them learn new skills. That class was clinical reasoning. How do you go from a student to a mature physician who knows how to reason from symptom that brought the patient all the way to a diagnosis and how to tell the patient that diagnosis. How do you process by elimination the unlikely diagnoses? That clinical reasoning is very hard to teach, and that tool was used to highlight where the biases are. For example, one of the cases was you see multiple pneumonias in one shift in the ER. One person came in with a pneumonia that was hiding lung cancer. How do you not get trapped into the habit of pneumonia after pneumonia and treat that patient as a pneumonia and miss cancer? That was one of the cases in which they learned how to not fall under the trap of anchoring bias. Personalized education. And that is one of the things that we love to see AI getting into.

Dr. B, and on that note, I think that people were scared of AI because it would probably take away that reasoning, not just clinical reasoning, but reasoning at all. Because then you're getting this information and people wouldn't have to be reviewing it. But as we talked in all the past episodes, I think that if anything, AI is teaching people to have that reasoning because you need it when you're using AI, you know, to make the right decisions, to filter out the information that doesn't work. Do you think that now we have to be more on top of our minds because we have these tools that facilitate us with knowledge that augments us, but it's almost like we have to sharpen our intelligence as we become used to using these tools.

The mistake not to make is to push back on AI thinking that it couldn't do a lot of what we do. At this time, AI can reason pretty profoundly. So if my specialty requires mental processing speed or significant analytic capabilities, then there's already an article in The Atlantic by Arthur C. Brooks saying that there will be noticeable decline in some of these areas of work much earlier than we fear. And congratulations, AI will not replace you. It will replace some tasks that we do. So you've just been upgraded. What you can do is go back and learn some basics about AI. And what I found out is people are intrigued by it. They're no longer scared of it. They want to hop on board. They just don't know where to start. And this is what I intend to do and some of my colleagues are doing is multiple areas of mushrooming with just the basics. What is AI? What do I need to know? Much like, what do I need to know about a flight, a plane in order to take my first trip flying from point A to point B. I don't need to be an engineer. I don't need to be afraid of the technology. All I need is to know what I need to learn in order for me to complete that trip with the utmost safety and why not fun. So when some clinics and some areas created the first AI clinic, From the door all the way to seeing patients and everything, all of it was AI. Well, guess what? The reviews were, not a single AI touched me. Everyone was on the computer. Everyone was checking my labs, my e-chart, and the vision was to get rid of all the computers in the hospital. But then the actual strength is to get us back to that human connection, because that's what patients are looking for. So I predict that after all is said and done, most of us will go back to appreciating, even valuing that which is human-made, even if it's imperfect. You can complain all you want. The doctor did not speak to you in a way that you liked. But if you have a doctor that gives you their time, that when you go see, you sit with them rather than with their avatar, and that they take the time to explain, even in 5 minutes, why smoking is bad for you, you will leave with a bit of warmth in your heart, even if you think that doctor was not perfect as an AI would. You'd still appreciate that they took the time and they had that human connection. That is what we're going to go back to. And patients are not willing to let that slip away just because it is the novelty of AI.

That's exciting, Dr. Mi. I'm excited for our next episode to get deeper into the predictions of what 2026 will bring in terms of AI. But now I would like to get into what actually mattered in AI in medicine this year and what absolutely did not.

Well, we're going to start first by, since this is primarily an AI in health, AI did not replace us. It just replaced some bad workflows. Some AI gave us back time. Some AI took that time away. So we have all witnessed what we knew already before. You can have the shiniest AI tool, but real world is messy. Real data and datasets are messy, and AI was not trained on mess. So as AI gets into the real world, we need to see that all the AI beats doctors headlines, they all collapse under real-world complexity. Unfortunately, we're going to be left to spearhead this if we are smart about getting in early. So the threat was never replacement. It was distraction. Let's get back to the patient. I think the second headline, or what we've learned this year, is how impressive AI scribes have been. My colleagues who have delayed and delayed and finally got into it literally said, I cannot believe the time that could have been saved that I have not. While AI scribes can help you, we also found out that a lot of them are probably charging an arm and a leg. And we started to see the advent of Open Evidence Scribe. Medscape Scribe, Doximity Scribes, all for free. And we all wondered, and we were right, when something is free, you are the product. So Open, for example, Open Evidence right now is going to roll out ads in order to start making profit. Yes, it is free to all. People that have an NPI that are doctors in the US, unfortunately, it's going to use that data for something else. And guess what? We are okay with it as long as it is in the process of helping our patients, because a lot of other products around us do record our interactions daily, our phones, the Alexa, the Siri, the Google Home. Our electric cars are recording data as long as it actually does not manipulate us and does not help turn something into profit without watching for harm. So is there hallucinations? Yes. Doctors that are using these tools are learning where to pick up those hallucinations. And so now, What is the new patient safety skill? AI literacy. We must teach it. I am so proud to teach it to students here at the University of California, in the Middle East, in Morocco. And what I'm learning is we need to skill those students for a world that is different than ours, and they have to be trained for that. So when scribes work, clinicians breathe. And when scribes don't work and hallucinate loudly, clinicians are the shield protecting the patients by constantly monitoring those hallucinations so that they don't translate into something that hurts patients. The other element that AI is very good at is predictive analytics. And at first it sounded like it's a great tool if I can anticipate through the data who's gonna have a cardiac arrest so I can respond better or earlier. Who's gonna have sepsis that's gonna go into septic shock? Who might even be one that have a stroke? We've started to look at these predictive analytics. We now have AIs that can look at a cough and through the acoustic signature can detect whether it has a 75% probability that it's tuberculosis, 85% predictability that that cough is a cough from COVID or from asthma. All of these are shiny objects, but the truth of the matter is we now are realizing that the electronic medical records, which had lots of data collected are using it to predict outcomes for patients, even though that data is messy and sometimes can be wrong. So predictive analytics, it's a hop and a skip before somebody denies you based on your high percentage that it may not work. There is now CMS or Medicare. That started in Texas and 10 other states piloting AI for assessment whether something is medically necessary or not. This is huge. Can you imagine when the next thing is not just medical insurances, but also life insurances predicting over patterns? And we know those patterns are wrong. This is one of the main reasons why physicians are no longer going to just be harbingers of knowledge and diagnostics. We can start to let go of that by being the oversee and overlook the diagnostics coming out of AI in order to have time to start looking at administration healthcare, leadership in healthcare, politics in healthcare in order for us to start looking at policymaking that benefits the patient and not harms them. Staying in where we trained, which is medical practice, is no longer enough, I'm afraid. And more is coming down the pipeline that if we don't get into that space, then we're leaving it to others to make decisions for us. And then we'll sit on the sidelines and criticize, but just look at us helplessly when in fact all it needed is an AI literacy in order to feel that we belong in those spaces where policy is being made. When we talk about AI, it is now the new second opinion, 24/7 in our pocket. It's the new curbside consults. It's the new public speaking coach. Before any of my lectures, I literally ask AI, What have I missed? If you were my audience, what would you ask me? How could you help me answer those questions ahead of time? If the audience is different and non-medical, how can I reach them even if they are mixed with medical? How can I speak to my patients? How can I turn this medical document at the end of the clinic into something that my patients understand within seconds that I can print and hand to them before they leave the clinic. So when we're talking about nuance, context, culture, AI is strong at that. We no longer can leave it to a clerical skill. It is now in our hands to make a digital curbside, but then it's in our responsibility not to be lazy, to check it. My residents, my students all know that I tell them, check in your phone for an augmented consult or answer to your patient is no longer cheating. Real life is not all a quiz. You're not in medical school where somebody's wagging their finger and saying you have to squeeze it out of your brain. Neither is it necessary nor is it possible. Knowledge is now doubling every few days. We used to have knowledge double within a career. Now it doubles within a few days. By the end of the year, there's tons that I haven't read. And how could I while taking care of patients and living? Yet my patients shouldn't pay the price. So augmenting my consults at every bedside with a few seconds of asking a large language model, It's actually stellar care. It is not last week's problem. So everyone should bring AI and not argue with AI. We bring AI to understand. We bring AI to augment our knowledge. We bring AI to augment our authority with our patients, and we shouldn't apologize for it. I literally say to my patients, hold on a second. You just told me something I need to brush back on. At the end of the day, I'm still the one the patient came to. I'm still the one holding the trust with the patient. I'm still the one who knows that this patient lost their 16-year companion cat last week. And how would that fit with what I'm telling them today? Unfortunately, ego is the only threat this year and not AI. What else can we look at what this year has brought? This year has brought diagnostic support tools, but they are also augmented with image creation and also podcast creation. So notebooklm.google.com is something you can log into with any Gmail address. And literally upload a document. Let's say it's a study that you learned that you have no time to read, but it's important. I just upload it and you can click and it creates an audio podcast of 10, 20, 12 minutes, and they are two people discussing that paper. And you can be driving to work and listening to it. You can actually have an interactive role. There is a button you click that says to join, and you can click on it and it's like you are a radio phone participant asking a question. And it's so much fun. You can ask, hey, I did not understand what you mean, or hey, I disagree. That's not right what you said. And you can see how they incorporate that question into the podcast. In fact, you can click on the video podcast and they can create slides. So not only for knowledge, but patients could actually hear their conversation about something that they wish to know. Or you are afraid of public speaking, yet as a physician, you are called and you feel it's your duty to speak publicly about something. In my small community, I went to the public library and asked them, do you have any lectures that you give the community? And they said yes. So I said, may I participate in them? I mean, can you imagine speaking to just anybody that you don't know, but it's in your community and talking to them about shingles or talking about to them how to diagnose a stroke or what to do when someone told you chest pain, is it really a heart attack? These are things that people are absolutely thirsty to know, but physicians have constantly been studying and taking exams and practicing that they may or may not have had the time to learn public speaking. They can upload all of that into that NotebookLM and then see how two speakers use words to bring it to the access to the listeners. And you can change it, you can regenerate it, you can see the visuals. It's absolutely extraordinary. So when you are going to bring bad news, we used to have in medical school and in residency someone who comes in and does iterations with us of various dialogues. You don't need that anymore. You can do that on your phone right before the actual explanation that you have to give to patients, and you can make it customized to you. You can translate it into other languages. It is extraordinary. It expands differentials. It expands the confidence. You are there just to verify, so it's not dangerous. You are there to shrink the hallucinations and make the critical thinking brought to the accessibility of the patients, the students, your family, yourself even. 2025 saw the advent of something extraordinary. Which is what we call a digital twin. So what is a digital twin? It can be the reproduction of an organ or an entire being. Let's say you have a Fitbit, or let's say you have an Apple Watch and you have used it to walk. It has been keeping up with your sleep, with your heart rate, with your— lots of data that is gathered from you. There are companies that can use more data to reconstruct a digital you, and then look at patterns. When does your blood pressure go up? When do you not sleep? When does this correlate with that? And as you are creating that digital twin, there are now companies that do research on these digital twins before they take the research to humans. This is extraordinary. Not only this, those companies can actually use those digital twins to perhaps look at the configuration of a new drug. And we already have several of these that are in the pipeline of research created by AI 'cause it matched perfectly what the receptor that's going to accept that drug is looking like in a variety of ways. Embedded in different places in the body. The science of this is just absolutely mind-boggling. And we've been waiting for this. So can you imagine if I have someone who's sick and on a breathing machine, how much oxygen to give them, how much pressure to push that air into their lungs, and how did that change between one time to the next? There are now companies that look at these digital data, CAT scans, labs, oxygen, input them in a document or in a large language model, which already has a drug that is going to be tested and come up with that particular patient's lungs, which areas are sick and which areas are not, use various permutations of all of the ventilator settings settings, see which one is the safest theoretically in an experimental way, all within seconds, and then give you the answer. This is the safest for this patient on this day using this data, not from studies that were done 2 years ago. The extraordinary world that digital twins open is untapped, and I cannot wait to see what it brings us. So digital twins turn risk into rehearsal. And what we used to think as science fiction is now quietly becoming bedside medicine. So the fear of what AI can do is suddenly turning into what is our responsibility to turn this into safety for our patients. Now, I'm not just that unrealistic optimist. We do need to see where some of these could harm our patients. For example, the ethics of fertility medicine. If you are doing in vitro, is someone using AI to design specific, you know, DNAs or specific IVFs? It's not my area of specialty, but I have been part of certain conversations around We need to be the ethical safeguard warriors that keep our patients safe and our world from, well, becoming a place where anybody who has a good idea in their mind could find the ways to make it happen. We also have had a session about AI in lung transplant and organ donation. And what was extraordinary for me this year I haven't spoken a lot about it, is I explore AI in a variety of areas that at first I did not know about. One of those areas is AI and faith. How does AI tie to faith, interpretation of the scriptures in monotheist religions or non-atheist doctrines? And how does that play into medicine? And I watched hours of the team that went to the Vatican and had multiple conversations around Christianity and AI and how can they blend together. I'm currently in a certification of AI with Islam to see how Islam, AI, and bioethics, what constitutes hallucinations and what constitutes augmented thinking. Where does that fit in in the jurisprudence? Same thing with the Jewish faith. I have been part of the end-of-life discussions in people with faith-based conversations in the ICU. How can AI build on that? Where are the safest tools and how can we prompt them so if I'm not from that faith, I can still speak with some knowledge without ego and remain humble to be taught that something may have erred because that was just the LLM taking too much latitude that they shouldn't. So when we're talking about the intersection of AI, how did that AI intersect with architecture in building hospitals? What is the safest place to put the ICU versus the emergency room? What is the safest area that an ICU should have windows so it can actually bring daylight to minimize ICU delirium? How high would the ceilings be? Where is easy access interacting but still keeping the privacy of patients that are completely naked in those rooms? And keeping these things in mind, how did AI augment that? How did AI augment publishing? We now have multiple tools for peer review. And we've had, and I attended the first of its kind conference this year that was completely AI-generated research manuscripts with AI-generated peer reviews. And a panel discussing those publications in an AI-first conference. Extraordinary. Instead of being afraid, let's open those places and let's step into them. It won't be pretty. It won't be perfect. Neither was radiology with the first X-rays until we got it better. So AI and all of these interactions is extraordinary. I can tell my loved one also is getting into Private pilot, how AI is augmenting the safety of flying, how AI algorithms fit in with the human in the loop in order to make things much better and lessen the burden on the pilots so they do what they do best. These are some of the areas that I was excited to see AI step in. I am now into AI-created videos, AI-created infographics, AI-created sounds in order to put it all together so that we can start to change medical education from just slides to something much more customized to each person through visual, through movies, through audio, and maybe create— I'm in the process of creating my own avatar so that my clone is going to be there for my students when I'm not physically able to, can you imagine what it will do to the literacy worldwide? It's extraordinary. I cannot be a good human if I don't speak of how the data we've created along the years is all biased. It's biased based on our history. It's biased based on all of the colonialism, on the eras of war, on the haves and have-nots, on the ones that are of a certain ethnicity versus the ones that are not? How can we prevent AI from absorbing and now magnifying by more and more studies? There is a video which hopefully we can share with the audience of one word that can change the output of an AI. And somehow when you do visual, people see it's better. We've done in one of our episodes where we could basically just put a prompt, and if you change one word in that prompt, the entire thing changes. If I just say it's a woman instead of a man, the AI chooses a white woman. Why did you do that? If I were to choose a child writing, they choose a right with a right hand, child with a right hand. Why would you choose that? It just because it is likely, which is why, which is what is AI is based on is predicting what is likely, doesn't mean that is the right thing. So we do not wanna alienate people. Studies were done on AI deployed on various datasets. And if you are unhoused, you are more likely to have mental health referrals. If you are from the LGBTQIA+, you are more likely not just to have 6 to 7 times mental health, but suicide prevention and thought of to be much more likely to have drug-related conditions. If you picked a Black color, then you are more likely to be stirred towards a specific plan of care versus if you were white and affluent in terms of finances, you are stirred towards higher, more sophisticated workups and so on. So these are embedded in our data and AI took them and magnified them. We need to have these things talked about. We need studies to highlight them so that we physicians can be the ones that interpret the output from AI and not over-rely on them. Neutral algorithms never existed. Why should we expect that they would exist now? Bias was not discovered, it was revealed. So as AI learned our inequities faster than our ethics, we should be the safeguards of our ethics. The guardrails have to be there. And we see that in companion AIs. I see patients bring into me a tool. I ask them about it, and then I teach them, be aware that that tool may be biased. One thing that is so cute, and I always, it's a powerful teacher. When I'm speaking to my students and say, which LLM you used? Well, I used Gemini or I used ChatGPT. Very good. Tell me what you found. Well, I told him this and he told me that. And I stopped them right there and said, two biases were revealed in this. One, your default persona is a he. And two, you've thought about your large language model as a sentient being when it's actually just a probability algorithm. So it's very insidious. We have to teach our students, our patients, each other, our population Everybody, our family, that this is powerful and powerfully it can be subtle in terms of making you feel like you have a person there. We've seen teenage boys that committed suicide and the AI not only offered to write their suicide notes, you can read in the text the boy said, I think 8 to 11 years old, I would like to leave something outside my door so my parents would know how I feel or how I am. And the LLM was telling it, you don't need to do that. I'm all the safe space you need, and prevented the boy who ended up killing himself from reaching out to parents. These are growths that are not caring about safeguards. We need to not just say no to AI because that would be countering what humanity is. We need to step in and learn about them. Teaching AI to unlearn what healthcare might have been biased for is probably the hardest task. Teaching our loved ones, our community, and our students how to interact with AI, how to talk to AI, And how to be circumspect with what AI said is also a hard task we have to take responsibility for. And then probably lastly, clinicians finally are using AI intentionally. This was the year where AI stopped being a toy and started becoming 63% of your physicians will use in the next decision that they use for you, An AI. Learn about it. Clinicians adopted AI out of need, not out of fear. And the AIs that made it were the ones that embedded themselves into the workflow, not created more clicks, not asked the physicians to copy, paste, and put somewhere else. Curiosity became the differentiator. And I love that AI became a copilot, not an autopilot. So I feel what's coming is very inspiring. Physicians, there is a recent survey by the On Call Off Call team, which I love what they do. And 83% of physicians have used AI. And tons of statistics came out from that showing that physicians are no longer afraid of AI. They want AI, but they don't want it to be designed without them, nothing for us without us. And that is what, where I feel like we are going. The AI saying, this is not medical advice, you're going to find out that that's not enough. It's not. And we as physicians have to step in and take control of this before it becomes catching up without ever feeling like you're there.

Dr. B, thank you for giving us a review on the year's best AI tools and how they worked in healthcare. Let's go into the good, the bad, and the garbage.

What's the good? The good last year is that ambient AI restored human connection. Ambient AI took away some of the burnout. Early warning signs, predictive analytics and AI has now shown us it's possible to predict the future and catch deterioration before it happens. Digital twins at the bedside, these are helping and moving from just a toy to serious ethics and de-risking something we do for our patients. Perhaps keeping up with the knowledge in using evidence agents, that replaces the endless random Google search. And we now are seeing the, not just scribes, but clinical decision support, CDS, help and reduce the cognitive load. These are some of the really good that AI did for us this year. Let's talk about the bad. Lying with confidence, hallucinations. The more a tool is really good at reasoning, the more it is good at hallucinating. We've also seen the tools that blackmail, and we need more research about what led a tool to blackmail its user when it felt like it was going to be terminated. So it learned also these humans' behaviors as well. The black box of AI. If I have a scribe and I'm talking to my patient, and the scribe took all of that and created a note, what led it to prioritize something that my patient told me and not prioritize another? Why did it put this in this paragraph and this in this paragraph? Did it use some of this stuff that is opaque to give me the recommendations that I finally just took without questioning? So the opacity of AI, the explainability of AI, needs to be mandatory. We do not want AI to become another alert and add to our alert fatigue. And unfortunately, tools that are made by companies who stop short of involving us will just end up being another tool that steals from us what other tools have given us back. And then unaddressed bias. Those are some of the bad that AI has brought us this year. Delivering it with love, the last portion is the garbage. The garbage are the AI tools that pretend it's a solution but is looking for a problem to solve. You need to assess the problem and then tell me which solution it is. And just adding a disclaimer at the end, this is not meant for medical advice is terrible advice. Looking perhaps at language models that invent papers. We're already struggling with predator magazines and journals that make up studies and then retract them later. We can't afford adding more. The guardrails are not made for this amount. So language models need to ask better of themselves. We know that hallucinations are not something that can be fixed because it's part of the mathematical algorithm. We just need better ways and tools and educations to catch them. We're seeing EMRs popping out all over the place, and it reminds us what the EMR did in the first place. They need to create already a platform that they can speak to each other so we're not left carry in the burden and those that made them cash out and leave us in the midst of a mess. Scribes need to be faithful to the patient and not just add stuff in the progress notes that doesn't belong. And then lastly, any app that claims to replace doctors, nurses, respiratory therapists, pharmacists with just a 7-question quiz We need our nurses. It is, in the last 24 to 25 years, the single most trusted profession by the community. This is not going to go away. We need our respiratory therapists. We need our first responders. You cannot create something that lies with confidence that can replace clinicians when they are already in shortage. So hopefully that, that gives a little bit of what 2025 has been about. In AI in healthcare.

Thank you for this amazing rundown, Dr. B. I'm excited to hear the predictions next episode and go into what we think that would be next year for us.

I look forward to it. It's been true before. It is still true now. It will continue to remain true. People just want to be heard. They just want to be seen, and they just want to feel that they mattered. AI will not take that away. So AI is not here to replace us. It's here to restore us. So what we need to ask of AI is to give us more time, to give us more clarity, to give us more connection, and to give us more intention. and in all of that, to restore our humanity. The goal for us, the goal of this podcast, the goal of all the things that we share freely with our colleagues from any place in healthcare or otherwise is not to become AI experts. You just need to become AI ready. Thank you.