In July last year, a surgical procedure, which involved a lengthy gallbladder removal, made headlines. What really surprised was not only the success of the operation, but an unprecedented detail: The procedure was performed without human intervention, with a robot. Although this advance was discovered by researchers at Johns Hopkins University in the United States, today there are implementations of artificial intelligence in the field of health in Argentina that have achieved improvements of more than 1000% in certain aspects.
A recent case in these latitudes took place at the Dr. Ramón Carrillo Central Hospital in San Luis, where several problems were detected some time ago when caring for patients. “We observed that Doctors in outpatient offices lost between 20% and 40% of their time on administrative taskshow to fill out medical records and prepare prescriptions,” explains Dr. Adolfo Sánchez de León, president of the hospital’s board of directors.
In addition, it indicates that, when conducting quality surveys, patients were satisfied, but complained that doctors spent more time looking at the screen instead of visually interacting with patients. “That removed empathy and undermined trust in the relationship,” adds Sánchez de León. Finally, a third frequent problem that they observed in consultations was that, although the doctors filled out the medical records, they were often incomplete and full of abbreviations that were difficult to understand.
In this scenario, Carrillo integrated artificial intelligence into medical practice in December 2025 and achieved very promising results. “There are other institutions that ask us about this implementation. Everyone is talking about AI in health; in fact, the two things that are coming – although they are already present – are AI and robotics,” explained Sánchez de León.
The project consisted of developing a product that transcribes information from consultations directly to the medical record. The application, called “Nodica”, captures the conversation between the doctor and the patient with a microphone, and transcribes it orderly in each field: medication, diagnosis, history, etc. It is worth clarifying that the conversation is not recorded, but rather is converted into text at the time; The system also has the ability to distinguish between information linked to the query and conversations from another context (such as the result of a soccer match) and records only the former.
The results? If before the doctor wrote down on average only 22 words per consultation, Nódica began to record 253 wordsthat is, an increase of more than 1000%. Another indicator that measures the quality of the annotations (words of importance, according to medical terms) improved by more than 263%. A parameter linked to the registration of medications, in which 0.4% was previously recorded, today grew to 10.9%.
In relation to the doctor’s vocabulary, there are several ways to name the same disease. With Nódica they managed to unify words that have the same meaning to 139, while before this measure was located at 20. For their part, abbreviations disappeared from the registry and are located at 0.01%. It is worth adding that Before, 13% of the fields were completed, while now this number has climbed to 81%.
In a second stage of AI adoption, the Carrillo hospital incorporated an application on the doctors’ and nurses’ cell phones connected to the health center’s medical records system. When conducting an interview, the phone orders it and then sends it to TASI, the digital system they use. Before sending it, it must be validated and signed by the health professional. In a next stage, they will take this system to the Intensive Care and Emergency services.
Where are we going? “Towards the possibility of a doctor taking a medical history and asking you questionsand ask, for example, how many episodes of fever the patient had, information that Nódica can provide and even distinguish by day, time, etc.,” says Sánchez de León and explains that this will allow epidemiological studies to be carried out, since the system can be asked to classify all medical records and distinguish how many patients there were with a certain disease, of a certain age, in a certain period, among other variables.
From the Laboratory AI for Good from Microsoft there is a team of artificial intelligence experts, machine learning and computer vision that They work with specialists on various topics to help them solve difficulties they face. “We realized that, although there are extremely different problems in society, such as helping to diagnose pancreatic cancer or being able to detect if there is deforestation in the Amazon, from a computational point of view they are practically the same challenge; in the sense that one works with image data and with algorithms to detect patterns,” explained Juan M. Lavista Ferrés, one of the authors of the book. Artificial intelligence for gooda work that compiles the work they have done until last year in the Laboratory.
Among the projects they are working on, there is an initiative that seeks to prevent retinopathy of prematurity, one of the leading preventable causes of blindness in children worldwide. It is a disease that did not exist a couple of decades ago, but which, according to Lavista Ferres, is now experiencing exponential growth in many countries: thanks to advances in medicine, more and more children born prematurely are surviving. “However, they are still not prepared to live on the planet, in the sense that their retinas are not fully developed. In fact, the singer Stevie Wonder, for example, is blind due to having suffered from this condition,” explained Lavista Ferres.
In some cases, When they are born prematurely, they have a window of between 24 and 36 hours to detect this disease, have an intervention and prevent blindness.. “The problem is that the world has only 200,000 ophthalmologists, of which only 10,000 are pediatricand there are millions of cases of children who are born prematurely, who will need to undergo an analysis to know if they have this condition or not,” says Lavista Ferres and explains that, due to the numbers, it is mathematically impossible to diagnose these cases.
With this panorama, they began to work with a doctor, who then contacted them with another doctor, Guillermo Monteoliva, from the Italian Hospital of Buenos Aires, to develop an algorithm to help in detection. This It is installed in a cell phone and, using the camera, it takes a photo of the retina and can detect retinopathy of prematurity in children in the same way that a doctor would do. They explain that they are still in the first uses, a lot of testing remains, but it is a tool that they are already working with in Argentina, Ghana, Colombia and Mexico.
The Carrillo Hospital in San Luis is not the only one that is implementing tools for transcribing consultations. The Austral University Hospital is also going through a digital transformation and is working on the adoption of new AI solutions, to enhance medical practice and optimize care processes. Among its implementations is an AI agent designed to assist the doctor during the general consultation, streamlining access to relevant patient information, generating clinical summaries and facilitating the preparation of the discharge summary.
At the same time, are starting to develop tools Ambient AIwhich transcribes the consultation directly into the medical record. “These capabilities not only improve operational efficiency (which we saw at booths at a global event organized by the Healthcare Information and Management Systems Society), frees up two hours of professional administrative time), but They also strengthen the quality of the clinical record and patient safety, key pillars in the HIMSS maturity models,” explains doctor Magdalena Princz, Health Informatics Manager at the Austral University Hospital.
The investigation was also strengthened by AI, as allowed structuring and exploiting data that was dispersed in your electronic medical records system. In this way, it made it possible to correlate variables and the information was transformed into actionable knowledge on a population scale. Doctor Manuel Mendizábal, head of Hepatology at the Austral University Hospital, details how his research benefited, thanks to the integration of this technology: “From this work, when analyzing 42,594 adult patients with fatty liver associated with metabolic alterations, We saw that it is not an isolated problem, but rather a disease closely linked to other common conditions in the population. In this cohort, six out of 10 patients had diabetes, a similar proportion had dyslipidemia, more than half had high blood pressure and 52% had obesity.”
In October 2025, the same hospital incorporated a tool, SwiftMR, based on a state-of-the-art deep learning model, which uses AI. This technology allows the sequences of virtually all MRI protocols to be accelerated. “The implementation resulted in an average decrease of 42% in net study timesallowing not only to increase the number of patients we serve, but also to improve their experience, by spending less time within the team,” explained doctor Guido Vázquez, biomedical technology manager at the Austral University Hospital, and added: “The results of the implementation were even better than we imagined during the pre-feasibility analysis stage of the project. For example, a knee MRI today can be done in 6 minutes of scanning”.
