Explore the applications and challenges of AI in healthcare, and discuss how technology influences clinical care and operational efficiency.
In this episode of podhauz, Carlos Pedrotti, medical manager of the Telemedicine Center at Hospital Israelita Albert Einstein and chair of the board of Saúde Digital Brasil, offers a clear and practical view of the current moment and the challenges of using artificial intelligence in healthcare.
Pedrotti has 16 years of experience at Einstein, 13 of them dedicated exclusively to telemedicine. Alongside Fabio Ferraretto and Caio Vahanian, the three experts discuss, among many other topics, the main trends and insights from HIMSS 2025, the largest global conference on health technology and data, held in Las Vegas.
Caio, who attended the event, says that “there wasn’t a single sentence there that didn’t mention artificial intelligence.” Pedrotti, however, notes that “many promises are not yet reality, and the deliverables, in many cases, fall short of expectations.”
Pedrotti emphasizes that every new technological solution must, first, improve the patient experience; second, improve the healthcare professional’s experience; third, raise the quality of care; fourth, reduce costs; and, fifth, ensure equity in access and service delivery. This sequence of priorities must always guide technology adoption.
Practical uses of AI in medicine: 4 key applications
Beyond the hype, Pedrotti brings real examples of AI in medical practice. He stresses that it’s essential to understand AI in health as a broad set that goes far beyond large language models like ChatGPT.
In telemedicine, for example, data collection is entirely digital, which offers a unique opportunity to integrate information and improve processes. Among the practical applications mentioned are:
- Ensuring proper use of clinical protocols – monitoring adherence to guidelines, such as the appropriate use of antibiotics in sore throat cases.
- Demand forecasting and staff allocation – AI models help adjust the number of physicians and nurses according to patient flow.
- Data analysis to identify patterns – assessing seasonal trends of diseases and outbreaks, such as dengue, and quickly adjusting operations.
- Judicious adoption of LLMs – Pedrotti warns of the risk of technological “laziness,” reminding us that not everything requires LLMs. Often, simpler and cheaper methods do the job.
Efficiency or clinical care?
For Pedrotti, the healthcare market is still maturing in balancing operational efficiency and quality of care. AI offers paths to increase productivity and reduce costs, but it’s essential to remember that, in healthcare, technical validation and patient safety must always come first.
In other words, there is an enthusiasm for AI that can end up reversing priorities: using AI for its own sake, without ensuring that the technology actually improves the patient experience, the professional’s experience, and the quality of care.
AI in clinical decision-making: risks, limitations, and evolution
When AI begins to participate in clinical decisions, such as recommendations for conduct or predictions of case evolution, important risks arise. The limitations of the models and the need for rigorous validation are central, since each decision directly impacts a patient’s life.
Therefore, the specialist emphasizes the need for caution so as not to let AI dictate conduct merely because it is technologically advanced. Care should define the use of AI, not the other way around. And this evolution should be closely monitored, with continuous and transparent validation.
Data applications in telemedicine today
Telemedicine, by its very digital nature, offers fertile ground for data. And the use cases are increasingly concrete: from staffing to monitoring protocols, through predictive demand analysis and the use of algorithms to ensure quality and safety.
But Pedrotti reminds us: everything starts with data collection and organization. Without this solid foundation, any AI becomes just a promise. “There’s no point in having a super-advanced model if the data source is flawed,” he points out.
This episode of podhauz shows that the future of digital health runs through AI, but always in a balanced and responsible way.
Want to listen to the full conversation?
Find it on YouTube or Spotify, and follow so you don’t miss upcoming episodes!
