Finally, Healthcare Is Becoming a Learning System with AI as its Catalyst

1. Making the Invisible Visible

Last century, the first major leap in our medical technology revolution came with the rise of modern medical technologies: X-rays, CT scans, MRI, robotic surgery, and advanced imaging systems. These tools allowed clinicians to see inside the human body with unprecedented clarity. Conditions once hidden beneath the surface became diagnosable and treatable with extraordinary precision.

Parallel to these breakthroughs, digital health records and early health informatics systems began shifting clinical information out of paper files and into searchable, portable formats. Although still rudimentary, these systems laid the foundation of our modern system.

2. When Data Meets Access

The second major transformation began when computing power, cloud infrastructure, and connectivity started reshaping healthcare operations. Electronic health records became ubiquitous, not necessarily as knowledge systems, but primarily as billing, documentation, and compliance infrastructures that structured how data moved through organizations.

Telemedicine, once niche, was thrust into mainstream adoption during the COVID-19 pandemic. Overnight, we redefined what it meant to “be present” with a patient. Virtualized care expanded access, enabled continuous follow-up, and pushed healthcare beyond the four walls of hospitals and clinics.

This era started conversations about interoperability, patient engagement, and the democratization of information. We can also call it the digital and telemedicine era, with the ability to store, share, and act on data with increasing speed and scale.

3. Intelligence Everywhere

Today, we are entering the third—and, I hope, the most dynamic—transformation: the age of AI. Unlike previous waves focused on tools and connectivity, this one is defined by agency, continuous learning, and exponential improvement.

Three characteristics make this era uniquely transformative:

Unlimited Learning Potential

Modern large language models (LLMs) are not static. They improve continuously as they encounter more data and more use cases. In healthcare, this isn’t about achieving artificial general intelligence; it’s about systems getting measurably better at highly specific tasks such as documentation, coding, eligibility, predictive modeling, and quality reporting. Improvements are not incremental; they are exponential.

Novel Data + Feedback

Generative AI doesn’t merely analyze existing data; it creates new structured information, new interpretations, and new insights. By converting unstructured content into actionable knowledge, AI accelerates everything from clinical documentation to risk stratification and operational optimization.

A New Innovation Lifecycle (My Favorite One)

In previous technological cycles, systems could plateau or fail. In the AI paradigm, failures become fuel. Every mistake becomes a learning event that strengthens the next model iteration. This continuous feedback loop fundamentally changes how innovation scales and sustains itself.

We’re already seeing the impact: AI-driven workflows are reducing administrative burdens, augmenting diagnostics, and transforming patient engagement. Tools such as automated medical scribes are giving clinicians back valuable time, one of the strongest levers against burnout and a direct contributor to quality of care.

An AI Bubble? I am not sure…

Each wave of healthcare innovation opened a new frontier. But the AI revolution is different. It isn’t simply about working faster or remotely—it is about redesigning the operating system of healthcare itself.

Agentic AI systems can autonomously execute complex workflows. Predictive intelligence can anticipate needs before symptoms arise. Operational systems no longer just record what happened, they increasingly shape what will happen.

This is also why I’m not convinced an “AI bubble” is imminent.

• First, I don’t believe there is a cap on how much better LLMs can become, not because we are chasing AGI, but because real-world performance continues to compound across tasks.

• Second, generative AI introduces a fundamentally different innovation lifecycle: a self-reinforcing loop of intelligence → novel data → improved intelligence that does not behave like traditional technology diffusion.

• In this world, failures don’t destroy progress, they accelerate it.

Healthcare is not adopting AI as a novelty. It is embedding AI into the core of clinical, operational, and financial decision-making. That creates durable value, not hype-driven volatility. And that gives me a lot of hope.

Originally published on LinkedIn, 19 December 2025.