Blog

Patient-First: the Future of AI in Healthcare

How does AI solve healthcare's most pressing problems?

Giovanni Lanzani

Updated January 12, 2026

3 minutes

It’s been millennia since humankind realized that, if you follow the advice of medicine, you can live a better and longer life (that, and avoiding carbs, apparently). The craft has been perfected ever since, and each major technological advancement has contributed to furthering this craft.

AI is the latest advancement, and to see its impact, I flew ✈️ to Lisbon for the AWS GenAI for Healthcare Summit!

My takeaway is that AI will augment doctors, instead of replacing them, and remove inefficiencies in the backend. Carbs will have to be skipped manually, though.

The "Human-Out-of-the-Loop" Application

Approaching AI as a partner rather than a replacement was a major talking point.

This brings us to the first major trend in GenAI healthcare, the “human-out-of-the-loop" application.

Unlike direct diagnosis systems, these applications operate in the background, optimizing the healthcare infrastructure. They offer efficiency gains scaled across hospitals, without adding to the workload of doctors and nurses.

Some highlights:

Enhancing medical imaging: Algorithms adjust X-ray imaging to compensate for missing data, low resolution, or poor contrast. This enhances the data radiologists examine, directly impacting diagnostic accuracy.
Enhancing data storage: Large hospitals pay millions per year for storage solutions. During the day, they presented algorithms that discard up to 70% of the space required to store image data, while reconstructing it perfectly when needed. Direct impact 💰
Reducing invasive examinations: A non-trivial percentage of cancers is linked to radiation exposure from CT scans. Lower-radiation scans are less precise, making diagnosis tricky. TheraPanacea trained a variational autoencoder on 250,000 CT scans to generate high-quality images from lower-radiation/quality scans. The demo was impressive.

A Moving Application

The Human-in-the-Loop Lifeline

The summit's most moving example was a highly "human-in-the-loop": in parts of Africa, a shortage of radiologists sadly coincides with a high prevalence of tuberculosis.

The solution is elegant in its simplicity and powerful in its impact: a portable X-ray box with built-in AI models, transported from village to village via motorcycle.

The motorcyclist has no medical training and AI does the heavy lifting, analyzing the X-rays in real-time, flagging individuals with potential signs of tuberculosis. Those are then referred to a hospital for specialized care and confirmation.

AI is not there to replace doctors but to extend their reach into underserved communities, turning a motorbike into a mobile triage unit and saving countless lives using early detection.

Building Trust Through Specialization

Another recurring theme was trust. To adopt AI, clinicians must believe in its outputs. To do so requires moving beyond general-purpose large language models (LLMs) towards customized versions for medical subdomains.

Why is this specialization a non-negotiable?

Medical accuracy and clinical relevance: A model trained on generic data is rife with wrong medical advice and irrelevant information. A model trained on academic work and clinical data provides better answers.
Evidence-based practice: Medicine runs on evidence. Specialized AI must access and quote reputable sources, allowing professionals to verify its output.
Contextual understanding and didactic communication: Generic AI models might be too straightforward, meandering, or miss the nuances of a clinical scenario. Specialized models prioritize clinical information and a didactic, educational tone mirroring how health professionals communicate.

Accuracy, relevance, evidence, context, and communication: essential building blocks of trust. From there, adoption (and impact) will follow.

Agents Tomorrow, Pragmatism Today

The AWS GenAI summit was not about agentic systems, but targeted tools solving actual problems:

Either creating silent, scalable efficiencies within healthcare systems, or
Designing human-centric solutions, enhancing diagnostic.

The future of AI in healthcare is then about leveraging human-out-of-the-loop AI for backend gains while creating human-in-the-loop systems, empowering professionals to do more and reach more people.

To achieve this, specialized models are key to earning trust and making an impact to, ultimately, put the patient where they belong: at the center.

Tags: