At a time when healthcare in Europe is under constant pressure from staff shortages, a technological solution is emerging that should help make the system more efficient. The Akeso holding, led by entrepreneur Sotirios Zavalianis, has begun implementing artificial intelligence that does not aim to replace doctors, but rather to serve as a highly intelligent assistant. This change is not just about technology, but about an entirely new architecture for working with health data.
Symptom Checker: The end of endless form-filling at the doctor's office
The traditional process of visiting a doctor often begins with the patient sitting in the waiting room and then filling out complex forms in the consulting room. Akeso is turning this process upside down. Using a digital interface, patients complete medical history questionnaires before the actual visit.
According to František Vlček, Director of Development and Innovation at Akeso, this approach can reduce the time a doctor spends documenting basic information to approximately half. More detailed information about the system's operation can be found in professional media, which thoroughly analyze pilot testing conducted on a group of 330 patients.
What does this mean for the patient? Instead of the stress of filling out paperwork at the doctor's office, the patient can describe their symptoms comfortably at home, often via a mobile device. The result is a structured digital picture of the patient that the doctor can immediately work with.
Technical background: Why is a regular ChatGPT not enough?
Many people ask: "Why can't a doctor just use ChatGPT or Gemini to take a patient's notes?" The answer lies in safety and accuracy. General-purpose language models (LLMs), such as GPT-4 or Claude, are designed for a wide range of tasks but can suffer from so-called hallucinations – that is, generating factually incorrect information that sounds convincing.
Specialized medical systems like Symptom Checker use RAG (Retrieval-Augmented Generation) technology. This method allows the AI model to draw answers exclusively from verified sources, such as clinical practice guidelines and the facility's internal regulations. This minimizes the risk of error and ensures compliance with current medical practice. While general-purpose models tend to "guess" the next word, medical AI must "verify" facts.
Comparison: General AI vs. Specialized MedAI
| Feature | General LLM (GPT/Gemini) | MedAI (e.g., Akeso) |
|---|---|---|
| Data accuracy | Medium (risk of hallucinations) | High (RAG, clinical guidelines) |
| Compliance | General data protection | Strict GDPR and EU AI Act compliance |
| Intended use | Creativity, text, code | Diagnostic support, triage |
Regulation and safety in the EU context
Implementing AI in healthcare is not only a technological challenge, but also a legal one. Within the European Union, these systems must meet strict standards set out in the EU AI Act. Healthcare applications are classified as high-risk systems. This means they must undergo stringent audits, be transparent, and provide human-in-the-loop oversight.
