AI in Healthcare: Prescription for Digital Transformation

In today’s episode of the “Business Tech Talks powered by BlueSoft” podcast, Robert Wójcik, Founder & CEO of Healthcare Innovation Embassy, and Artur Mirończuk, Project Manager handling AI projects at BlueSoft, sit down with tech journalist Paweł Pilarczyk to discuss the strategic and safe deployment of artificial intelligence (AI) in medicine. The conversation focuses on categorizing solutions by risk level, building trust in algorithms through the “Human in the Loop” mechanism, and the practical steps of implementing technology in healthcare organizations and pharmaceutical companies. Below is a summary of the episode transcript.

AI Safety and Confidential Data 

Using AI in critical areas is safe, provided it is treated as support — a “second pair of eyes” or a double-check tool — rather than an autonomous decision-maker. A key trust-building element is the algorithm’s ability to cite its sources, allowing specialists to verify the information.

AI Applications in Healthcare

Artificial intelligence in medicine can be divided into three domains:

• High risk (Deep Tech): Solutions in medicine and life sciences that support diagnostics, where errors can cost lives.
• Medium risk: Patient routing management, known as triage.
• Low risk: Administration, scheduling, and paperwork. AI is also used to streamline internal processes (e.g., in pharmaceutical companies), saving employees time and allowing them to focus on scientific work or patient care.

Certified vs. General-Purpose Tools (e.g., ChatGPT)

The key difference lies in data validation. Professional medical tools are based on verified textbooks and scientific publications, while general-purpose tools often hallucinate. Certification distinguishes administrative algorithms from those that support diagnosis.

Real-World Deployments vs. Hype

Beyond media reports about disease detection, there are also very concrete, tangible business implementations. One example is knowledge base automation, which saves several minutes per interaction, adding up to entire full-time equivalents at the organizational scale. Another is algorithmic physician scheduling, which reduces conflicts and cuts the process from weeks down to a few clicks.

Trust in Results, Non-Determinism, and Hallucinations

Trust is built through the Human in the Loop principle — specialist verification is always required. Technically, the hallucination problem is mitigated through source citation (RAG) and multi-agent systems that cross-check several generated responses to verify their accuracy.

Consequences of Incorrect Decisions

In life-saving areas, the risk is enormous, which is why it is recommended to start deployments with low-risk use cases, such as administrative tasks or patient satisfaction analysis. This allows organizations to learn the technology without putting patient health at risk.

The Right Process for Implementing AI in an Organization

Implementation should follow four stages:

1. Inspirational meeting: Discussing the capabilities and limitations of the technology.
2. Workshops: Selecting specific areas for solution deployment.
3. Prototyping (Proof of Concept): Quickly testing the idea at low cost.
4. Scale or pivot decision: Based on the results of the POC.

Deployment Strategy: Small Steps or Multiple Tasks?

There is no single right answer, but workshops are the critical element that delivers the greatest impact. The most important foundation is well-organized data — without it (following the garbage in, garbage out principle), the system will not function properly.

Biggest Risks and Barriers

The biggest barrier is not the technology itself, but the organization’s mental readiness and employee adoption. Key challenges include the lack of space for experimentation (due to day-to-day operational pressure) and insufficient internal competencies.

Advice for Companies Hesitant About Implementation

• Test and experiment: Use safe, low-cost environments to build prototypes (POCs).
• Strategic partnerships: Don’t reinvent the wheel — leverage the expertise of experienced partners.
• Expert consultation: Ensure regulatory compliance and data protection.