Healthcare AI provokes both hope and doubt. On the one hand, it promises to enhance medical processes, including diagnostics and treatment planning. On the other hand, it has access to sensitive data, and its reasoning may not always be transparent. This creates a dilemma that was discussed at the “Healthcare x AI” event in Paris on March 9, featuring speakers from Theremia and Sanofi, companies pioneering AI in this industry. The tone was optimistic: AI is already enabling innovative use cases for medical institutions, while solutions are being developed with explainability, data security, and regulatory compliance at their core.

Are you wondering where the industry is headed, what role AI plays in biotechnology, and how to navigate challenges related to explainability, data management, and Europe’s regulatory framework? Take a look below – we’ve compiled the most thought-provoking takeaways from the event for you.

Shift from Functionality to Targeted Impact in Healthcare AI

One of the topics addressed during the event is the smart and gradual way of implementing AI and machine learning (ML) in healthcare systems. While some see the implementation of AI in healthcare as a complete overhaul of existing systems, the time and budget required for such a transformation can exceed available resources. Luckily, healthcare AI companies are shifting away from the ambitious, yet overwhelming idea to reinvent their entire systems from the ground up and are instead opting for impact-focused usage. This means that AI use often focuses on particular pain points like, for example:

  1. Improving patient outcomes with early disease detection and AI-driven treatment recommendations;
  2. Personalizing diagnostics by enabling precision medicine and computer vision-driven imaging analysis;
  3. Streamlining physician workflows with automated administrative tasks and efficient triage and routing.

In practice, this highly targeted approach is beneficial in many ways: from reducing upfront costs and meeting compliance requirements step-by-step to gradual updates of long-established infrastructures and uninterrupted day-to-day operations. This was proved by the AI-powered iOS application we developed for a B2C aesthetic wellness company.

Our client asked us to start a small project, specifically addressing aging and skincare. The core feature was AI-driven face analysis that we enabled thanks to computer vision, image recognition, and ML. The app scanned the image of a face, analyzed it to detect signs of fatigue, stress, or imperfections in real time. Based on the analysis results, the app offered 90% accurate facial mapping and, consequently, relevant skincare recommendation.

This purpose-driven strategy allowed our client to offer unparalleled functionality and attract a solid user base to guarantee faster ROI. What’s more, this AI-driven face analysis not only generated insights for users, but also helped skincare professionals to streamline the process of clinical planning. Now, the client can gradually enable other features with AI, while maintaining user engagement.

AI-Powered Convergence of Biotech and Technology

The “Healthcare x AI” event also brought up a possible convergence of biotechnology and AI to tackle previously intractable challenges such as understanding biological mechanisms (e.g., brain functions in neurological diseases) that remain poorly understood. 

The use of biotech with AI in healthcare examples include:

  1. Drug discovery via AI-driven molecular modeling and virtual screening to evaluate how potential drugs bind to target molecules;
  2. Drug trials via digital twins that allow researchers to run virtual experiments, predict side effects, and optimize dosages before human testing begins;
  3. Cell engineering through predictive models that can refine CRISPR guide RNA design;
  4. Neuroscience, specifically neuroimaging, thanks to deep learning models that detect subtle brain pattern changes in neurological diseases like Alzheimer’s, Parkinson’s, or epilepsy.

These are just a few of many examples, yet they illustrate how many diverse tasks AI can take on. This is why this technology has the potential to connect tech firms, biotech startups, and medical researchers to create a rich ecosystem where data-driven methodologies intersect with biological expertise. In this manner, healthcare solutions can be designed, tested, and enhanced much faster. This AI and biotech synergy can ultimately lead to leaps in tackling unmet needs like slowing aging, curing chronic diseases, or preventing epidemics.  

For example, AI proved to be effective in the development and delivery of COVID-19 vaccines. Moderna, a US-based biotech company, was one of the first companies that allowed fighting against the virus. The company utilized ML algorithms that helped the team of researchers to leverage reverse vaccinology and identify potential antigens from genetic sequences. At the same time, deep learning was used as a more advanced subset to handle unstructured data and design novel proteins that could serve as vaccine candidates. As a result, the first trials proved the vaccine to be nearly 95% effective, and, ultimately, 155 million doses of the Moderna COVID-19 vaccine were administered in the EU/EEA.

Building Trust Through Explainability and Responsible AI

Despite the existing practical results that AI brings to the industry, the use of AI in healthcare is still often met with skepticism from patients, doctors, and pharmaceutical companies. The reason for that is the AI’s black-box nature: models produce accurate predictions or classifications but they do not offer a clear understanding of how or why they reached those decisions. 

When it comes to high-stakes decisions typical for the healthcare industry, the reasoning behind diagnoses and treatments directly influenced the wellbeing of patients. When doctors cannot see why AI provided a particular diagnosis and treatment for a patient, they cannot rule out the possibility that AI was trained on unrepresentative data and has bias in its decisions. This uncertainty can mean inadequate and even dangerous medical advice. For this reason, only 15% of patients felt comfortable with AI making diagnostic decisions.

So, instead of completely delegating tasks to AI health tech, doctors and researchers must treat AI as a supportive tool for extracting insights. In such a way, they can come to certain conclusions faster thanks to AI and then validate and act upon these findings. This is called a human-in-the-loop (HITL) approach, where AI results are combined with human oversight. The HITL process looks like this: AI can analyze patient histories, genetic profiles, and imaging, which are hard for a doctor to process quickly and with unparalleled accuracy. Afterward, doctors can verify the results of AI reasoning, cross-check recommendations thanks to their experience and specialized knowledge, and consider patient-specific nuances that a purely algorithmic process might overlook. 

This HITL approach should also be combined with data governance frameworks (e.g., GDPR, EU AI Act). It is a crucial step in ensuring that AI decisions are verified by knowledgeable specialists and, thus, offer practical insight as well as comply with laws and regulations like HIPAA.

To further ensure fairness of AI-driven decisions, medical institutions can use explainable AI (XAI), which aims to open the “black box” to provide insights into the internal reasoning or decision-making path of a model.

Data Challenges and Strategic Partnerships

The future of AI in healthcare largely depends on whether AI systems can use data fully and ethically. The former means harnessing not just structured electronic health records (EHR) but also genomic data, lifestyle metrics, imaging, and patient-reported outcomes. The latter refers to using data strictly adhering to privacy standards, transparency in data use, and meaningful patient consent.

These complexities can be navigated and overcome when research institutions, hospitals, and, in some cases, even patients themselves can agree on data collaboration. With such a strategic view of data, it becomes possible to fill in the gaps in the data, train AI models better, and, ultimately, enhance AI-based decisions. At the same time, to ensure peace of mind over data, companies that utilize AI must opt for such techniques as anonymization, federated learning, or differential privacy. They imply that individuals’ personal data cannot be easily traced back to them, even when organizations collaborate and derive insights from shared datasets.

However, the collaboration should not be limited to data alone. In many cases, healthcare AI also needs high-performance computing and scalable cloud platforms. This is essential because training large neural networks or running complex digital twin simulations is a massive workload on hardware. This is why this hardware must be equipped with powerful servers, specialized processors or clusters of machines.

Therefore, it is crucial for both medical companies and cloud providers like AWS or Microsoft Azure to be open to partnerships or collaborative research initiatives. Our client Roche, a Swiss-based healthcare company with a focus on pharmaceuticals and diagnostics, partners with tech leaders like AWS and NVIDIA to tap into their expertise and enhance ML models with accelerated computing. This collaboration aims to speed up drug development and boost research success, uniting diverse disciplines to deliver medicines to patients faster.

Europe’s Regulatory Framework as a Strategic Advantage

Last but not least, the event highlighted the growing importance of European regulations in healthcare. Europe’s focus on ethics and patient consent fosters a disciplined, trust-centric approach to AI deployment thanks to such regulations as GDPR, which enforces strict data privacy requirements, and the EU Medical Device Regulation (MDR), which sets strict standards for software classified as medical devices. Nevertheless, this strict legal framework can be regarded not as an obstacle but as a springboard for credible solutions. This is particularly helpful in targeted areas like clinical research and physician support tools.

European healthcare companies must put emphasis on precision, quality, and ethical innovation in the use of AI. It might not be a competitive advantage in terms of scale if compared to US companies. However, in terms of privacy and data security, European AI systems will have a significant advantage, which is more valuable in such a sensitive niche as healthcare. 

This approach is further promoted by Europe’s nuclear energy advantage, which can provide low-carbon and stable energy for research facilities and data centers, leading to more sustainable and cost-efficient infrastructure for AI development. Added to that is Europe’s robust research ecosystem, where academic institutions, government bodies, and private enterprises collaborate closely on cutting-edge initiatives, often funded through pan-European programs like Horizon Europe. Thus, European healthcare innovators have the opportunity to build trust with healthcare providers and patients alike and solidify the position of their ventures on the market.

Conclusion

The use of AI in healthcare is far from simple, yet “Healthcare x AI” showed that there’s confidence in this new technology and its transformative impact on the industry. Medical institutions are exploring and even actively using AI in their work by gradually implementing it for particular features that drive the most value, like diagnosis and treatment plans, and biotech companies have the greatest potential to leverage AI for innovation and discovery. 

While AI decisions may be met with cautiousness, healthcare providers try to eliminate all the doubts about AI reasoning and decisions. This is why XAI and regulations (GDPR, HIPAA, and MDR) come into play as well as the initiatives for data and infrastructure partnerships, meaning the benefits of AI in healthcare can be fully realized only by fostering transparency, compliance, and collaboration.