Last month, our team traveled to Gothenburg, Sweden, for Vitalis 2026, one of Scandinavia’s leading healthcare innovation forums. Over two intense days, we spoke with physicians, startup founders, researchers, hospital IT teams, and AI specialists. And one idea kept surfacing again and again:
Healthcare is approaching an infrastructural inflection point — and the bottleneck is no longer AI models.
What matters now is the foundation beneath them: data architecture, compliance, observability, and trust.
Health Data Is Not the New Oil. It’s Shale.
One of the most memorable sessions at the conference was “Evidence-Based Data for Evidence-Based AI.” The speakers challenged the popular analogy that “data is the new oil.” In healthcare, they argued, data is more like shale: abundant, fragmented, inconsistent in quality, and extremely expensive to refine into something useful.
That framing explains why so many healthcare AI initiatives struggle despite impressive models and funding.
A second insight was even more unsettling. Multiple organizations are already moving toward policies where most software is generated by AI. We’re entering a world where the systems powering healthcare may become too complex for humans to fully inspect line by line.
That breaks the traditional chain of:
Explanation → Understanding → Trust
And in healthcare, where audits, certifications, and patient safety depend on that chain, the implications are enormous.
The answer is not to slow down AI adoption. The answer is to build systems where:
- Outcomes can be independently verified
- Data provenance is traceable
- Policies are enforced automatically at execution time
In medicine, that is not philosophy. It is operational necessity.
Send the Algorithm to the Data
Another major theme at Vitalis was data sovereignty and zero-trust architecture.
The traditional healthcare model assumes data must be centralized to be useful. But many teams are now inverting that logic:
Send the algorithm to the data, not the data to the algorithm.
In practice, this means hospitals run AI inference locally inside their own environment. Only aggregated or verified outputs leave the perimeter.
This architecture aligns naturally with the regulatory direction of:
- GDPR
- NIS2
- EU AI Act
- EHDS (European Health Data Space)
These frameworks are converging on a shared principle: sensitive health data should remain under institutional control, with strong guarantees around access, traceability, and purpose limitation.
For Ukrainian HealthTech companies, this is a strategic opportunity. Teams that design for these constraints early will move faster internationally because compliance becomes embedded in the platform instead of bolted on later.
What People Were Saying in the Hallways
Some of the most revealing conversations happened outside the official sessions. A few themes came up repeatedly:
- “We have more data than ever — and less trust in it than ever.”
- Replacing an EMR is now a strategic architecture decision, not just a vendor choice.
- Compliance only becomes painful when it is added after the fact.
- Most healthcare AI failures stem from poor data quality, weak labeling, or workflows that ignore how hospitals actually operate.
Real-world examples were discussed repeatedly, including the well-known issues around Epic’s sepsis prediction model and Obermeyer’s research on algorithmic bias in healthcare. These cases shifted the industry conversation from “Can AI work?” to “How do we build systems that remain trustworthy in production?”
The Ukrainian Context
Ukraine’s medtech ecosystem is growing rapidly: telemedicine, healthcare information systems, and AI-assisted diagnostics are already moving beyond pilot stage.
But many teams are still building on foundations that will struggle under regulatory scrutiny or enterprise-scale adoption.
A few examples:
- Scaling into EU markets is not just a legal problem — it is a cloud architecture and DevOps problem.
- Running AI on sensitive medical data without exporting raw records is a platform engineering challenge.
- Making every processing step transparent and auditable is fundamentally an infrastructure and SRE problem.
At Gart Solutions, this is the layer we focus on with healthcare and medtech teams: compliant cloud infrastructure, Kubernetes environments, secure CI/CD pipelines, zero-trust access, and operational governance that can survive real audits and real hospital integrations.
The Real Takeaway from Vitalis
We went to Vitalis expecting conversations about new healthcare AI models.
What stood out instead was a broader realization:
The competitive advantage in healthcare AI is increasingly architectural.
Organizations that build compliance, scalability, observability, and data sovereignty into the foundation from the beginning are not slowing themselves down. They are removing the friction that stops others from scaling.
In healthcare, AI is not just a model.
It is a foundation.
