Conference talk · AI in healthcare · VivaTech 2019

AI in Health: Past, Present, Perspectives

VivaTech 2019 · Microsoft Health & Life Sciences · English

Historical document — 2019. This conference talk was delivered at VivaTech 2019 while Jérôme Vetillard was Director of Health & Life Sciences at Microsoft France. The observations on AI readiness in healthcare, data quality constraints, and the gap between technological capability and operational deployment remain structurally valid — and directly inform the architecture of TweenMe and the PREDICARE programme six years later.
VivaTech 2019 · Microsoft · English
🇫🇷 Cette intervention est également disponible en français — Voir la version française →
"The question is no longer whether AI can perform clinical tasks. It already can. The question is whether the data infrastructure, the regulatory frameworks, and the organisational cultures are ready to support industrial-scale deployment."

Talk structure

Past — The trajectory of AI in medicine

From early diagnostic algorithms to deep learning: a structured retrospective on the milestones that made AI clinically relevant. Expert systems, medical imaging, genomics, and the turning point of large labelled datasets. Why medicine lagged behind other sectors, and what changed.

Present — What AI can and cannot do in 2019

State of the art in 2019: AI-assisted radiology, sepsis prediction, drug discovery acceleration, population health analytics. The Microsoft Health & Life Sciences perspective on deployment patterns across European health systems. The gap between proof-of-concept and operational production — the data quality problem, interoperability barriers, and governance constraints.

Perspectives — What needs to happen next

Conditions for industrial-scale AI deployment in healthcare: standardised, ambient, high-quality data; regulatory frameworks adapted to AI as a medical device; organisational transformation of clinical workflows; and the shift from AI as a tool to AI as infrastructure. The framing that would directly shape PREDICARE and TweenMe.

Reading it in 2025

The three conditions identified in 2019 — data quality, regulatory readiness, organisational transformation — are precisely the three problems that PREDICARE addresses at territorial scale and TweenMe solves technically. The HDLSS challenge (High-Dimensional Low-Sample-Size), the heterogeneity of clinical data sources, the need for standardised digital twin generation: all of this was named here, six years before the operational solution.

VivaTech 2019 is one of the earliest public formulations of the problem space that the Twingital Institute now works to resolve.

The problem was named in 2019.
The solution is operational in 2025.