Why Healthcare’s AI Revolution Is a Strategy Problem, Not a Tech Problem

A conversation with Matthieu Divet — Healthcare AI Manager at Alcimed and AgriTech Startup Founder — on bridging the gap between engineering, high-stakes consulting, and the entrepreneurial frontier of healthcare innovation.

By Louise Servoin · 2026-05-01 · 10 min read

## Matthieu Divet – Healthcare AI Manager at Nautilus·ai by Alcimed | Founder of TacoTech

Expert in Life Sciences innovation and AI implementation. Centrale Paris and Georgia Tech alumnus. Leader at Alcimed’s healthcare digital & AI activity, Nautilus.ai, and AgriTech startup founder, bridging the gap between data science, software, and life sciences strategy. Based in New York, United States.

Some professionals understand the code. Others understand the boardroom. Rarely do you find someone who can navigate both while operating in the high-stakes environment of American healthcare innovation. Matthieu Divet’s trajectory — from the rigorous engineering halls of Centrale Paris and Georgia Tech to managing complex AI portfolios for global Pharma and Biotech leaders at Nautilus·ai by Alcimed, a consulting firm specializing in innovation and the development of new markets, all while founding his own agritech venture, TacoTech — places him at a unique intersection of technology and entrepreneurship.

We sat down with Matthieu to discuss why most “AI revolutions” in healthcare stall at the pilot phase, how being an entrepreneur makes him a better consultant, and why the future of the industry depends on translating raw data into human strategy.

## The Hybrid Architect

Matthieu, you have a dual identity as both a high-level innovation consultant and a startup founder. How does the “builder” mindset of an entrepreneur influence how you advise massive healthcare organizations?

They are two sides of the same coin. In the startup world, with TacoTech for example, you learn very quickly that technology is not valuable if it doesn't solve a friction point. Technology is a means to an end. And building new technology can be very resource-consuming, so you have to be solution-oriented, lean, fast, and obsessed with building for the user.

When I step into my role at Alcimed to advise a multinational Pharma or Biotech firm, I bring that same “builder pragmatism”. Large organizations sometimes get enamored with the idea of AI — they want “innovation” as a concept. My job is to act as a reality check — the goal is to build concrete, valuable and long-lasting solutions. So, with my clients, our starting point is always: “What is the specific business problem we are solving?” Then, once the answer is clear, and only then, do we start talking about the tech.

So, whether you are a two-person startup or a global Pharma leader, the goal is the same: building solutions that convert data into actionable decisions and eventually improve patient outcomes or make operations run more smoothly.

## The Foundations of Technical Leadership

Your background at Georgia Tech and Centrale suggests a deep technical foundation. In an era where AI is often treated as “magic,” how important is it for managers to actually understand the underlying engineering?

It’s critical. The opportunities that recent AI tools have opened are amazing. But if you don't understand the limitations of the models underneath — the bias, the hallucinations; or the infrastructure required to scale a solution, it is very hard to lead a project effectively.

My academic journey provided the bedrock for this perspective: Centrale instilled a rigorous engineering discipline, a structural integrity in problem-solving that is often dismissed when companies feel the pressure to “ship fast.” While that speed might win a sprint, that underlying rigor is what proves critical in the long run, especially when dealing with high-stakes environments where failures are not an option. Complementing this, Georgia Tech reinforced a “problem-solving first” mentality that bridges the gap between theory and execution.

In healthcare, the stakes are too high for black boxes. When I manage AI projects at Alcimed, this technical background allows me to be a translator. I operate at the interface between data scientists, talking about hyperparameters and architecture, and business stakeholders who want to understand why this specific model will reduce their time-to-market for a new therapy. It’s for this reason that my teams and I often favor explainable AI over more advanced black-box models, because explainability matters for our clients.

“Innovation in healthcare isn't about having the best algorithm; it's about having the best integration of that algorithm into a human workflow.” — Matthieu Divet

## The Implementation Gap

You’ve seen the healthcare AI landscape from both the European and American perspectives. Where is the “bridge” breaking down between having great data and actually using it?

Despite different regulatory frameworks, with Americans favoring speed and Europeans prudence, I believe American and European companies actually share the same challenge in the end. The bridge usually breaks at the “last mile”: adoption. Most companies have plenty of data, and many have built impressive pilot programs. But moving from a successful pilot to a standardized tool that a doctor or a researcher uses every day? That’s where the friction is.

In healthcare, on both continents, we have regulatory hurdles, data privacy concerns, and, most importantly, a high-pressure environment where practitioners don't have time to learn a complex new interface. The gap isn't technical; it’s strategic. We need to stop building “tools” and start building “solutions.”

## Entrepreneurship and AI

As the founder of TacoTech, you’re navigating the entrepreneurial side of tech. How do you think AI is changing the barrier to entry for new founders in the healthcare space?

It’s democratizing it for sure. Ten years ago, if you wanted to launch a health-tech venture, you needed a massive team of developers and data cleaners. Today, AI-powered tools allow a small, agile team to do the work of twenty people.

But it’s not a magic wand. Just because you can build something faster doesn't mean you should. For TacoTech and other ventures I advise, I always emphasize that AI should be the engine, not the driver. The driver must be a deep understanding of the domain knowledge, the healthcare ecosystem in this case — how clinical trials are run, how payers reimburse, and how patients actually behave.

## The Question of the Future

As AI begins to commoditize “average” knowledge, what is the one human skill or nuance in your field that you believe an algorithm will never be able to replicate?

Empathy-driven strategy and field investigation — navigating the intersection of complex, fast-evolving healthcare ecosystems and disruptive innovation — requires more than just processing power. An AI can find a pattern in a million patient records, or scour the internet to find useful information, but it cannot sit in a room with a client and understand the cultural resistance to a new technology or the nuance in a physician’s opinion about a new therapy. It can’t feel the “weight” of a strategic pivot and has trouble finding a consensual answer where there is limited and nuanced data.

In consulting and entrepreneurship, the most valuable skill is judgment. It’s the ability to look at the data, even scarce, listen to the human intuition, and decide which risk is worth taking. AI gives us the map, but the human still has to choose the destination.

## The Golden Thread

We often see a gap between what technology can do and what professionals actually adopt. How do we ensure AI becomes a “co-pilot” that elevates human judgment rather than a tool that encourages intellectual passivity?

There are two levers here: how we design AI systems, and what we design them for.

On the “how,” the key is to deliberately keep the human in the loop. The best AI systems are not decision-makers — they are decision accelerators. They should challenge assumptions, surface insights, and reduce cognitive load, but always leave the final judgment to the human. If a system removes the need to think, it’s poorly designed.

On the “what,” I believe one of the most transformative areas is education. AI has the potential to shift us from a one-to-many model to a truly personalized, n-to-1 model — where each individual can effectively learn from the aggregated knowledge of thousands of experts. In that sense, AI is as disruptive as the invention of the printing press. But the goal shouldn’t be to replace thinking — it should be to elevate it, by adapting to how each person learns and reasons.

Ultimately, this is a design responsibility. You can train users to avoid passivity, but behavior follows structure. If you build systems that reward critical thinking, users will stay engaged. If you build systems that optimize for convenience alone, you risk creating dependency

Tags: AI in Healthcare, Pharma Innovation, Alcimed, Georgia Tech, Data Strategy, HealthTech Entrepreneurship