The Story of Savana: The Company Building the Future of Predictive Healthcare

The Story of Savana: The Company Building the Future of Predictive Healthcare

The cafeteria conversation that changed everything happened in a hospital where Ignacio Medrano treated patients with Alzheimer’s and stroke. He was explaining machine learning to fellow neurologists, describing how computers could learn from examples rather than rules, mimicking the way children acquire language.

Their response? “Yeah, that’s cool stuff, but it’s never going to work.”

In a recent episode of Category Visionaries, Ignacio Medrano, Founder and Chief Medical Officer of Savana, shared the decade-long story of building a healthcare AI company that’s now raised $44 million. It’s a story of pivots, near-death experiences, and the patience required to build in a market where being right too early looks a lot like being wrong.

The Spark: When Statistics Meets Probability

Ignacio’s insight was simple but powerful. “Everybody is trying to predict things with data. You know, like the statistics is the past and probability is the future. And the more data you have, the more you can predict,” he explains. Food companies were doing it. The stock market lived by it. Sports analytics had transformed entire industries.

But healthcare? In 2014, healthcare was still trapped in the past, with mountains of valuable data locked away in medical records, never systematically analyzed. “Back in 2014, it wasn’t that obvious, but we thought that Maybe with all that information that gets written in the medical records every day, we could do something if we through that into the proper machine learning models,” Ignacio says.

The vision was ambitious: predict which patients would respond to specific drugs, identify who would experience disease progression, transform reactive medicine into predictive medicine. “And with that idea in mind is that we started then. So we created our prototypes, started going to hospitals and it worked.”

But working technology and having a viable business are two very different things.

The First Crisis: Nobody’s Job

Savana built something remarkable—natural language processing sophisticated enough to extract structured data from the free text doctors write in medical records. “We have to invent the methods to make this reliable from a scientific point of view,” Ignacio notes. “The idea that you can transform the free text that doctors write into variables into a database.”

The technology worked. The value proposition was clear. The problem? “The idea was great, but then who would use it? And we found that it was no one’s job to use this kind of tool.”

They’d built a solution without a budget owner. In hospitals, especially in Europe where Savana started, there was no line item for AI-powered medical record analysis. No decision-maker whose responsibility included this category. No organizational structure to support adoption.

Revenue came quickly—within a year—but it wasn’t sustainable. They had early adopters and innovators, but no path to scale. “Of course the budget were not big, especially in Europe where we started, so we had to pivot very quickly.”

The Survival Pivot: Five Years in Pharma

The pivot that saved Savana was pragmatic, not visionary. They took the same NLP technology and started selling to pharmaceutical companies. “We started selling this to pharmas because they really have this interest of knowing what is happening to patients by certain disease,” Ignacio explains.

With hospital partnerships to access de-identified patient data, Savana became a real-world evidence provider for pharma companies studying disease progression and treatment patterns. “So we stayed there as a company for five years. We survived. We’re incredibly close to die, as probably every entrepreneur would tell you.”

It wasn’t the original dream of transforming hospital operations. The impact wasn’t as direct. But it kept the company alive while healthcare slowly, painfully caught up to the possibility of AI. “And then only when the technology evolved, the mindset evolved, the culture evolved around the idea of data, healthcare data, intelligence, especially, sadly, thanks to the COVID situation. Only at that moment, the hospitals were ready to catch up with budgets and with people waiting to use our tools.”

Five years. That’s how long Savana waited for their original vision to become viable. “And that’s how we came back somehow to the original idea, which was selling this to hospitals that could then use their own information, aside from pharmas, for all kinds of use cases.”

The Product They Had to Kill

But returning to hospitals meant confronting a brutal truth about their first product for doctors. Savana Consulta seemed brilliant in theory: show doctors what their colleagues were doing in real clinical practice, enable decision-making based on collective wisdom, digitize the clinical rounds where physicians discuss cases.

“We thought that was a good idea. We were thinking about the wisdom of the crowds, the idea that if many people are thinking about that, it means something,” Ignacio says. They were trying to replicate in software what happens naturally when doctors consult each other.

The market rejected it viciously. “It was incredibly rejected by my colleagues. And the reason was very simple. The fact that the majority is doing something doesn’t mean it’s the right thing to do. And that’s quite opposite to what Evidence means. Evidence means this is the right thing, no matter if few people or a lot of people are doing it.”

They’d accidentally positioned themselves against evidence-based medicine—the foundation of modern healthcare. “In a way, were trying to go against the status quo, which is using the right evidence instead of using the majority, the wisdom of the majority.”

“And we realized before, probably before it was too late, we completely killed the tool,” Ignacio admits. They salvaged the underlying NLP technology and rebuilt it for a fundamentally different purpose: “Creating reports of disease and how patients are behaving and how disease is behaving. And that’s how we started building something more interesting, on top of which we finally could come back to our tools that helped decision making, but in a totally different way.”

The Reality of Healthcare Innovation

Ten years in, Ignacio has hard-won wisdom about what makes healthcare different. It’s not just that sales cycles are long or that regulation creates friction. It’s something more fundamental about how decisions get made.

“What differentiates healthcare is that a bigger amount of decisions that your customers will make will not be based on the rational laws of the market,” he says. He’s seen brilliant innovations fail not because they don’t work, but because the political and social dynamics of healthcare trump pure efficiency.

“A typical case would be you optimize something with which you need, I don’t know, 10 nurses less. And it doesn’t really matter because politically it’s against everything not to count on 10 nurses,” Ignacio explains. The result? “You come with an innovation that apparently is useful, but then at the end of the day, you have double expense. You now have the software and you have the people, and that keeps replicating and replicating in the sector.”

This reality extends to fundraising. Traditional VCs expect returns in three to four years. Healthcare almost never moves that fast. “The pace is lower so it’s very regulated. Cultural change needs time. And even if your tool is great, it’s going to need time,” Ignacio notes. “For us and for any company in healthcare, it’s very difficult to give returns. And in that amount of time you normally, you need longer.”

The Future: AI as Healthcare’s Gatekeeper

After a decade of pivots, near-death experiences, and killed products, Ignacio’s vision for the future is coming into focus. And it’s radical.

“Things are really going to change,” he predicts. In three to five years, people will have access to digital tools where their health data—ECGs, medical records, test results—gets uploaded and analyzed by predictive algorithms. “And the level of precision of that tool is going to be high enough so that people are going to want to go there for advice and for medical care.”

The shift isn’t just about technology. It’s about where healthcare begins. “Probably these tools, if the companies behind those things, well, will be connected to physical sites, to physical hospitals, of course, will still be in place. But the gatekeeper, I don’t think it’s going to be the hospital anymore. The gatekeeper is going to be a sequence of personalized algorithms on digital that people are going to use on a normal basis.”

Two forces are converging to make this possible. First, clinical validation. “The machine learning AI algorithms that we’ve been creating for five, 10 years, five years from now, will be finally validated with clinical trials and everything,” Ignacio explains. “Something that we don’t have today, but we’ll have it in five years. So the science behind the AI algorithm for you will be robust.”

Second, accessibility through generative AI. “Generative AI doesn’t really add new knowledge, but it improves access to already existing knowledge,” he says. “So generative AI will help these customers, these patients, navigate those complex algorithms in an easy way.”

Put them together? “I see that AI in the phone for healthcare is a reality that is very close.”

It’s a future where the skeptical doctors in that hospital cafeteria will finally understand what Ignacio saw a decade ago. Machine learning in medicine wasn’t a hundred years away. It just required the patience to survive until healthcare was ready to listen.