- Unicorner
- Posts
- Rethinking AI in Healthcare
✍️ Notes from the Editors
Last week, we covered Ferrum Health, a startup building accessible clinical AI for hospital systems.
It’s no surprise that the role of AI in healthcare is an ongoing topic of contention. With that in mind, we took the opportunity to dig deeper into the space and learn more from Pelu Tran, CEO of Ferrum (and previously exited healthcare founder).
ICYMI: we’re hosting a seed+ founder dinner in SF on May 8. If you (or an awesome builder you know) would like to join us, RSVP soon. Our dinners are pretty small (<20 people) and fill up quickly!
Rethinking AI in Healthcare: A Deep Dive with Ferrum Health’s CEO
We spoke with Pelu Tran, CEO and Co-Founder of Ferrum Health, to explore the origins of Ferrum Health, the evolving landscape of healthcare AI, and the systemic dysfunctions that continue to plague our medical system.
Written by Neo Phuchane
A word from our partner
The OOO Summit is the ultimate gathering for entrepreneurs, creators, and founders in New York City on May 16. Join 700+ ambitious individuals for a full day of actionable insights, tactical workshops, and curated networking.
Hear from industry leaders like Andrew Yang, Sahil Bloom, Simran Kaur, and Justin Welsh as they share strategies on scaling businesses, building personal brands, and navigating the future of entrepreneurship. Whether you're bootstrapping or venture-backed, this summit is designed for you to 10X your growth.
Tickets are limited and subject to approval. Don't miss out on this opportunity to connect with the best in the business.
Save your spot today and use UNICORNER for $100 off! 👇️
Why build Ferrum?
Pelu Tran founded Ferrum Health after a close family member passed away from lung cancer, despite years of visible signs in imaging and clinical AI algorithms that could have prevented a missed diagnosis. The experience taught him the barrier to better patient care wasn’t a lack of tools, but rather hospitals' inability to safely and efficiently adopt them.
Legacy systems, security requirements, patient safety, (and) conservatism play a real difficulty in taking high-performing tools and getting them in the hands of patients.
Thus, Ferrum Health was born—a secure, on-premise platform that enables hospitals to evaluate, deploy, and govern AI models at scale without compromising data privacy or patient safety.
Product-market fit in a shifting AI landscape
Having built and sold a digital health company before, Tran understood that solving a pain point isn’t enough to succeed. People have to be willing to pay for your solution, too.
For hospitals, the value of adopting Ferrum quickly became clear.
AI has some pretty profound impact on hospital margins, provider productivity, (and) patient outcomes. Some of the most fundamental components of a hospital’s (profit and loss statement) are increasingly requiring AI to actually deliver their standard of care.
The shift from AI as a “fun toy” to an operational necessity created an opening that Ferrum fills to become a governance platform hospitals could trust.
The first major partnership
Ferrum’s first deployment was at Sutter Health in the Bay Area.
Despite being an innovative hospital system with a high-performing clinical and admin team, even Sutter struggled to implement AI. According to Tran:
It costs an average of $700 thousand to $1 million to be able to onboard a new vendor … and unfortunately with AI, there are a thousand use cases on the market.
That economic bottleneck made it impossible to scale AI responsibly. Tran states he was looking at a growing list of AI models clinicians wanted but couldn’t take advantage of due to a health system that was struggling to sustainably engage the ecosystem. In response, Ferrum offered an end-to-end platform that let hospitals adopt multiple models safely, quickly, and without risking patient harm.
Lessons from Augmedix
Ferrum was built on lessons from Tran’s first startup, Augmedix, which developed AI that listens to patient-provider conversations and automatically generates medical notes. The company was acquired by Commure in 2024 for approximately $139 million.
His first lesson was that building in healthcare is incredibly challenging, but that difficulty is part of what makes the work so meaningful. As he puts it:
If you want to make an impact, it's hard to find an area that's more impactful than healthcare.
Another crucial lesson Tran took from Augemedix was that a great sales team was the cornerstone of a successful healthcare enterprise company.
The best product doesn’t always win, the best sales team does.
When hiring salespeople, Tran states that he specifically looks for those willing to go the extra step. For him, great salespeople don’t stop at winning over a single champion. They go deep, making sure to influence everyone surrounding the decision maker in an organization.
That broader influence, he adds, is what truly sets top performers apart. Paired with the right tools, slides, brand, and an exceptional product, sales is a key ingredient to building businesses that truly scale.
As Tran states:
There are plenty of companies that have such great product-market fit that they are able to thrive without a sales team, and there are companies that have mediocre products that have amazing sales teams … But the ones that generate the vast majority of venture capital returns that become these massive names are ones that have both, and I feel like that was really something that was ingrained within me over the years that I was building Augmedix.
Scaling successfully
As Ferrum scaled, Tran’s responsibilities transitioned from serving as a generalist to strategically assembling and leading the right team.
The most impactful decision I (have) made as a sales leader is finding really, really good people and having them take (it) over.
This required him to intentionally step back from day-to-day execution and redefine his role around long-term vision, including anticipating what’s next, identifying new opportunities, and making high-leverage decisions to drive the company’s next phase of growth.
Staying ahead of the AI curve
Tran and his product teams spend a lot of time monitoring and adapting to what AI governance means in the rapidly evolving world of generative AI. Most recently, they have been considering the shift from closed- to open-source LLM models.
According to Tran, the unprecedented rise of open-source models like DeepSeek has dramatically shifted industry expectations. Whereas many once believed that large, expensive foundation models from companies like OpenAI and Anthropic would dominate the AI landscape, investors are now questioning the viability of that approach. Instead of envisioning millions of companies relying on closed APIs, there's growing momentum around open-source LLMs that offer comparable performance, greater flexibility, and lower costs.
Additionally, Tran emphasizes how these models can be deployed securely, are more customizable, and are a fraction of the cost—or even free. And in the healthcare space, this changes everything.
The problem with OpenAI and healthcare
Deploying ChatGPT is problematic for several reasons.
For one, clinicians are often uncomfortable with the idea of uploading patient records to ChatGPT, even if they’re demystified. This largely has to do with OpenAI still reserving the right to train its models with the data, which, as Tran observes, isn't “something that flies in healthcare.”
Open-source models, on the other hand, can be deployed privately. In Tran’s perspective, this eliminates what is most likely the biggest security risk frightening hospitals.
Second, there’s the challenge of making sure the model actually performs well in real-world medical settings.
I’m not asking it to write an essay on The Odyssey, I’m asking it to generate medical codes for some super specialized clinical use cases in a population that is not the average American population. So, (for example), the responses that the model needs to have in New Orleans needs to be very, very different than it has in Boston.
ChatGPT, as it stands today, isn’t fully equipped to deliver the precision, privacy, and deep contextual awareness that healthcare demands.
Competing with cloud providers and consultancies
One of Ferrum’s main sources of competition comes from hospitals attempting to build their own clinical AI solutions, often by assembling in-house data science and engineering teams in collaboration with major cloud providers.
However, Tran points out that this approach requires hospitals to invest heavily in infrastructure and manage complex integrations. They also have to handle technical components like compute, middleware, and AI orchestration—areas that fall far outside their core mission of patient care.
Then, there are firms like Deloitte and Accenture that offer outsourced IT services, but those approaches don’t replicate the kind of platform Ferrum is building, according to Tran.
The stakes for healthcare’s future
For Tran, the urgency behind Ferrum’s mission stems from what he sees as an unsustainable trajectory in American healthcare.
We’re looking down the barrel of a gun.
Tran points out what he sees as horrific industry trends, including exponentially increasing healthcare costs, the flattening curve of life expectancy, the lack of growth of providers compared to the growth of applicable data, the decline of reimbursement, and the possibility of Medicare going bankrupt within our lifetime. Given these, Tran believes fundamental changes need to be made in the healthcare system.
The system, he argues, is overburdened by inefficiencies, many of which fall on the shoulders of the very people it’s supposed to serve.
It’s not just about delivering care anymore. It’s about doctors, nurses, and even patients having to do dumb stuff like remembering when to refill a prescription, handling medical codes, or tracking down specialists … that just cannot continue to exist.
This growing administrative burden has real consequences. Tran believes we’re already seeing a decline in care quality, not because clinicians care less, but because they simply don’t have the time.
If you went to see your primary care doc 10-15 years ago, they would be seeing a much smaller panel of patients. They would have the time, the ability, (and) make enough money, to reach out and say, “Hey, I noticed you’re not getting your flu shot.” They would actually have time to care about you. We are now living in a system where we are ever-shrinking cogs in an ever-growing machine.
He warns that if things continue on this path, patients will face even longer wait times, more fragmented experiences, and worsening outcomes.
That’s where he sees AI playing a critical role in rewiring the system.
Either we build a world where AI connects every step of the patient's journey, or we end up with a system even more dysfunctional than what we have now. I’m hoping it’s the former. But it’s going to take fundamental changes, from regulation to reimbursement models, to get there.
Written by Neo Phuchane
🥲 That’s all, folks
 | Hope you enjoyed today’s founder deep dive. To check out all 200+ of our past startup covers, visit our website. The best way to support us is by visiting our sponsors—today’s sponsor is OOO Summit. See you next Monday! |
Enjoyed the founder deep dive? |
If you enjoyed today's article, forward this email to a friend!
If you're just seeing this email now, subscribe here.
Made with 💜 by the Unicorner team