From Hype To Infrastructure: The Great Compression Of Health AI

Forbes

20-05-2025

Amber Nigam is CEO and cofounder of basys.ai, a Harvard-based company streamlining prior authorization for health plans with generative AI.
getty
The generative AI boom in healthcare feels inevitable. Every week, a new vendor promises to reimagine clinical workflows, redefine care management or reshape revenue cycle operations. Billions are pouring into startups, tech giants are repositioning and healthcare incumbents are scrambling to stake their claim.
But beneath the noise, the real AI arms race is quietly underway—and winning it requires more than flash, jargon or even technical prowess. It demands discipline, clinical credibility and a long view of how the industry's foundations are shifting.
As the cofounder of a health AI startup and someone who has spent a decade in healthcare and data science—from the halls of Harvard to the trenches of health plan operations—I've seen the real trenches and the real hype. And I can tell you: The AI arms race in healthcare is very real. But it's not what you think.
Let's be clear: Healthcare has been burned before. Blockchain, virtual reality, wearables: Each had moments where hype outpaced reality. I think generative AI is different, but not because it's immune to exaggeration.
The difference is that GenAI genuinely has the potential to replace rote administrative tasks, streamline care delivery and change how data moves through the healthcare system. Early proofs are real, and we are seeing firsthand streamlined prior authorizations, faster risk adjustment coding and more humanlike patient engagement.
But these examples are narrow. A model that drafts a prior authorization letter isn't the same as a system that understands patient context, payer policy nuance and clinical appropriateness at scale. Copywriting is easy. Clinically meaningful transformation is not.
Still, I see some vendors are serving up the same old drink. If a pitch centers around an AI that can "talk like a doctor" without having deep clinical oversight, be skeptical. If it relies solely on fine-tuning open models without access to proprietary, real-world healthcare data, be even more skeptical.
And if it suggests that technology can replace clinical judgment wholesale—not just augment it—you're not dealing with a serious player. You're dealing with an "intention impostor": a founder who may believe in their mission but lacks the operational, regulatory or clinical sophistication to actually deliver it.
Healthcare isn't a sandbox. It's a high-stakes, highly regulated environment where the cost of failure is patient harm, and the cost of overpromising is industry fatigue.
So, how do you separate noise from signal? Start with team composition.
True healthcare AI players blend technical excellence with clinical credibility. Look for former clinicians, policy experts and health plan operators at the leadership table—not just advisors used for window dressing.
Second, scrutinize the problem they're solving. Is it rooted in collaboration with existing workflows, or does it presume clinicians and operators will change overnight? Healthcare doesn't bend easily to technology. The solutions that work will complement human decision-making, not try to outshine it.
Third, assess their attitude toward regulation and liability. Serious players view HIPAA, CMS guidelines and AI-specific governance as the price of entry, not a hurdle to "hack."
Finally, and most importantly, watch who trusts them. Meaningful pilots, real integrations with major payers or providers as well as tight partnerships matter. The best early indicators aren't flashy logos; they're renewal rates, expansion contracts and references from customers who are hard to impress.
In an arms race, speed matters—but substance wins.
The startups and incumbents pulling ahead are moving fast and building for durability. They're investing in proprietary data partnerships today to future-proof their models. They're building modular, interoperable platforms instead of narrow-point solutions.
They're also deeply integrating into existing vendor ecosystems. The players who realize early that they must be an API, not an island, will pull ahead. Those who assume they'll own the whole stack will face slow adoption, frustrated users and dwindling investor patience.
Healthcare is the ultimate trust-based system. Every new technology must prove it deserves a seat at the table—not just once, but every day.
Perhaps the least discussed but most profound outcome of this AI arms race will be the collapse of the current vendor landscape.
Historically, health plans and providers bought solutions for prior authorization, care management, risk adjustment and member engagement from different vendors—each with siloed data, rules engines and interfaces.
I view AI as eventually breaking that model. Large language models (LLMs) and foundational healthcare AI platforms aren't task-specific; they're context-specific. The same longitudinal patient record that powers faster prior authorization can also power smarter care management workflows, risk stratification and value-based care initiatives.
What does this mean? It means many of today's seemingly distinct categories will likely converge into a few dominant layers:
• Data orchestration
• Clinical knowledge modeling
• Application-specific workflows
I believe vendors that can plug into these layers will be the ones to thrive. Vendors that can only operate in isolation—no matter how good their specific application—will risk becoming redundant.
Expect a wave of acquisitions, mergers and even outright closures over the next 24 to 36 months. Expect a few dominant AI platforms to emerge—ones that health plans, providers and life sciences companies use across functions.
In the end, healthcare won't have 50 different AI partners. It will have a handful of trusted foundations, each deeply embedded into clinical, operational and financial lifeblood.
The AI arms race in healthcare is real. But it won't be won by those shouting the loudest. It will be won by those who build quietly, partner wisely and think long term.
In a world where data, models and workflows converge, the winners won't just automate tasks—they'll transform how healthcare is delivered, trusted and experienced. Sorting glitter from gold has never been more urgent—or more valuable.
This article was co-written with CEO and cofounder Arpan Saxena, a Forbes Business Council member.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?