As a founder building in the medical intelligence space, I’ve seen firsthand how commercializing innovation is not about hype or singular breakthroughs — it’s about understanding the entire technology stack, as well as many orthogonal critical pieces of the puzzle — not least, regulatory considerations, interoperability, security, and integration.
Building AI into the healthcare stack may be slower than in other industries, but that’s a good thing.
With the ever-increasing hyperbole around the exponential advance of AI, anyone can be forgiven for thinking that if we don’t embrace it today, we’ll be left behind. That might be true in markets where “moving fast and breaking things” is acceptable, but obviously, healthcare is not such a market.
It’s worth looking past the headlines and seeing AI as part of a technology stack — one that will inevitably require application layers, regulatory approvals, interoperability, and workflow integration before it becomes commonplace in healthcare. I’m certainly not a naysayer. Don’t ignore this “revolution” but don’t panic either.
Just like the web and mobile revolutions before it, foundation models like LLMs and multimodal systems are today’s equivalent of TCP/IP or operating systems — essential, powerful, but not complete products on their own. Real-world applications — especially in areas like healthcare — need to be built on top of those models and will come when systems, workflows, regulatory frameworks, quality processes, security and IT departments adapt.
The personal computer. The internet. The smartphone. Each of these breakthroughs felt explosive at the time – but in reality, they emerged gradually, built on foundational technologies and timed with shifts in trust and infrastructure.
AI is following this same path. Its utility, particularly in complex and regulated domains like healthcare, won’t come from the models alone. It will come from the applications, workflows, and safety protocols built on top. And the complexity of the domain, diversity of patient populations, and critical need for unbiased, secure data and interoperability, trust, and traceability, will take time.
What’s coming will be profound — saving lives and billions of dollars — in numerous applications including diagnostic AI support tools that help doctors see patterns in data that they might otherwise have missed.
It won’t feel like a revolution — until, quietly, it is.