AI Is Changing a Lot in Healthcare. Drug Development Isn't One of Them — Yet.

Every few years, something comes along that's going to "revolutionize" drug development. In the early 2000s it was genomics. Then it was combinatorial chemistry. Now it's AI.

I'm not here to tell you AI isn't a big deal — it is. But after nearly two decades investing in global healthcare, I've learned to ask one question before getting swept up in any new wave: what exactly is it changing, and when?

When I apply that question to AI and biotech, the answer is a lot more nuanced than the headlines suggest.

The biology doesn't care about your model

Here's the thing that gets lost in a lot of the AI-in-drug-discovery enthusiasm: most drugs don't fail because researchers couldn't find the right target fast enough. They fail because biology is messy, unpredictable, and deeply resistant to shortcuts.

A molecule that looks perfect in silico — clean binding, great selectivity, no obvious red flags — can still blow up in a Phase II trial because the human body does something unexpected. Off-target effects. Metabolic degradation. Immune responses nobody predicted. These aren't data problems that a better algorithm fixes. They're fundamental features of biological complexity.

AI is genuinely good at pattern recognition across large, well-structured datasets. Drug development, especially in novel mechanisms and rare diseases, often happens at the exact edges of what's known — where you don't have the training data to begin with.

The regulatory clock isn't going anywhere

Even if AI cuts the early discovery phase in half — and some companies are making real progress there — you still have to run the clinical trials.

A Phase III oncology study still needs years of patient follow-up to generate meaningful survival data. A rare disease program still has to find and enroll a small, globally dispersed patient population. The FDA still needs to see a safety record built over time. None of that compresses because you used a better machine learning model upstream.

This matters a lot when you're evaluating companies. If the investment thesis depends on AI shrinking a ten-year development cycle to five, that's a bet on something that hasn't been demonstrated. And in my experience, the market tends to price that in about three cycles before it actually happens.

Where I do think AI is making a real difference

I don't want this to read as pure skepticism, because that's not where I land.

There are places where AI is already doing genuinely useful work in healthcare — just not always the places getting the most attention.

Patient stratification is one. Better ML models are helping identify which patient populations actually respond to a given therapy, which reduces the late-stage failures that have historically been so brutal for biotech investors. Biomarker discovery is another — the ability to process genomic, proteomic, and metabolomic data at scale is surfacing signals that weren't visible before.

Manufacturing and process optimization is probably the quietest story but one of the most real. Stability prediction, formulation optimization, quality control — these are areas where AI is generating compounding efficiency gains without running into the regulatory friction that slows clinical development.

And then there's diagnostics. Medical imaging, pathology, clinical decision support — this is where I think the AI story in healthcare is most legitimate right now, and it's already moving from pilot programs to actual commercial deployment.

These are real contributions. I just think they're more accurately described as meaningful improvements to an existing pipeline, not a reinvention of one.

What this means for how I think about healthcare investing

The reason I spend time on this isn't to be contrarian for its own sake. It's because the gap between narrative and evidence is where a lot of money gets lost in this sector.

Healthcare has always attracted waves of enthusiasm around technology. Each wave brings real advances. Each wave also brings a cohort of companies that get valued on the promise of compression — faster timelines, lower failure rates, shorter paths to approval — that biology and regulation don't ultimately deliver on schedule.

The durable investment case in healthcare doesn't depend on any of that. Aging populations, rising chronic disease burden, expanding healthcare infrastructure in emerging markets — these are structural, long-horizon trends that don't require a technological leap to play out. AI will be part of the story. I just think it'll be one chapter, and we're still in the early pages.

Alex Forschner is the President of Exome Asset Management LLC, an SEC-registered investment adviser focused on global healthcare investing.

The views expressed in this article are solely my own and do not represent the views of Exome Asset Management LLC or any of its affiliates.

Past performance is not indicative of future results. For qualified investors only. © Exome Asset Management LLC