Something keeps me up at night that most people never need to consider. Currently, there are 10,000 human diseases without treatment. At the current rate of around 50 treatment approvals per year, it would take centuries to treat all known human illnesses.
Drug discovery is not the main problem. Industry investments are heavily concentrated in early-stage science, like finding new molecules and identifying new targets. The bottleneck is the limited capacity of human clinical trials. If we want to find treatments for thousands of untreated diseases in any reasonable timeframe, we need to solve that problem.
I first started working on solving the problem when conducting clinical research as a resident at Stanford nearly 10 years ago. Clinical trials then were rate-limited because we couldn’t engage enough participants. Most research efforts were confined to a city or even a single site, while patients were globally distributed. We had promising therapies. We had willing researchers. What we didn’t have was a way to run trials at the scale the science required.
The digital revolution brought a paradigm shift. Technology expanded access to clinical research, connecting patients to therapies they otherwise might never reach.
While digitization decreased patient burden and opened access, it revealed the next layer of the challenge. The system is still constrained by the operational complexity of running human trials at scale. About 80% of clinical trial work is still painstakingly manual and dependent on extensive low-value human labor; logging into a dozen systems, chasing data across fragmented workflows, and reconciling discrepancies for accuracy and completeness.
This rigor exists for a reason. Precision matters when lives are on the line. The consequence is a system that simply cannot scale in its current form.
Each technology wave I’ve lived through has taught me the same thing: Science is rarely what holds us back. The operating paradigm is what limits us.
In October 2024, Dario Amodei, the CEO of Anthropic, wrote an essay called “Machines of Loving Grace.” In it, Amodei envisions a “compressed 21st century” in which AI-accelerated biology delivers a hundred years of scientific progress in five to 10 years. He argues that artificial intelligence will lead to better targets and better molecules, which in turn will lead to therapies so effective that the clinical trial system naturally speeds up in response.
I share his ambition. We all should.
Every generation of clinical research has had promising science that the system can’t execute fast enough. What’s new is that AI-driven drug discovery is accelerating the supply of candidates into a pipeline that was already overloaded. The science is getting faster, but the system that tests that science hasn’t kept pace. That doesn’t resolve itself just because the molecules are better. We also must transform clinical trials.
This is the kind of problem you only fully understand from the inside. From the outside, clinical trials look like a science problem. From the inside, we understand the operations challenge of extraordinary complexity—thousands of sites, millions of data points, fragmented systems, and human workflows designed for a different era.
The question isn’t whether we can discover better therapies. That helps for sure. But if we want to get them to people faster and deliver a compressed 21st century, we need to build an operational infrastructure capable of proving the science works in humans, fast enough to matter.
This is where agentic AI enters the picture. It’s the beginning of the next paradigm shift—one that could be as consequential as digitization was a decade ago.
Agentic AI now enables us to easily work across the fragmented, manual workflows of clinical operations. We can build agents that orchestrate in real time—automating repetitive tasks, continuously reconciling data, surfacing risks as they emerge, and ensuring protocol adherence without constant human intervention. This isn’t theoretical. It’s happening now, and it immediately frees up capacity. Clinical teams are shifting from administrative work to more impactful decision-making.
Agentic workflows are only the beginning. The real destination—if we’re serious about the compressed century—is clinical operations that are largely autonomous. This does not mean they are unregulated or unsupervised. Instead, they operate within rigorous guardrails so humans can focus on decisions requiring clinical judgment.
The operational machinery that currently consumes 80% of the effort would all but run itself. Autonomous clinical research could unlock a new era of progress and make possible a dream that animates researchers to launch a clinical trial in one day, enroll participants in one day, and close out the trial in one year.
The FDA’s 2025 draft guidance on AI in regulatory decision-making already points in this direction. It introduces a risk-based, lifecycle-oriented framework that distinguishes between operational systems that can continuously learn and safety-critical decisions that require the highest scrutiny.
The compressed century doesn’t happen with better science alone. It happens when we build an operating system that can finally keep up with the science.
The paradigm has changed before. It can change again, but only if we treat this as a shared problem.
Regulators need to keep moving in the direction the FDA’s draft guidance signals, creating frameworks that distinguish between operational autonomy and clinical decision-making. Technology companies need to build for the reality of clinical research, not the abstraction of it. Pharmaceutical and biotechnology companies need to be willing to rethink the operating models they’ve relied on for decades. Clinicians—the people closest to patients—need to be at the center of how these systems are designed, not an afterthought
The treatment of 10,000 diseases is waiting. The science is ready and moving fast. Now let’s build the clinical trial system to match the pace of the 21st century.
Michelle Longmire, MD, is cofounder and CEO of Medable.