The one-click clinical trial
Starts with a one-click protocol
You’ll just have to read today’s episode of Clear Thinking. If you like it - subscribe for more of my writing.
👉 If you enjoy reading this post, feel free to share it with friends! Or feel free to click the ❤️ button on this post so more people can discover Clear Thinking. 🙏
This week on Life Sciences Today
What if clinical trial protocols could go from months of track changes in Word to a clean protocol generated by AI?
This week, I sat down with Pedro Coelho, founder and CEO of Biorce, to explore how his team went from tagging mistakes to building AI that gets it right the first time.
Biorce is the first company to submit an AI-generated protocol to the FDA.
Pedro breaks down how Biorce can generate a best-in-class clinical trial protocol in under 5 minutes, what is the industry’s biggest anti-pattern, and what a “one-click clinical trial” could look like by the end of 2026.
Clarity of thought
The first thing that stands out in a conversation with Pedro is his ambition, and the sharpness of his thinking.
Biorce didn’t start by designing AI prompts to write clinical protocols.
They started by identifying problems - the first is a tough fact of life that everyone involved in clinical trials knows - the number of changes to the protocol during a clinical trial.
The number of changes, the messiness of track changes, multiple versions and implementation into the clinical trial data collection systems for sites and patients is a nightmare for life sciences companies/
Someone told Pedro early on: you’re like McKinsey finding problems and not fixing them.
That comment helped trigger a pivot. Instead of only spotting flaws, Biorce began using AI to correct and improve protocols directly.
The result is a system that generates clinical protocols in 5 minutes; protocols that are 85–95% complete, more robust than the usual process with Word, and less likely to require endless revisions in the first six months.
I was involved in 70+ clinical trials and I know what a mess trial amendments can be.
When Biorce scales it will be a major change to a major blocker in clinical trials.
The anti-pattern: believing in things that don’t work
Pedro’s biggest anti-pattern in clinical operations is the continued belief that risk-based remote monitoring really works.
A little background first.
Monitoring is a manual activity that involves people (CRA - clinical research associates) going to research sites and comparing pieces of paper with digital records.
Monitoring can consume 20–30% of a trial budget, while updating less than 3% of the data points ( a study Medidata did on 14,000 drug trials over 10 years ago). At FlaskData we analyzed 10M+ data points in medtech trials and found that the number updated by CRA monitoring activity was < 1%.
You could save $20-30M in a Phase 3 oncology trial and sustain 97% data quality by doing nothing.
In 2009, FDA issued guidance for RBM (risk-based monitoring).
During COVID, the industry went to RRBM (Remote risk-based monitoring, doing the same thing over zoom) and then to RBQM (Risk-based quality monitoring, doing the same thing but prioritizing the massively non-value add activity under the guise of improving quality).
Monitoring is part of a broader anti-pattern that people in drug development know but rarely say out loud: huge parts of the system persist because they are inherited, not because they are effective.
Biorce is interesting because of the clarity of their thinking, their logic and execution.
They’re part of a wider movement of clinical software startups challenging assumptions and rocking boats of bureaucracy.
Listen to the show here
You need a good nights sleep to think clearly
Anyone who dealt with a cyber event appreciates the value of a good nights sleep.
I know. I once got a 2AM call from a customer ICU, back in the day.
FDA-cleared devices in hospitals are attacker entry points.
Today, hospitals are attacked by AI-driven ransomware-as-a-service platforms.
Cyber risk doesn’t stop with FDA clearance.
It stops the night you can sleep well.