Only 1% of trials involve AI. Look at who runs them

Someone counted how many studies on ClinicalTrials.gov actually involve AI. It comes out to about 1%. That’s the number people will share, but the part worth reading is who is behind it.

Almost three quarters of those studies are sponsored by hospitals and universities, while industry barely shows up. Most of them are not drug trials at all: they are clinicians validating a diagnostic model on their own patients, one centre at a time.

So the picture isn’t pharma quietly transforming clinical development. It’s doctors testing whether a tool works before they trust it with anyone.

Which sits oddly next to the noise. Every week there’s a new claim about halved timelines and smarter recruitment. Then you look at what teams say when surveyed honestly, and most of them have barely started.

I recognise that distance. Getting a model to do something impressive takes an afternoon. Getting it to survive an audit, with someone willing to sign their name under the output, takes a year. That second part is the whole job, and nobody puts it on a slide.