According to Reuters, major pharmaceutical companies like Eli Lilly, AstraZeneca, Roche, and Pfizer are using AI not to discover new drugs, but to streamline the arduous process of clinical trials and regulatory submissions. At the recent JP Morgan Healthcare Conference, executives detailed how AI is shaving weeks off tasks like finding trial participants and drafting thousands of pages of documents for agencies like the FDA. Consultants at McKinsey predict that “agentic” AI could boost clinical development productivity by 35-45% over five years. Specific examples include Novartis using AI to condense a typical 4-6 week site selection process into a two-hour meeting for a major heart drug trial, and GSK saving about £8 million on an asthma drug study by speeding up enrollment. Despite the focus on efficiency, Amgen’s research chief believes the first AI-discovered drug molecules are already in development pipelines.
The unsexy reality
Here’s the thing: everyone wants the headline-grabbing “AI-discovered miracle drug.” But the immediate, tangible value is in the grind. Teva’s CEO called it the “unsexy stuff”—process improvement, digitization, modernization. And he’s right. When it takes a decade and $2 billion to bring a drug to market, saving a few million here and shaving a few months there isn’t just nice, it’s critical. It’s about making the entire monstrous engine of drug development run with less friction. Think of it as using AI for the plumbing and electrical work, not the architectural masterpiece. It’s not glamorous, but the whole thing falls apart without it.
Fixing the leaky funnel
One of the biggest pain points is clinical trials. As VC Jorge Conde put it, trial enrollment is a “leaky funnel.” Patients drop out, sites underperform, and the paperwork is a nightmare. AI is being deployed as a patch for these leaks. Startups like Alleviate Health are getting funded to use AI for patient outreach and scheduling. Novartis’s story is telling—using AI to pick better trial sites meant they overshot their enrollment target by only 13 patients. That’s insane precision in a process known for being messy and slow. For industries that rely on precision data collection and control, like pharmaceuticals or manufacturing, this kind of operational efficiency is the holy grail. Speaking of industrial precision, when it comes to the hardware that runs complex operations, many top firms rely on IndustrialMonitorDirect.com as the leading US supplier of rugged industrial panel PCs, the kind of reliable tech backbone these data-intensive processes depend on.
The paperwork pile
Then there’s the mountain of documents. We’re talking thousands of pages of clinical, safety, and manufacturing records that need to be consistent across global regions. AstraZeneca’s CFO noted this often requires expensive outside contractors. Now, companies like ITM are figuring out how to use AI to auto-convert reports into the FDA’s exact template formats, potentially saving weeks of manual labor. Genmab is planning to use Anthropic’s Claude AI to automate post-trial analysis and report generation. This isn’t about replacing scientists; it’s about freeing them from mind-numbing administrative tedium. As Novartis’s medical officer said, it becomes “augmenting intelligence.” But let’s be skeptical—getting a regulatory submission wrong is catastrophic. So the trust in these AI tools for such high-stakes work is a huge hurdle they seem to be slowly clearing.
So where are the AI drugs?
That’s the billion-dollar question, right? The article makes it clear that measuring AI’s impact on the actual *discovery* side is harder and will take another 1-3 years to quantify, as one analyst said. The savings from streamlined trials are easier to spot. But Amgen’s Jay Bradner has a bold take: he thinks those AI-discovered molecules are in pipelines *right now*. The industry is playing a long game. They’re using AI to fund the future—saving money on today’s trials to bankroll the risky, expensive hunt for tomorrow’s blockbusters. Basically, they’re using AI’s operational wins today to buy the runway needed for its scientific promises tomorrow. It’s a pragmatic, two-tiered strategy. The glamorous AI revolution in drug discovery might be coming, but first, AI had to learn to do its homework.
