The pharmaceutical industry’s AI revolution isn’t arriving through miracle molecule discovery. It’s sneaking in through paperwork.
While tech headlines obsess over AI designing breakthrough drugs from scratch, pharmaceutical executives gathered at the recent JP Morgan Healthcare Conference revealed a different story. The real gains? Automating the mountain of administrative tasks that make drug development so painfully slow and expensive.
Key Takeaways:
- AI tools are compressing weeks of clinical trial site selection into hours and trimming months off regulatory document preparation
- GSK saved approximately $10.87 million on late-stage asthma drug trials using digital and AI tools last year
- Industry consultants predict AI could boost clinical development productivity by 35-45% within five years
Bringing a single drug to market typically devours a decade and $2 billion. Most of that time vanishes into tasks far removed from laboratory brilliance: tracking thousands of regulatory documents, recruiting trial participants, cross-checking data across dozens of countries. This is where AI has found its first real foothold.
“Everything else that’s around that needs to be as efficient and as small as possible,” Teva Pharmaceutical CEO Richard Francis explained. “This is where I think AI digitization, modernization, process improvement, all the unsexy stuff that we get actually quite excited about, makes a difference.”
The Paper Mountain Problem
AstraZeneca, Roche, Pfizer, and smaller biotechs like Spyre and Nuvalent all described wrestling with identical challenges. Thousands of pages covering clinical results, safety data, and manufacturing records must be compiled, verified for consistency across multiple countries, and submitted to regulators. Outside contractors typically handle much of this work at substantial cost.
AstraZeneca Chief Financial Officer Aradhana Sarin noted that documents must be cross-checked and kept consistent across geographies—a painstaking process ripe for automation.
Venture capital is flowing toward these operational bottlenecks. Andreessen Horowitz general partner Jorge Conde recently invested $4.3 million in Alleviate Health, a startup targeting what he calls drug development’s “messy middle.” Trial enrollment functions like a “leaky funnel,” he explained, with participants dropping out at every stage. Alleviate aims to deploy AI for patient outreach, education, screening, and scheduling.
Novartis Turns Hours Into Minutes
The most striking example comes from Novartis. In 2023, the Swiss drugmaker launched a massive cardiovascular outcomes trial for its cholesterol drug Leqvio, enrolling 14,000 participants.
Traditional site selection for such a trial consumes four to six weeks. Novartis completed it in a single two-hour meeting. AI identified higher-performing clinical sites and optimized enrollment so precisely that the company exceeded its participant target by just 13 patients.
“AI becomes augmenting intelligence, not artificial intelligence,” Chief Medical Officer Shreeram Aradhye said.
A company spokesperson confirmed these time savings accumulate to months over a complete drug development program.
Quantifying the Gains
British drugmaker GSK has deployed digital and AI tools to reduce manual data collection and accelerate trial enrollment, targeting a 15% speed increase across all clinical trials. The approach saved roughly 8 million pounds ($10.87 million) on late-stage studies of asthma drug Exdensur, which received U.S. approval last month.
Danish antibody developer Genmab plans to implement Anthropic’s Claude chatbot-powered AI to automate post-trial work. Hisham Hamadeh, Genmab’s head of AI, said the technology will handle data analysis and transform results into graphs, tables, figures, and clinical study reports.
German radiopharmaceuticals company ITM has developed a method to convert lengthy trial reports into FDA template formats—work that previously required several staff members and weeks of effort. The system awaits deployment.
The Waiting Game
TD Cowen analyst Brendan Smith noted that large language models like Microsoft Copilot have become fairly common for administrative tasks across the industry. Yet measuring AI’s full impact on drug development timelines may take another one to three years, he said. Quantifying savings depends heavily on how and where companies deploy these tools.
McKinsey predicted last year that agentic AI—autonomous systems requiring minimal human oversight—could boost clinical development productivity by 35% to 45% within five years.
Major partnerships continue forming. Eli Lilly has teamed with chipmaker Nvidia, betting AI will improve not just efficiency but actual drug success rates.
Still, the question everyone truly wants answered remains unanswered. Amgen research chief Jay Bradner put it directly: “What everybody’s waiting for is the AI drug. When do I get the AI drug?” His answer: “I actually think those molecules are in pipelines right now.”
The proof will come when those pipelines deliver.
Written by Alius Noreika
