On the same day Anthropic staged a record-shattering IPO on Nasdaq, the artificial intelligence tech giant’s lone pharmaceutical board member, Novartis CEO Vas Narasimhan, was in London pitching investors on how the Swiss drugmaker is using AI to shrink drug development timelines and one day solve “biological latency.”
Speaking during an investor event on Thursday, Narasimhan laid out Novartis’ goal to use AI to improve its medicines’ probability of success—from first-in-human testing to approval—from the industry average of 8% to 12%, while cutting four years off the standard 12-year R&D cycle from entry into research to the market.
Narasimhan, who joined Anthropic’s board in April, gave an assessment of where AI can move the needle immediately and where missions such as increasing the drug development success rate will be “a long journey.”
Narasimhan broke drug development bottlenecks into three buckets: information latency, operational latency and biological latency.
Information latency, which concerns gathering data to make decisions, currently takes roughly a year in the standard 12-year development timeline, and Narasimhan expects AI to soon drive that down to zero.
Operational latency, which covers designing and setting up clinical trials, drafting regulatory paperwork and implementing the experiments, accounts for another two-and-a-half years, give or take, that AI can steadily compress, he said.
Already, Novartis is seeing tangible gains in those workflows. About 90% of all content in Novartis development is now AI-assisted, cutting clinical report timelines by 25% as statistical programming in the company’s 2026 trials got 90% faster, Narasimhan said.
“It’s been remarkable to see with the recent readouts that we have had that our statisticians can completely analyze both positive and negative results in a day or days rather than months,” Narasimhan said, calling the depth of understanding from trial results for remibrutinib, pelacarsen, del-desiran “astonishing.”
Using an in-silico trial planning platform, Novartis can run thousands of enrollment simulations in minutes rather than a few different scenarios in weeks, Max Lawson, head of AI in Novartis’ drug development, said during a session dedicated to the technology at Thursday’s event.
“What we found interesting is those simulations are also 50% more accurate in predicting enrollment timelines even than our prior process,” he said.
AI has also helped Novartis get “much more sophisticated” with choosing clinical endpoints, beginning with neurodegeneration and neuromuscular diseases, Narasimhan noted.
Beyond R&D, Novartis expects production to see the biggest profit improvements over the next three years amid “a massive effort to scale AI in our manufacturing units,” Narasimhan said.
Meanwhile, on the commercial side, almost all field representatives now use AI tools, enabling faster patient enrollment, insurance verification and content creation.
Biological latency, however, remains the hardest nut to crack, Narasimhan argued.
AI drug developers are already good at designing better molecules with stronger pharmacological properties, but getting better at solving biological questions will take longer, he observed.
“There are folks in the tech sector who think that solving that biology and [being] able to model what happens in a cell and then what happens in 40 trillion cells in this very complex dynamic is going to be something that you can solve much faster,” to then improve the probability of success—but Narasimhan suggests that will be a “10-year-plus journey.”
That comment could be interpreted as a direct shot at a prediction made by Anthropic CEO Dario Amodei that goes: “Cancer, Alzheimer’s, heart disease, mental illness—these will be relics of the past in, I think, less than a decade.”
Narasimhan’s position that biological latency will take a long time to overcome echoes what Eli Lilly CEO David Ricks said in response to Amodei’s statement.
“I think that would be amazing, and I think it’s not likely,” Ricks tweeted on Nov. 11. “The map of biology is still largely a blank page, and randomized human proof is the long pole in the tent. I am excited about the use of AI in our sector, but progress will likely be steady and incremental.”
Fiona Marshall, Novartis’ president of biomedical research, also pushed back against the idea that AI will make major diseases obsolete within a decade. While tools like AlphaFold revolutionized protein structure prediction, converting a promising molecule into an approvable medicine still requires navigating complex human biology, which “we still are not very good at even in the context of AI," she said during the investor event.
Speaking to AI’s power, Marshall cited examples where AI has helped Novartis crack targets that the company had previously failed using conventional high-throughput screening and structure-based design methods.
“In terms of finding molecules that you don’t find before, what AI is particularly good at when you do the generative chemistry is it actually does take you into a wider area of chemical space that is not normally explored within a chemistry program,” she said.
Bringing together datasets from multiple sources, including cell-based multi-omic profiling to clinical data, AI has already helped Novartis identify three novel neurodegeneration targets that are advancing in its later-stage preclinical program. Even so, AI is not yet able to answer critical biological questions purely in silico.
“What’s very important, we believe, is having really robust ‘lab-in-the-loop’ profiling assays so that you take the output from the AI models into relevant either cell-based models or then later on even in vivo models where you can validate what comes out of the AI models,” Marshall said. “Because actually still a lot of the things that come out just don’t translate into even the cell models, let alone into the clinic.”
Because increasing R&D success rate is essentially a biology problem, Narasimhan argued that the massive amount of biological data and insights that Novartis has generated through 25 years of experiments will give the company a competitive edge.
To capitalize on that advantage, Novartis spent the past decade structuring its internal data even before generative AI went mainstream. About 10 years ago, the company began building “data42”—a nod to The Hitchhiker’s Guide to the Galaxy—to pull research and clinical data out of organizational silos and harmonize them into an ontology.
During the dedicated AI session, Novartis’ chief data & digital officer, Christian Diehl, noted that computing power and open-source models have become a “level playing field.” Instead, Novartis’ differentiator is an internal vault of more than 150 petabytes of research data, equivalent to 40 million years of 4K Netflix streaming.
On top of that AI-enabled capability is human judgment that can decide whether a correlation is a hidden causality or just noise, Diehl said.
Those two features, plus a “mindset of the willingness to reimagine the work processes that you have,” will eventually get Novartis to reduce biological latency, Diehl said. “We’re accelerating our learning on biology.”