Skip to content
From drug discovery to clinical trials, AI is being used, studied, refined, and advanced as a means to slingshot biotech innovations into the future at an ever-increasing speed and with ever-increasing accuracy.
But that is not to say that the technology is perfect, or its application without issue—far from it. And the reality is that the technology is still in its toddler years.
“If you think about ChatGPT, it just turned 2 years old as of last November,” said Ricardo Vilanova, Partner, AI and Intelligent Automation at Ernst & Young, LLP, during the opening AI Summit at the 2025 Biotechnology Innovation Organization (BIO) International Convention. “And how many things have happened?”
Panelists discussed where AI applications are showing the most utility in biopharma development and how executives should prepare for the next wave of applications.
AI grows up
“At Novartis, I got access to the first versions of GPT as part of our partnership with Microsoft,” recalled Iya Khalil, Ph.D., Vice President and Head of Data, A.I. & Genome Sciences at Merck & Co. “We saw GPT grow from 1 to 3. At 3, you could do some predictions. And then we literally go through a step-wise transition of being able to go from 3 to 4, where finally you feel like you’re interacting with an AI that’s intelligent and emergent.”
“We’re talking about December 2022 to January 2023,” she adds.
Though AI has shown rapid growth in a few years, that growth is a result of a tsunami of behind-the-scenes effort. And when it comes to biotech, that effort must be sustained, and effectively focused, for companies to meet projected endpoints.
“It takes a really long time to build AI infrastructure—bottom up, starting from nothing and developing into a completely different fit-for-purpose way,” said Mary Rozenman, Ph.D., CFO/CBO of Insitro.
“Just like you invest in building out your lab and lab technology and automation and generating data at scale, you also build out the infrastructure needed to support AI,” added Khalil, including teams to “leverage the data, come up with new AI models, build the AI models, and apply those AI models, and apply it.”
With different AI models come different potential uses—from early-stage drug discovery, to clinical trial development, to in-clinic care and doctor’s assistants.
“You could think about things in terms of generative models or predictive models. You could think along the value chain, from early discovery to translational to clinical trials, and you could think about things in terms of patient recruitment and how do we do trials more efficiently,” noted Molly Gibson, Ph.D., co-founder and President, Future Sciences at Lila Sciences.
‘We’re in a new era’
Take early-stage drug discovery, where AI is proving itself very useful. So useful, in fact, that it is changing the way scientists work with compounds they are trying to untangle and decode.
“We’re in a new era now where AI is not just doing the things that humans previously weren’t good at, like thinking about protein structures or chemical structures, but now, AI is also starting to enter spaces that humans are innately good at: reasoning and critical thinking and hypothesis generation,” said Gibson. “So that’s a whole new wave and a whole new era that we’re just now entering the last 6 to 9 months, and it’s starting to transform the way we think about the role of a scientist.”
And the more AI is used, the better AI becomes. In the hands of a discerning researcher able to direct AI counterparts to use and understand the right kinds of data, its growth becomes more effective, as well as more vast.
“Where do we have an abundance of data that we can start to learn from, and then learn beyond just simple statistical signals?” asked Khalil. “Those are the places where you want to go first.”
As Luba Greenwood, Executive in Residence at PureTech Health and Chief Executive Officer of Gallop Oncology, pointed out, AI can help biotech companies grow and improve, even while its users help it grow and refine its skills.
“We can use AI to find inefficiencies and leverage them as a biotech company, to get to places like clinical trials faster,” she said.
“The things that you’ve always done before may look similar, but with AI, you’re actually doing something completely novel,” added Gibson.
The real world is the next frontier for AI in biotech
“What’s nice is that a lot of things are working,” said Greenwood. But she conceded that not everything is working—yet.
“You can’t even tell if one thing is working better than the other because it is based on data, and it’s going to be pretty hard to figure out what’s working or not,” she continued. “Ultimately, our question is, Can these tools get you into the clinic?”
“And in drug discovery, most of the biotech companies that are using AI for discovery, there are very few that are going into the clinic today.”
This was the part of the conversation that highlighted where AI data sets are lacking: the real world. As Gibson noted, AI needs data loops to grow its algorithms in order to perform well.
“The places where we don’t have the data—things like clinical trials, things like human trials, things where you actually have to get into the real world—are places where we have to either figure out how to get the data or innovate in new ways,” Gibson continued. “This…is where I am seeing where the next innovation needs to happen.”
“We know AI gets better with learning groups, and AI starts to learn things that go beyond human intuition with learning groups,” Khalil echoed. “That is where what we’re doing and where we’re going.”
However, as Rozenman pointed out, biotech and its sister industries need to focus their priorities on the right areas.
“There’s an error, I think, that we have made as an industry of over-focusing on speed,” said Rozenman. “This idea that AI should make things faster – let’s just run as fast as we can. No. Let’s advance the highest-quality programs that we can.”
With the right focus, the panelists asserted, the sky’s the limit.
“We’re at the precipice of understanding disease biologies in ways we could not before,” said Khalil, explaining that AI is helping biotechs ask what’s specific or unique about this patient. “This is going to hopefully bring a host of drugs that are working in different ways, ways we hadn’t imagined and anticipated,” she continued, “ways that are better for patients.” Instead of a “blunt instrument,” you can “target the disease itself. And then as we fold in our genetic and genomic data, that will make it more precise for that patient.”
To do this, collaboration is key.
“It’s not a solo effort,” said Rozenman. “I cannot wait to see what all of us in this room do over the next decade, as these forces continue to converge.”
