Reporters’ notebook from Ruth Reader: This week, I heard from and chatted with executives in the pharmaceutical space about AI and drug development. At a small event in Manhattan’s Noho, a flock of biotech executives settled at white clothed tables to talk about the challenges and opportunities in AI. Clinical trial innovator Quant Health hosted the gathering, which was off the record, so I won’t be able to share who said what. But I can offer some insight into how top players are thinking about using AI in drug development. 

Though pharma is famously stodgy, some of the largest drug makers are transforming how they work in order to best utilize AI. What they are not doing, these executives said, is layering AI onto their existing processes, which, they said, is counterproductive and inefficient. Instead, they are having to rebuild how they develop drugs in order to meet the AI era. 

There is a lot of optimism for the use of AI in this category, but challenges remain. Pharma leaders seemed more concerned with the disconnect between the executives who want to use AI and the scientists who fear that AI will replace them. 

“When you're actually addressing the heart of what scientists are actually trying to solve,” said Orr Inbar, CEO of Quant Health, told me on the record at the event. “There becomes a bit more tension on, can we trust this tool, how does it fit into the current workflow?” He also said that these workers often don’t have the information they need to understand prospective AI tools and how they work. 

Ultimately, companies need buy-in from rank-and-file scientists in order to scale the AI systems they think will ultimately reduce the time it takes to bring a drug to market (the holy grail). 

Two executives also spoke about the emerging problem of having AI that disagrees with the company's roadmap. “What do you do when the AI tells you to kill that phase II trial?” one executive asked. Another told me that top leaders love it when AI supports their expectations, but have a harder time trusting it when it doesn’t. This fascinated me. I thought AI was supposed to be the data-backed independent voice in the room? Maybe not.

The Regulators: The Food and Drug Administration is trying to support AI in drug discovery, albeit slowly. Last year, the agency put out draft guidance on using AI-generated data in regulatory decision-making around drug safety and efficacy. This year, it announced ten guiding principles for the use of AI in drug development, forged in partnership with the European Medicines Agency. It also announced two very limited real-time clinical trials, which could help the agency understand the kind of infrastructure it will need to build in order to support the AI-assisted drug trials of the future, one of the executives said.  

The agency will have to completely reinvent how it takes in data, they said. These very small pilots—proof of concepts, really—start to deal with the nuts and bolts of rewiring how clinical trials happen with AI. But there is still a lot more work to do, I’m told. 

The biggest problem of all: AI for drug development is expensive. Two executives explicitly mentioned concerns around cost. Companies are eating up tokens, the units of text that large language models process, and the adoption of agentic AI is going to explode budgets. 

“Agentic AI is expected to drive a 24-fold increase in token consumption by 2030,” says a Goldman Sachs report issued earlier this month. We’re already seeing these trends. As Fortune wisely captured in a story last week, companies are spending more on AI than they do on labor. This week, Axios broke the news on a company that spent half a billion dollars on tokens in a single month. While some startups may be token-maxxing, many large companies aren’t keen to jump on the bandwagon. 

For this reason, one executive said, the opportunity for AI use is especially strong with small biopharmas who are nimbler than the legacy behemoths.

What this means: Smaller outfits may have an opportunity to use AI more effectively, potentially reducing their reliance on the big pharmaceutical players.

Oura’s latest ring, the Oura 5, is the company’s slimmest ring to date. The ring is 40% smaller and packed with new health features, including blood pressure reading while you sleep, night breathing tracking, and integration with Counsel Health’s medical AI chatbot. Oura Ring 5 starts shipping on June 4 at a base price of $399.

Pope Leo published a cautionary encyclical on AI this week. In it, he urges developers to build for the common good and not to erect a modern-day Tower of Babel and avoid “the idolatry of profit that sacrifices the weak sacrifices the weak, a uniformity that neutralizes differences, and the pretense that a single language — even a digital one — can translate everything, including the mystery of the person, into data and performance. The risk of dehumanization — of building a future that excludes God and reduces the other to a means — is an ancient and ever-new temptation that today takes on a technical guise.” 

American Medical Association CEO John Whyte says the challenge in health care will be whether AI preserves the physician-patient relationship.

The company has expanded its chart-aware intelligence, technology that reads through a patient’s chart history, to include everything that happens between admission and discharge. The tool synthesizes prior notes, labs, imaging, medications, vitals, and orders across every workflow.