How do we apply the rapid-fire efficiency of generative artificial intelligence to a field where the cost of error is not a broken webpage, but a human life? This is the central scientific and operational question facing medicine today. At the annual Adobe Summit in Las Vegas, where more than 14,000 people convened this week, the intersection of technology and healthcare took center stage. The consensus among clinical leaders was clear: while algorithms can optimize processes, the highly individualized nature of human biology demands a level of precision and trust that technology alone cannot replicate.
Overcoming the Inertia of Legacy Healthcare Systems
To understand why digital transformation has lagged in medicine, one must look at the structural barriers that define the industry. Tory Smithe, who leads digital strategy for healthcare at Adobe, points out that strict regulations, siloed data systems, and legacy technologies have historically made healthcare organizations deeply conservative. Yet, this caution is not without merit, given the high stakes of patient care. Jigar Shah, Chief Medical Officer of Blue Shield of California, argues that the path forward lies in reframing these barriers. "The intent behind regulation is to protect the consumer," Shah noted, suggesting that companies should view compliance as a springboard for innovation rather than an obstacle.
This philosophy is already yielding practical applications. For instance, Blue Shield of California implemented price transparency measures well before they became a regulatory mandate. Today, their proprietary app proactively alerts consumers when a physician prescribes a medication, informing them if equivalent, lower-cost alternatives are available to prevent unexpected financial shock at the pharmacy. Lesley Spellmeyer, who manages personalization for Lilly, notes that such proactive communication is vital because healthcare providers and patients now expect the same seamless digital experiences they encounter as everyday consumers.
What the Technology Actually Does vs. What Headlines Claim
Popular media headlines frequently depict a near-future where AI diagnostic tools autonomously treat patients and write medical textbooks. However, a look at how leading institutions actually deploy these tools reveals a much more disciplined reality. At the Cleveland Clinic, which maintains a massive public library of medical literature, AI is used strictly as an operational scaling tool rather than an autonomous content creator. Amanda Todorovich, who manages this collection of medical articles, emphasizes that the institution does not use AI to generate any consumer-facing content or medical images.
"We don't publish a single thing without medical review," Todorovich stated, highlighting the strict boundary between algorithmic assistance and clinical validation. The methodology here relies on using technology to streamline backend workflows, while preserving human expertise as the final, absolute gatekeeper. As Smithe observed during the summit panels, maintaining patient trust requires keeping a human face at the forefront of healthcare communication, with clinicians remaining the primary, authoritative voice.
Re-engineering the Bottleneck of Drug Development
Beyond patient communication, the most profound scientific application of AI is occurring in the grueling arena of clinical trials. Over the last decade, the number of potential drug candidates has nearly doubled, yet the number of therapies approved by the FDA has remained stubbornly flat at approximately 50 per year. This massive discrepancy reveals that the primary bottleneck in modern pharmacology is not drug discovery, but the complex, expensive clinical trial process where the vast majority of candidates fail.
To address this systemic inefficiency, computational biologist Benjamine Liu and computer scientist Linhao Zhang cofounded Formation Bio in 2016. Rather than seeking to discover new molecules from scratch, the startup is acquiring a portfolio of 10 early-stage drug candidates that stalled or failed in early clinical trials, using AI to optimize and accelerate their clinical development. The venture has captured the attention of major Silicon Valley investors, including Andreessen Horowitz, Sequoia, Thrive Capital, Kleiner Perkins’ chairman John Doerr, and OpenAI’s Sam Altman, raising $615 million at a $1.8 billion valuation. Michael Moritz, the former Sequoia chairman who wrote the first check for the business, noted the startup's potential to disrupt an industry that has resisted fundamental structural change for decades.
Limitations to Consider
Despite the excitement surrounding algorithmic drug development, significant limitations remain. AI can optimize trial design, streamline patient recruitment, and predict molecular behavior, but it cannot alter basic human biology or bypass the physical reality of clinical testing. A drug that fails due to fundamental toxicity or lack of efficacy cannot be saved by software. Furthermore, legacy data silos mean that historical clinical trial datasets are often fragmented, limiting the accuracy of the machine learning models trained on them.
Even when a scientific breakthrough succeeds, navigating the healthcare system's infrastructure presents monumental hurdles. Dr. Katherine High, a co-recipient of the $3 million Breakthrough Prize in Life Sciences alongside Jean Bennett and Albert M. Maguire for developing the landmark gene therapy Luxturna, points out that transformative treatments face severe challenges regarding payment models. Luxturna, which was commercialized under Spark Therapeutics before the company was acquired by Roche for $4.3 billion in 2019, offers a one-time cure for a rare blinding disorder, but its high upfront cost strains traditional insurance frameworks. This structural reality applies equally to newer frontiers, such as Eli Lilly’s recent acquisition of Kelonia Therapeutics for $3.25 billion upfront (with the deal potentially worth up to $7 billion including milestones) to develop in vivo CAR-T therapies, as well as their $2.4 billion purchase of Orna Technologies earlier this year.
The Next Signals to Watch
As these technologies move from theoretical frameworks to clinical practice, the coming months will offer clear indicators of their true efficacy. Rather than watching stock fluctuations, the key metric to monitor is the clinical trial progression rate of Formation Bio’s 10-drug portfolio. The success or failure of these specific trials will serve as a measurable signal of whether machine learning can truly solve the clinical development bottleneck, or if the biology of the human body will continue to dictate its historical, slow-moving pace.
