Amgen CEO Bob Bradway is using AI to find drug candidates 50% faster

Amgen CEO Bob Bradway is not your typical biotech executive. While many of his peers are currently scrambling to justify their AI budgets, Bradway has already moved past the questioning phase and into the implementation phase. The results are measurable and significant: Amgen is now selecting drug candidates for clinical development roughly 50% faster than it was previously. This acceleration is not the result of a recent pivot, but rather the culmination of a decade-long bet on the intersection of massive datasets and machine learning.

Bradway's strategy centers on a specific type of talent he calls the 'autodidact' - the curious, self-taught individual who can master a new tool over a single weekend. To embed this mindset into the company's DNA, Bradway has required his entire senior executive team to take AI courses and spends his own weekends 'vibe-coding' to stay ahead of the curve. However, this push for rapid, generalist adaptability creates a profound internal tension. Amgen's primary competitive advantage remains its deep bench of specialized experts - scientists who understand the microscopic, delicate consequences of a single amino acid substitution. Bradway's central leadership challenge is reconciling the speed of the AI-driven autodidact with the precision of the irreplaceable specialist.

This journey began much earlier than the current generative AI hype cycle. In 2012, Bradway led Amgen to acquire DeCODE Genetics, a small Icelandic firm holding longitudinal genomic data on nearly the entire population of Iceland. At the time, the move was met with skepticism from industry peers. Even Andrew Ng, a pioneer in the AI space who would later become a collaborator, initially advised Bradway that the genetic data was not yet 'tractable' and suggested focusing on more immediate applications like computer vision. Bradway, however, played both sides of the bet. Today, that foresight has manifested in a massive technological infrastructure, including an Nvidia SuperPod humming in Reykjavik and the development of proprietary protein folding models.

Amgen is now pushing the boundaries of what is possible in molecular engineering through zero-shot antibody design. This process uses AI to engineer drug molecules without needing prior experimental examples, a feat that represents a massive leap in efficiency. Rather than simply adopting off-the-shelf solutions like DeepMind's AlphaFold, Bradway has directed his teams to look past the hype, identify the specific limitations of existing models, and build custom layers on top of them. This refusal to rely on a single approach allows the company to maintain a bespoke technological edge that is difficult for competitors to replicate through mere software subscriptions.

Looking toward the immediate future, Bradway has set an explicit timeline for the next wave of disruption: 2026. He identifies this as the year the industry must master 'agentic AI.' His vision for these autonomous systems is not one of human displacement, but of administrative liberation. He sees a future where agentic AI absorbs the 'menial' tasks that currently plague high-value researchers - filling out forms, requisitioning lab materials, and summarizing massive datasets. By automating the friction of scientific inquiry, Bradway believes AI will allow scientists to devote more time to the 'essence of their skill,' effectively turning AI into a force multiplier for human intelligence rather than a replacement for it.

Despite his optimism, Bradway is acutely aware of the cultural and psychological hurdles facing AI integration. He acknowledges that there is significant misinformation and a general human aversion to the unknown, which can breed anxiety among a workforce that fears being rendered obsolete. While he admits he does not have a perfect roadmap for managing this transition, he advocates for radical transparency to mitigate the fear of the dark. For the broader biotech and pharmaceutical sectors, the lesson is clear: the cost of being late to the AI party is high, but the complexity of being early requires a leader who can manage both the machine and the human soul of the enterprise.