The vast majority of human suffering from chronic illness stems not from a lack of scientific insight, but from the limits of how new therapies are discovered and brought to patients. Traditional drug discovery remains lengthy, uncertain, and costly, with profound implications for health systems and economic sustainability in aging societies. Extending healthspan—the period of life spent in good health—requires rethinking how we discover and develop drugs, and artificial intelligence (AI) is emerging as a critical tool to help meet this challenge.
Despite years of methodological refinement, drug development remains inefficient. Estimates indicate that about 90 percent of drug candidates entering clinical trials ultimately fail to achieve regulatory approval. Roughly 40–50 percent of failures occur because the drug does not demonstrate sufficient clinical efficacy, another 30 percent due to toxicity or safety issues, and the remainder due to poor pharmacokinetic properties or strategic considerations. These figures have been consistent across decades of pharmaceutical research and underscore the structural difficulty of discovering new effective therapies.
The cost of this attrition is immense. The average cost to bring a single new drug to market exceeds US $2 billion, with development timelines often surpassing a decade. Each unsuccessful candidate diverts billions of dollars and thousands of person-years of research effort away from more promising avenues.
A root cause of these high failure rates is that many traditional drug discovery pipelines focus on targets identified late in the disease process, after significant damage has occurred. Age-related diseases, such as Alzheimer’s disease, illustrate this challenge starkly, where trial failure rates for novel therapeutics have exceeded 99 percent over extended research periods. By the time a candidate enters clinical testing, underlying pathology is often established, limiting the likelihood that interventions can meaningfully alter disease progression.
Traditional preclinical models further constrain discovery. Animal studies and simplified cellular systems often fail to recapitulate the complexity of human biology. Compounds appearing safe and effective in these models frequently fail in human trials, contributing to high attrition despite significant investment in earlier development stages.
Artificial intelligence offers a fundamentally different approach. Machine learning algorithms can analyze vast datasets—including genomics, molecular structures, biological pathways, pharmacokinetics, toxicity profiles, and real-world clinical data. Rather than testing large numbers of compounds empirically, AI can prioritize candidates with higher likelihoods of human efficacy, reducing wasted effort on molecules unlikely to succeed.
Early evidence supports this potential. Molecules designed or selected with AI assistance have demonstrated substantially higher success rates in early clinical phases, with Phase I outcomes reportedly in the 80–90 percent range, compared with traditional Phase I success rates of roughly 40–65 percent. By de-risking candidate selection, AI shortens development timelines, reduces costs, and increases the probability of delivering clinically meaningful therapies.
AI also transforms target selection. Traditional approaches often focus on well-studied pathways, leaving more than 96 percent of the human proteome underexplored for therapeutic potential. AI-guided strategies can identify novel targets and previously unrecognized biological mechanisms, creating opportunities to address disease earlier and more effectively, before irreversible damage occurs.
Beyond discovery, AI can optimize clinical trials. Algorithms can simulate trial protocols, forecast outcomes, and refine patient stratification, improving statistical power and minimizing avoidable failures. Adaptive AI frameworks enhance trial design iteratively, further increasing the probability of success and accelerating the path from laboratory to patient.
The economic implications are profound. Reducing failure rates and accelerating timelines can lower development costs, while effective therapies that delay or prevent chronic disease reduce long-term healthcare spending on hospitalizations, medications, and long-term care. Extending functional capacity also enables older adults to remain engaged in work, caregiving, and community life, supporting broader economic stability.
Realizing this potential requires robust data infrastructure, access to diverse datasets, transparent governance, and collaboration across regulators, academia, industry, and healthcare providers. Equity and interpretability are essential to ensure AI benefits all populations rather than exacerbating disparities.
Drug discovery has historically been slow, costly, and marked by high attrition. In the context of aging populations and the urgent need to extend healthspan, this model is no longer sustainable. Artificial intelligence is not a panacea, but it represents a transformative step toward a more predictive, efficient, and cost-effective paradigm. Applied thoughtfully, AI can increase the likelihood that new therapies reach patients, reduce the economic burden of chronic illness, and ensure that longer lives are also healthier lives.