An analysis by the Tufts Center for the Study of Drug Development (Tufts CSDD), research that was shared first with Axios, finds that agentic artificial intelligence (AI) can both accelerate development and unlock multi‑million dollar efficiencies in oncology clinical trials.
Key findings
Tufts’ analysis estimates that deploying AI agents in late‑stage oncology development can shorten clinical development by roughly 10 weeks and reduce direct operating costs by up to $5.6 million for a single late‑stage trial. The report also models larger aggregate benefits when a drug addresses multiple tumor types: for a hypothetical experimental therapy with 50 active indications, Tufts projects net benefits as high as $565 million.
The analysis applied a clinical monitoring agent from Medable — a company that offers a platform to support clinical trials — to an unspecified oncology development program running phase 2 and phase 3 studies.
What the deployment achieved
Ken Getz, executive director of the Tufts CSDD, said using the agent produced tangible efficiencies such as reducing the number of on‑site visits, accelerating enrollment into the trial and speeding up data lock.
"To our knowledge, this is the first time that [predictive] modeling based on actual use and benchmark data has been applied to quantify the net financial impact of an agentic AI solution in a drug development program," Getz said.
Why this matters
AI tools could trim time‑consuming steps — recruiting and enrolling patients, monitoring outcomes and interpreting data — freeing resources for additional studies and potentially lowering the high failure rates in drug development. Medable representatives say these agents could become standard features in some clinical trials within three to five years, operating similarly to self‑driving systems that handle repetitive but necessary record‑keeping associated with regulatory submissions.
AI may also help track trial population diversity, a frequent concern in oncology that affects whether an experimental therapy will work across broader patient groups. Medable adds that the technology can enable earlier understanding of safety and efficacy, allowing human researchers to focus on more strategic aspects of drug development.
Limitations and caveats
The analysis notes AI does not guarantee a trial’s success. Key challenges often lie outside data processing: finding the right patients, obtaining informed consent, and producing and distributing the investigational drug remain human and logistical problems. Human review and verification of the AI agent’s outputs will likely still be required, which could offset some of the time savings.
Nevertheless, Tufts and Medable’s findings indicate that properly integrated agentic AI tools could deliver substantial efficiencies, particularly in complex oncology trials that typically carry large monitoring budgets.
