A Tufts CSDD analysis estimates that an AI clinical monitoring agent could reduce trial costs and shorten drug development timelines by around 10 weeks.

AI agent assessed using oncology trial data

A new analysis from the Tufts Center for the Study of Drug Development (CSDD) has assessed the potential financial and operational impact of using an AI agent for clinical trial monitoring.

The analysis examined Medable’s Clinical Monitoring Agent using benchmarked oncology programme and clinical trial data from Tufts CSDD alongside operational and contract data supplied by Medable.

Tufts CSDD estimated that use of the agent could increase expected net present value (eNPV) by approximately $7.5 million for a Phase 2 trial, $11.3 million for combined Phase 2 and Phase 3 development and $21 million for a Phase 3 trial.

The analysis also estimated reductions in direct on-site monitoring costs of approximately $4.4 million for a Phase 2 trial and $5.6 million for a Phase 3 trial.

The findings are based on modelling rather than a published peer-reviewed study. Tufts CSDD and Medable said a more detailed peer-reviewed paper is expected later in 2026.

Clinical monitoring could be accelerated

The analysis found that the use of an AI monitoring agent could reduce clinical development timelines by approximately 10 weeks.

Faster patient enrolment was identified as one contributor, with enrolment timelines estimated to be reduced by approximately 109 to 119 days.

The analysis also estimated that database lock could occur around two weeks earlier by reducing the time required for trial closeout activities.

Clinical trial monitoring involves reviewing trial data and site activity to identify issues that could affect patient safety, data quality or study conduct. Traditional monitoring can require clinical research associates to make regular visits to trial sites.

According to the analysis, an AI agent could reduce the number of on-site monitoring visits and associated travel costs while allowing clinical research associates to spend more time on other activities.

Tufts CSDD executive director Ken Getz said: “The financial value created by the investment and deployment of the monitoring agent was driven by operational efficiencies such as the reduction in the number of on-site visits and reduced travel costs as well as accelerated enrollment and database lock timelines.”

The analysis also identified approximately $600,000 in administrative off-site monitoring efficiencies for a Phase 2 trial and $1.7 million for a Phase 3 trial. These figures were not included in the eNPV calculations.

Evidence for agentic AI in clinical trials

The findings add to growing interest in using agentic AI to automate parts of clinical research, including activities traditionally requiring significant manual input.

Unlike conventional software tools that perform defined tasks, agentic AI systems are designed to carry out sequences of tasks with a greater degree of autonomy.

For clinical development, potential applications include monitoring trial sites, identifying operational issues and supporting researchers with routine activities.

However, the Tufts analysis does not establish that AI monitoring agents will deliver the estimated financial gains across all clinical trials.

The modelling used data from an oncology development programme and operational data supplied by Medable, meaning the findings may not apply equally to different therapeutic areas, trial designs or development programmes.

A peer-reviewed publication containing further details of the analysis is expected later this year.

If the findings are supported by further independent research and real-world implementation, AI-based clinical monitoring could become one route for reducing operational costs and shortening timelines in drug development.