Unlocking peak operational performance in clinical development with artificial intelligence – McKinsey

Despite dedicated efforts across the biopharmaceutical industry to streamline clinical development, clinical trials remain complex, costly, and time-consuming. These challenges are compounded by an increasingly competitive and crowded trial landscape. Yet accelerating clinical development remains crucial—not just for patients but for the enterprise. A 12-month reduction in the clinical development timeline can add more than $400 million in net present value (NPV) across a sponsor’s portfolio while delivering immeasurable benefits to patients and their families.
This article is a collaborative effort by Anton Mihic, Chaitanya Adabala Viswa, Gaurav Agrawal, Hann Yew, and Kevin Webster, representing views from McKinsey’s Life Sciences Practice.
Leading biopharma companies are addressing this challenge by adopting AI and machine learning (ML) to accelerate trials through scientific and operational improvements. Our analysis of biopharma data from operational AI/ML pilots indicates that AI/ML can be used to identify optimal trial sites, boost enrollment by 10 to 20 percent, and predict real-time enrollment performance, which allows for earlier, more proactive interventions. On average, applying these AI/ML techniques across assets and therapeutic areas has compressed development timelines by six months per asset, bringing innovative therapies to patients and communities faster. With the advent of generative AI (gen AI), clinical development organizations are building on these gains to further streamline trial processes, enhance stakeholder engagement, and inform decision-making with deeper insights. Even greater cost efficiency and trial speed gains can be achieved by combining classical AI/ML and gen AI techniques together with new ways of working.
Particularly promising is gen AI’s potential to enable “peak performance” in operational aspects of clinical development. For example, gen-AI-powered digitalized processes, such as auto-drafting trial documents, have cut process costs by up to 50 percent. Our studies show a 20 percent increase in NPV from gen-AI-enhanced health authority interactions, driven by improved data quality and signal management. Additionally, gen AI has enabled more than 12 months of trial acceleration through optimized site selection and rapid copiloted decision-making across trial operations.
We’ve identified 12 AI and gen AI use cases across all clinical development functions (Exhibit 1). These use cases can significantly improve the quality, speed, and efficiency of clinical development: improved trial design and planning, enhanced trial and site performance via AI-powered copilots, reimagined clinical data management through automated cleaning and query resolution, and faster “first right time” document generation.
In this article, we present three lighthouse use cases that are gaining traction with leading biopharma companies. These examples showcase the significant improvements in trial efficiency and organizational productivity that are possible with AI and gen AI.
Site selection, often based on geography or target indication, is critical to trial and site performance. Top-enrolling sites typically outperform median sites by two to four times, while 10 to 30 percent of activated sites fail to enroll any patients. Over the past decade, demand for trial participants has grown, along with rising site congestion and staff burden. These challenges have recently been exacerbated by increasing protocol complexity and high turnover among principal investigators (PIs) and clinical staff.
AI-enabled site selection (see Exhibit 1 above) drives superior trial outcomes by identifying and predicting top-enrolling sites. Clinical operations managers can use gen AI to analyze protocols and identify historical trials with similar characteristics, such as their end points and assessments, and by combining extensive trial- and site-level data about such performance metrics as activation timelines, site quality, and PI experience. From those analyses, a customized list of possible trial sites can be generated and ranked by enrollment potential. This approach significantly outperforms traditional methods, which often depend on limited, outdated sponsor or clinical research organization (CRO) data. We observed that AI-driven selection improves the identification of top-enrolling sites by 30 to 50 percent and accelerates enrollment by 10 to 15 percent, or more, across therapeutic areas (Exhibit 2). AI-driven site selection can enhance trial quality by predicting site performance on metrics such as data cleaning, while also increasing trial diversity by factoring in patient demographics.
To manage the deluge of data generated during trials, top organizations often implement a “clinical control tower”—a centralized dashboard that integrates live insights on trial performance, including milestones, site activation and enrollment, sponsor-CRO interactions, clinical drug supply, costs, and resourcing. However, monitoring this data—the job of a clinical trial manager—is challenging. They need to track thousands of trial data points daily and are often overwhelmed by “urgent” alerts, making true prioritization nearly impossible. In addition, control towers typically capture only about 70 to 80 percent of the critical insights trial managers need for active trial oversight.
A gen AI trial performance “copilot” (see Exhibit 1 above) can offer relief by prioritizing critical issues, executing specific actions, such as drafting targeted emails to sites based on historical performance and PI preferences, and enabling deeper custom analytics that traditional control towers are unable to perform (Exhibit 3). Over time, the copilot learns which interventions succeed in various contexts, evolving into an increasingly effective guide that enhances a trial manager’s productivity.
Dossier generation is a central activity in regulatory submissions, under which is the clinical study report (CSR) that summarizes trial results and is often on the critical path to dossier completion. CSRs typically span hundreds of pages and require weeks for medical writers to draft and refine. Gen AI has emerged as a transformative solution, enabling rapid, high-quality, and automated CSR drafting for pharmaceutical companies.
Gen AI can accelerate CSR timelines by 40 percent, cutting the process from eight to 14 weeks to five to eight weeks and increasing NPV per asset by roughly $15 million to $30 million (Exhibit 4). Within minutes, the gen AI tool generates a first draft from the protocol, a statistical analysis plan, and tables, listings, and figures. It can also achieve 98 percent (or more) accuracy with fewer errors than human-written drafts. One company reduced its first-draft timeline from three weeks to three days, halving the touch time from 200 hours to 100 hours. Medical writers have reported that gen AI platforms allow them to focus on clinical insights rather than on performing manual, repetitive tasks.
Building on the success of this lighthouse use case, we see many pharmaceutical companies rapidly expanding their document-generation capabilities across the R&D value chain, often starting with critical path documents such as clinical summaries and quickly moving into additional documents such as informed-consent forms, investigator brochures, and periodic safety update reports.
Scaling the impact of gen AI requires more than just technology—it demands conviction and commitment at all levels, from senior leaders investing in a long-term vision and modeling change to frontline leads adapting to new expectations and workflows. For operational gen AI use cases, the classic influence model has proved highly effective in driving large-scale change management. In a case study, a clinical development organization successfully used the following approach to deploy and adopt an operational trial analytics platform across more than 2,000 users over two years:
By adopting this holistic change-management approach, the organization achieved strong user buy-in, with hundreds of core users actively utilizing the platform weekly for trial decision-making. Users felt supported, energized, and saw the platform as a game changer and significant time-saver. Broad adoption saved tens of thousands of hours across the clinical development process and helped to accelerate cycle times from bottom quartile to top quartile for key assets.
Gen AI is poised to transform clinical trials by making such processes as documentation generation, protocol design, and performance monitoring more time and cost efficient. However, success requires organizational change. Leaders will need to model new behaviors, foster innovation, redesign processes, and develop talent to fully capture gen AI’s potential to enable faster delivery of treatments to patients.
Anton Mihic is a partner in McKinsey’s New Jersey office, Chaitanya Adabala Viswa and Hann Yew are partners in the Boston office, Gaurav Agrawal is a senior partner in the New York office, and Kevin Webster is a partner in the Bay Area office.
The authors wish to thank Leo Potters, Lionel Jin, and Michael Loebl for their contributions to this article.
This article was edited by Jermey Matthews, an editor in the Boston office.

source

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top