Moving Beyond Disappointing Pilots: Scaling Adoption of AI-Enablement in Drug Development

Nearly every sponsor and contract research organization (CRO) has now implemented initiatives to explore how AI-enabled solutions can improve productivity, accelerate study timelines, enhance quality, and reduce the cost of drug development. Yet despite substantial investment and widespread enthusiasm, few organizations have successfully scaled AI-enablement beyond isolated pilot projects.

Recent surveys and reports highlight a disconnect between aspiration and adoption and this has left sponsors and CROs struggling to prioritize potential use cases and determine where AI will deliver meaningful business value. MIT's NANDA initiative, for example, drawing on primary and secondary research activity found that the vast majority of AI pilots have underperformed. In a separate study, Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept. Stanford’s Institute for Human-Centered Artificial Intelligence has also documented a widening gap between adoption and impact.

The upside for those that deploy successful use cases is not speculative. In the same Gartner analysis, drawing on a survey of 822 business leaders fielded between September and November 2023, early adopters reported average gains of 15.8% in revenue, 15.2% in cost savings, and 22.6% in productivity (Gartner, July 29, 2024). The value proposition is real and demonstrably attainable. What separates the organizations that capture it is rarely access to better technology; it is the discipline to prioritize, sequence, and govern deployment.

AI-enablement is unlikely to achieve its transformative promise through opportunistic experimentation alone. Instead, organizations require a more disciplined framework for identifying high-value opportunities, and for prioritizing and sequencing investments to maximize both probability of success and long-term impact. Based on our experience informed by emerging best practices in drug development and adjacent R&D intensive industries, we offer insights and considerations to help organizations scale adoption of AI-enablement solutions.

Why many AI pilots may be failing

The wide gap between AI exuberance and measurable business impact reflects a common innovation adoption pattern characterized by Gartner's hype cycle. New technologies experience an initial "Peak of Inflated Expectations," during which organizations expect rapid, transformative benefits. As implementation challenges emerge and early pilots fail to produce anticipated returns, organizations descend into the "Trough of Disillusionment." Only after gaining practical experience do organizations progress toward the "Slope of Enlightenment," where successful use cases become better understood and scalable deployment becomes possible.This adoption pattern for numerous technology innovations has been well documented within drug development operations (e.g., EDC, eCOA, IVRS, DCTs).

The central challenge is not a lack of AI opportunities. Rather, organizations face an overwhelming number of possible applications. Without a disciplined prioritization framework, companies frequently pursue independent and isolated pilots that lack strategic rationale, broad support, or scalable infrastructure. Many organizations begin with technology rather than company-specific business problems. AI becomes the solution in search of an application. Teams become captivated by technical capabilities instead of identifying operational bottlenecks where AI can deliver measurable improvements in quality, speed, cost, or patient outcomes. What works in isolation for a single employee experimenting with tools often fails to scale within a function, let alone a complex biopharma R&D organization. As a result, many pilots demonstrate interesting functionality without addressing problems sufficiently important to justify enterprise deployment.

Organizations have become increasingly sensitive to workforce implications. Although AI promises substantial productivity improvements, senior management recognizes that successful transformation requires thoughtful workforce planning rather than automation followed by consolidation. But most organizations also admit they are unclear how to best enable personnel to perform higher-value activities. And some organizations have noted that many AI pilots are not clearly framed and lack well-defined outcome measurements.

And most organizations recognize that successful AI implementation requires maintaining "humans in the loop." Conceptually, organizations widely agree that human oversight is essential for validating AI-generated outputs, ensuring appropriate clinical judgment, identifying bias, and maintaining accountability for critical decisions. But most organizations concede that personnel don’t actually know ‘what to do in the loop.’ Effective organizations recognize the critical need to carefully define roles and practices for personnel in the loop including decision rights, review thresholds and escalation criteria.These needs become increasingly important as AI technology applications move toward AI management agents overseeing AI operations agents.

Increasing the likelihood of piloting and scaling AI-enablement opportunities

Our interactions and interviews with pharmaceutical and CRO company management reveal five primary approaches that help facilitate successful pilot experiences and the transition to organization-wide adoption:

1. Assess and shore-up organizational capability and readiness
2. Establish clear governance, trust and measures of success at the outset
3. Ensure data readiness and prepare for an agentic future
4. Develop and articulate an organization-wide AI-enablement strategy
5. Select initial pilot projects in areas where your organization is already effective

Organizational capability and readiness

Organizational capability to reimagine current and target state business processes, execute on those concepts, and incorporate a rapid learning feedback loop is paramount regardless of the transformative nature of the AI capability being considered. In addition, individuals and teams have valid concerns that may include—among other issues—quality and accuracy of AI outputs, data integrity and security, regulatory implications, and workforce impacts.AI-enabled solutions will be effective only if they are adopted broadly by the organization. Understanding and addressing employee concerns and existing barriers to adoption up front facilitates trust in the adoption process and promotes collective support for scaling initiatives.

Governance, trust, and measures of success

AI governance and trust principles including data governance, privacy, cybersecurity, and regulatory obligations are vital to successful adoption and scale-up of AI-enabled solutions, and should be established at the outset. Regulatory agencies including the FDA and EMA are encouraging AI use in drug development. The FDA's January 2025 draft guidance sets out a risk-based credibility assessment framework tied to a model's context of use, and the EU AI Act imposes obligations on sponsors operating in Europe regardless of where their models are built. Despite regulatory encouragement, many organizations have applied governance principles on a function-by-function basis instead of holistically. That’s fragmented approach limits organizational learning, duplicates effort, and creates a false sense of security where unmitigated risks remain.Leadership that addresses these uncertainties before asking teams to implement new AI-enablement solutions will facilitate a more favorable environment for successful AI adoption.

In addition, establishing clear business objectives with measurable expectations and outcomes is essential for teams to rapidly learn and adapt before abandoning what may initially appear as failed efforts. AI presents a unique opportunity to rapidly fail and recover with invaluable learnings in a uniquely rapid cycle time. The role of senior leadership and project sponsorship is more time-consuming and challenging for these initiatives: arm's-length oversight does not keep pace with this rapidly advancing technology. Hands-on, immersive understanding of AI applications is crucial as senior leaders are best positioned to apply visionary thinking and assessment of potential—even when an early solution may be underperforming.

Data readiness

Many sponsors and CROs underestimate the foundational importance of data readiness. AI models are only as effective as the quality, completeness, consistency, and accessibility of the data used to train and operate them. Clinical development environments typically contain fragmented and siloed data distributed across a variety of disparate systems (e.g., electronic data capture, clinical trial management systems, electronic trial master files, safety databases, laboratory platforms, imaging repositories, electronic health records, wearable devices, and external vendor platforms). Inconsistent standards, limited interoperability, and incomplete integration substantially constrain AI-enabled solution development and performance.Data readiness should be scored and weighed alongside business value when selecting and sequencing projects, rather than treated as a gate that must be fully cleared before work begins. The distinction matters: waiting for complete data defers organizational learning indefinitely, while sequencing deliberately against known data constraints does not. Furthermore, agentic AI changes what "ready" means. Data foundations built for human consumption—curated for people who read, interpret, and reconcile as they go—are inadequate for agentic consumption, where systems must access and correlate data across sources at speed with no human intermediary to resolve gaps. In some cases, data foundations must be rebuilt from the ground up. This is a substantial effort and investment, and it’s important for organizations to keep data foundations grounded in the clinical development domain, tempering enterprise-wide efforts that may take multiple years to complete.

Organization-wide AI strategy

Organizations often prioritize incremental automation rather than transformational change. Many pilots focus on automating relatively small administrative tasks that generate modest efficiency improvements but limited value, and leadership fails to communicate how these initial projects fit into a broader AI-enablement strategy that will drive significant impact. There is value in these efforts, where they build organizational capability and momentum, but they are a bridge and not the destination. Selecting initial AI pilot projects that fit into a comprehensive AI strategy and communicating internally how successful implementation provides the foundation for meaningful business and patient impact will drive organizational buy-in and increase likelihood of success.

Initial pilot selection

Organizations often choose to pilot and deploy AI-enablement solutions in areas where they are struggling. This approach, however, is more likely to lead to failure as the organization attempts to integrate powerful new technology with all its uncertainties into an existing area of weakness.Instead, organizations should first pilot AI-enablement solutions in areas where the organization excels. These are areas where teams best understand success drivers and constraints and can more readily establish and measure KPIs. These areas often have the organization’s top talent with strong leadership and a growth mindset. Teams can focus on addressing known bottlenecks and develop expertise in the best applications of AI solutions—knowledge that can then be shared across the organization. Piloting within areas of strength to build capabilities and credible KPIs facilitates transformational redesign. Strength-based pilot selection is a viable organizational learning strategy that ultimately informs addressing areas of weakness and drives more rapid adoption.

Understanding and addressing these challenges as part of an AI-enablement strategy can transform the success of pilot initiatives and the value and insights that they generate. The issue is seldom the underlying AI solution itself.Rather, it reflects deficiencies in conditions that would most facilitate success: company-specific prioritization, organizational readiness, data infrastructure, and coordinated execution.

Achieving clarity and, ultimately, scale

Successful AI adoption begins with clarity rather than additional experimentation. Leading companies distinguish themselves by carefully selecting and committing to investment areas where both organizational readiness and business value are greatest. They incentivize their teams to challenge the norm, invest in upskilling themselves, and provide the tools and executive support to pursue transformative outcomes.

Fundamentally, organizations must develop and articulate a comprehensive AI-enablement strategy. Initial projects should be selected in areas where the organization is knowledgeable and high-performing and supported by rich, integrated, and high-quality data. Availability of existing metrics with historical readouts and project-appropriate data readiness should therefore become explicit criteria when prioritizing AI investments.

Successful AI implementations extend beyond technical capability and leverage key components of culture and change enablement: executive sponsorship, regulatory alignment, workforce retraining, governance maturity, and process standardization. Even highly promising AI applications may struggle if organizations are unprepared to modify workflows or integrate new decision-support tools into routine operations.

AI-enablement strategy should provide a roadmap for transformational opportunities rather than isolated efficiency gains. Incremental automation offers some value, but the greatest return on investment often arises from redesigning entire business processes rather than optimizing individual tasks. Integrating AI across protocol design, document authoring, site selection, enrollment forecasting, patient recruitment, and operational monitoring, for example, may generate substantially greater improvements than independently automating each activity. But sequencing initiatives to generate early successes and build transferable organizational expertise is critical. Once this is achieved, organizations can concentrate resources on relatively few high-value initiatives capable of achieving enterprise scale. AI-enablement decisions also need to quickly move beyond the integration of discrete AI-enablement solutions to focus on multi-agentic systems that orchestrate across platforms.

Tiered success metrics should be predefined and tied directly to business outcomes: reductions in cycle time, improvements in enrollment performance, decreases in protocol deviations, enhanced quality, improved regulatory compliance, lower operating costs, or accelerated decision making. Measuring technical performance alone provides insufficient evidence of business value.

And organizations should recognize that successful AI transformation represents an ongoing and dynamic organizational capability. Sustainable competitive advantage will increasingly depend upon building integrated data ecosystems, developing AI-literate workforces, strengthening governance, continuously refining models, and embedding AI into routine operational decision making. These organizational capabilities become strategic assets that compound over time.

Fit-for-purpose scale

AI-enabled solutions possess enormous potential to transform drug development by improving productivity, accelerating clinical research, enhancing decision quality, and ultimately bringing innovative therapies to patients faster. Realizing this potential, however, requires moving beyond isolated pilots toward disciplined fit-for-purpose R&D transformation.

Piloting a wide variety of AI-enabled solutions alone is not the most direct path to success. Indeed, most organizations appear trapped within an expanding collection of disparate pilot initiatives that consume resources but fail to deliver outcomes that matter. Organizations will deliver value to their business and to patients by identifying and prioritizing high-value, company-specific business problems best addressed by AI-enablement that are supported by a well-articulated strategy, strong data foundations, mature governance, cross-functional alignment and organizational readiness.

The industry has been through this before. Electronic data capture, eCOA, and decentralized methods each spent years in the trough before becoming unremarkable infrastructure, and the successful organizations that emerged were not those that ran the most pilots. They were the ones that decided earliest what they were trying to change and had the leadership and resilience to stay the course. AI will reward the same discipline, on a shorter clock and with less tolerance for indecision. The question facing every sponsor and CRO is not whether to invest. It is whether leadership prioritize and articulate the three business problems AI is meant to solve over the next 24 months, support the rapid buildout of the necessary organizational capabilities, and have the confidence in the ability of their teams to deliver.

Gillian Bowden, and Sean Connolly; both with Evinova AG

Marc Blaustein, and Ken Getz; both with Tufts Center for the Study of Drug Development, Tufts University School of Medicine

Sources

• Challapally A, et al. The GenAI Divide: State of AI in Business 2025. Massachusetts Institute of Technology, Project NANDA; 2025.
• Gartner, Inc. Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept by End of 2025. Press release, July 29, 2024.
• Stanford Institute for Human-Centered Artificial Intelligence. The 2026 AI Index Report. Stanford University; 2026.
• Beede E, Baylor EE, Hersch F, et al. A Human-Centered Evaluation of a Deep Learning System Deployed in Clinics for the Detection of Diabetic Retinopathy. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems; 2020.
• U.S. Food and Drug Administration. Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products. Draft guidance; January 2025.
• European Parliament and Council of the European Union. Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union; 2024.