In 2025, the pharmaceutical industry stands at a crossroads. The adoption of artificial intelligence (AI) has accelerated, yet its impact on R&D productivity remains underwhelming. Despite the availability of advanced models and digital tools, clinical development still spans eight to 10 years from identification to submission and consumes the bulk of R&D budgets. Why hasn’t AI delivered the transformative gains we expected?

The answer lies in how AI has been deployed: fragmented, tactical, and disconnected from the broader operational realities of pharma.

The illusion of progress

Many organizations have launched hundreds of AI use cases, yet only a small fraction yield measurable benefits. Tools designed to enhance productivity often remain siloed, applied to individual tasks without rethinking the entire workflow. The result is a proliferation of pilots—what I call “pilotitis”—with little to show in terms of scaled, real-world impact.

Site selection has been touted as a top AI use case for pharma. Models can identify high-performing sites based on historical and external data, and they do so effectively. Yet the percentage of low-performing sites remains high. Why?

Because the process hasn’t changed. Feasibility teams still rely on manual data collection via email and Excel, repeating the same 12-week cycle for every study. Even when models suggest optimal sites, local teams override them without feeding back the rationale—data that could be used to train models and improve future recommendations. The result: a 26% performance boost from model-selected sites, diluted by manual overrides and process inertia.

To truly accelerate patient recruitment and improve site performance, we must digitize feasibility, automate data collection, and redesign workflows. Novel patient recruitment methods powered by agentic AI are being explored to transform low-performing sites into high-performing ones. With only 5%–8% of eligible patients currently entering clinical trials, the opportunity for impact is enormous.

Beyond protocol AI

Clinical document authoring is another area ripe for transformation. Protocol development typically takes three to four months, involves 20 to 30 teams, and connects to over 200 documents across clinical development. Accelerating protocol creation with AI is helpful but is just scratching the surface of true transformation. The real opportunity lies in reimagining document authoring across the entire program. Outsourcing agreements must be renegotiated to reflect AI-driven efficiencies. Documents must be linked, changes tracked, and systems automated. Without this, AI-generated versions remain disconnected, and the broader process remains unchanged.

The goal should be to author documents twice as fast across the program, not just the protocol. This requires end-to-end process redesign, revised SOPs, digital solutions for document storage and linkage, and flexible AI models.

A blueprint for AI transformation

To move from pilotitis to productivity, organizations must:

• Set clear goals. Define clear, measurable outcomes, such as doubling productivity or halving timelines. Track progress and reward teams based on real impact.
• Redesign processes end-to-end. Reimagine workflows with AI and digital at the core. Build change management into your strategy from day one to prepare to scale.
• Foster cross-functional collaboration. Avoid siloed AI solutions that may work in isolation but create more complexity when integrated into workflows.
• Align incentives. Promote digital and AI acceleration as criteria for career advancement, similar to successful practices in large tech companies.
• Choose the right digital partners. Work with partners who understand the complex pharma operating realities to integrate diverse AI models effectively.
• Focus on fewer, bigger bets. Prioritize initiatives with the highest potential for R&D productivity. Don’t be afraid to cull projects that aren’t delivering value or can’t scale.
• Leverage human expertise: Use expert judgment to guide AI priorities—site selection may be popular and seem like an easy win, but it may not necessarily be the most impactful for your organization.

Pharma is on the cusp of a digital revolution. But to unlock AI’s full potential, we must shift from isolated use cases to holistic transformation. By setting bold goals, redesigning processes, and aligning incentives, we can finally realize the productivity gains that AI has long promised and usher in a new era of innovation in pharmaceutical R&D.

Cristina Duran is President & CEO, Evinova