Pharma Embraces AI, But Only 24% Expect Better Drug Outcomes

Pharmaceutical companies have rapidly deployed artificial intelligence across drug research and development, but industry executives remain deeply skeptical about whether the technology will actually produce better medicines, according to a new Citi Research survey.

The investment bank polled 50 U.S.-based pharmaceutical executives responsible for technology strategy in research and development during June 2026. The findings reveal a striking disconnect between adoption rates and outcome expectations.

Widespread deployment, limited confidence

Seventy-two percent of respondents reported their companies were either scaling or had fully scaled AI across R&D operations. Not a single company surveyed indicated they had no plans to adopt the technology, underscoring AI's status as table stakes in pharmaceutical research.

Yet only 24 percent of these same executives expect AI to deliver significant improvements in drug success probability—the critical metric that determines whether compounds advance through clinical trials and reach patients. A majority, 56 percent, anticipate only moderate gains rather than transformative change.

The translation problem

John Yung, head of Asia healthcare research at Citi, identified what he considers the central challenge facing AI-powered drug discovery. The risk isn't that AI fails to accelerate the discovery process itself, he noted, but that speed gains won't convert into higher-quality drug candidates or improved clinical success rates.

"It's only through that translation that companies can turn AI-driven speed into profit, and patients into real beneficiaries," Yung said.

This concern reflects a fundamental tension in pharmaceutical AI applications. While machine learning can rapidly screen molecular compounds and predict binding affinities, the biological complexity of human disease and the unpredictability of clinical trials remain formidable obstacles.

Human bottlenecks persist

Thirty-four percent of survey respondents pointed to enduring constraints that AI cannot easily overcome: human biology, clinical judgment, and execution capabilities. These factors continue to limit drug development regardless of computational advances.

The pharmaceutical industry could potentially reduce drug discovery costs by $26 billion globally through AI deployment, according to estimates referenced in the survey. Yet realizing those savings depends on whether accelerated timelines actually yield approved therapies—a proposition that remains unproven at scale.

Why it matters

This survey captures a critical moment in pharmaceutical innovation. Companies are committing substantial resources to AI infrastructure based on the promise of faster, cheaper drug discovery. If those investments fail to improve clinical success rates—which currently hover around 10 percent for most therapeutic areas—the technology may accelerate failure rather than success. For patients, the distinction between discovering more drug candidates quickly and discovering better candidates that actually work represents the difference between hype and genuine medical progress.

The findings were first reported by Citi Research in a survey released Tuesday.