ICON on the real-world integration of AI in drug R&D

DDW’s Megan Thomas speaks with ICON's IT Director, Marco Heldoorn, about leveraging artificial intelligence (AI) to enhance efficiency, decision-making, and patient outcomes across the drug development lifecycle, from early feasibility to post-marketing. MT: Could you provide an overview of your findings when it comes to drug development and AI? TO: Our most recent survey data shows a cautious but steady integration of AI across drug development, with industry attitudes remaining broadly optimistic though cautious. Since 2019, sponsor confidence in AI’s ability to improve R&D productivity has increased by 5%, reflecting growing recognition of its strategic value. Now, nearly half expect a moderate improvement and another third expect higher improvements (51% to over 100%) because of AI. While enthusiasm is high, adoption is more measured than once expected due to ongoing concerns around data governance, regulatory clarity and technological interoperability. That said, we see a clear shift from experimental use and early-stage pilot programs to more practical integration of AI-driven tools in routine workflows across the development cycle. Sponsors are increasingly investing in AI tools that deliver strong returns, particularly in data analysis, where AI enables unprecedented speed and depth of insight. In fact, 88% of survey respondents expect their organisation to increase investment in digital-enabled technologies in the next one or two years. We see the industry moving toward a model where AI complements human expertise, embedded within routine workflows to enhance trial efficiency, accuracy and strategic decision-making. MT: Where do you see the greatest opportunity for AI to impact early-stage drug development? TO: AI is already reducing personnel burden through intelligent automation and by ingesting, interrogating and generating insights from vast datasets much faster than human experts can. This frees up expert time for strategic tasks and improves early decision-making. Beyond this, the most immediate opportunity lies in the feasibility, study startup and site activation stage. It’s a vital step, and a risky one, as it sets the trial up for success. Nearly half of industry leaders (47%) cite patient identification and site selection as the area where AI and big data can most positively impact speed, efficiency and productivity. AI solutions like ICON’s One Search are already demonstrating value by analysing vast datasets to match trial protocols with high performing sites that have both access to the right patient populations and a strong track record of recruitment. AI also reduces timelines for identifying key opinion leaders in rare disease from months to seconds, providing key insights for feasibility and a quicker path of study design. Faster site identification significantly reduces site startup timelines, lowers risk and supports more inclusive, data-rich studies. Beyond recruitment, AI offers tremendous potential across clinical phases through postmarketing. As integration becomes more seamless, AI will continue to optimise critical early-stage activities, laying a stronger foundation for later-phase success. MT: How is AI contributing to target identification or lead optimisation? TO: AI has exciting applications in small-molecule drug discovery. It’s being used to analyse massive biological datasets to uncover novel targets and predict lead compound behaviour more efficiently. It’s reducing time-to-candidate selection by 20-30% in some cases. With the availability of the right data, these tools help identify high-probability targets and optimise molecular properties before even entering the lab. MT: Where is AI making the most impact on the industry globally? TO: AI is reshaping global clinical development by delivering measurable ROI, particularly in enhancing efficiency and reducing risk. In our survey, 82% of biopharma leaders said digital tools like AI improve return on R&D investments, while 76% cited better product safety and efficacy, and 76% noted greater competitiveness in targeted medicine markets. These gains are most evident in trial planning, patient recruitment and adaptive design where AI can support faster, smarter decision-making. At ICON, we use integrated AI solutions across the development cycle to optimise startup, improve recruitment and retention, accurately predict postmarketing requirements as early as Phase I and accelerate insights. While confidence is growing in AI, its main application is still in driving efficiencies wherever it’s implemented. Efficiency saves time and resources and helps reduce risk in an otherwise risky, expensive, and time-consuming clinical development process. AI is making it easier to get therapies to patients faster. MT: Are there specific therapeutic areas or modalities where ICON believes AI adds the most value? TO: ICON sees significant value in AI across several therapeutic areas, particularly in medical imaging. While human expertise remains essential, AI tools help reduce workload, improve decision-making speed, and enhance data quality. In oncology, cardiovascular, metabolic (including MASH), and musculoskeletal trials, AI supports early cancer detection, treatment response prediction, prognostic biomarker identification, radiomics assessment, and tissue or lesion segmentation. In cardiac studies, AI assesses cardiac function and flow, plaque burden, and myocardial viability. It also aids obesity trials by estimating visceral and subcutaneous fat, and in musculoskeletal trials by assessing bone age, organ volume, scoliosis and fracture risk. At ICON, the Deep Learning (DL) subset of AI is frequently used to enhance image segmentation efficiency. Manual segmentation of organs like the liver, spleen, and various fat and muscle tissues can be time consuming. Manually segmenting the liver, for example, can take over an hour while using DL can this to 10 minutes or less. Similarly, DL assists in automating multiple sclerosis lesion segmentation from MRI scans, cutting annotation time from 45 minutes to around 10. MT: Looking ahead, what kind of infrastructure do you think is essential to fully realise AI’s promise for drug development? TO: Strategic investment in digital maturity and building organisational readiness is key to fully realising AI’s potential in drug development, and leveraging partnerships will be important to that success. Currently, only 13% of respondents in our survey had a? comprehensive and fully implemented AI program. Embedding AI across all stages of clinical development will seamlessly support data-driven decisions and enhance trial efficiency. As AI solutions mature, implementation within a more comprehensive digital ecosystem will enable interoperability across systems, stakeholders and solutions throughout development. We have invested in a connected digital environment to enhance these efficiencies, improve patient journeys and bring therapies to market faster while reducing risk and generating deeper insight for sponsors. Biography Marco Heldoorn, IT Director at ICON, has 15+ years in the CRO industry. He bridges business and tech, driving digital transformation in healthcare. With strategic insight and a hands-on approach, he partners with global labs and clinics to optimise operations and deliver innovative IT solutions.