AstraZeneca cuts protocol authoring time 85% with Modella AI multimodal models in oncology R&D

Traditional oncology drug development faces prohibitively high costs, long timelines, and high late-stage clinical trial failure rates. AstraZeneca needed to accelerate biomarker discovery, streamline complex multi-modal data workflows across pathology images, genomic data, and clinical records, and reduce document authoring overhead across its oncology pipeline.

AstraZeneca acquired Modella AI in late 2025 to integrate multimodal foundation models trained on diverse biomedical datasets directly into R&D operations. An 'intelligent protocol' system powered by generative AI was deployed for clinical document authoring, and AI-assisted 3D location detection was added to CT scan workflows to reduce radiologists' manual annotation burden.

Document authoring time was reduced by up to 85%. AI-assisted CT annotation cut radiologists' manual annotation time by significant margins. Oncology segment revenue reached $11.954B in H1 2025, a 16% year-on-year increase, with the AI-driven strategy expected to further reduce attrition rates in clinical trials and accelerate time-to-market.

• Acquiring an AI firm rather than licensing technology can enable deeper integration of multimodal models into proprietary R&D datasets and workflows.
• Generative AI for clinical protocol authoring offers among the fastest measurable ROI in pharma R&D, with up to 85% time savings on document-heavy tasks.
• Combining pathology image analysis, genomic data, and clinical records in a single multimodal model addresses a critical bottleneck in biomarker discovery.