On May 6, over 1,000 leaders representing more than 400 organizations in the life sciences sector gathered in New York City for the seventh annual Amazon Web Services (AWS) Life Sciences Symposium. The event, themed “Building for Breakthroughs: AI-powered Innovations Transforming the Pharmaceutical Value Chain,” featured 26 sessions led by 39 experts, emphasizing the practical application of artificial intelligence (AI) throughout the pharmaceutical discovery-to-delivery lifecycle.
A prominent takeaway was the realization that generative AI is no longer a distant vision; it is actively reshaping how the pharmaceutical industry discovers, develops, and delivers new therapies. This transformation was particularly highlighted during the Clinical Trials Breakout Track, where industry leaders presented how AI is optimizing critical aspects of clinical trials, including protocol design, site selection, patient enrollment, and data submission to regulatory authorities.
Clinical Trials in Need of Reinvention
Pharmaceutical clinical trials are ripe for innovation. Typically, it takes 6–7 years and can cost up to $2.6 billion to launch a new therapy. A significant bottleneck lies in patient recruitment, which consumes nearly a third of both time and expense. Alarmingly, 80% of trials fail to meet enrollment targets, while 85% struggle with participant retention. As trials become increasingly complex and globally distributed, traditional manual methods are proving inadequate.
Artificial intelligence presents a compelling solution. By leveraging organizational data and real-world evidence, AI can transform clinical research—from enhancing trial design and predicting recruitment outcomes to enabling efficient patient matching and real-time monitoring through automated data capture and anomaly detection. The result is a faster, more adaptable, and data-driven trial process.
The urgency for action is evident, especially as regulatory bodies like the FDA begin clarifying guidance on AI in clinical development. Early adopters have a unique chance to influence industry standards and spearhead the next wave of innovation. Those who act promptly can achieve significant improvements in speed, cost efficiency, data quality, and patient inclusion.
This sense of urgency and optimism was apparent during the Clinical Trials Breakout Track, where pioneering organizations shared their experiences with AI in clinical research. Here are the main highlights:
AstraZeneca: Streamlining Clinical Trials with AI Agents
AstraZeneca kicked off the track by presenting its Development Assistant, a generative AI-powered tool developed on AWS that facilitates access to clinical data and enhances decision-making. Sarah Mitchell, Senior Director of Patient Safety, elaborated on how this tool allows clinical operations teams to query both structured and unstructured data using natural language for real-time, evidence-based insights.
The platform, built on Amazon Bedrock, integrates retrieval-augmented generation (RAG) with text-to-SQL capabilities, rapidly surfacing insights from AstraZeneca’s extensive data landscape. Each response includes traceable source information, minimizing manual effort while ensuring transparency—addressing critical industry challenges such as patient recruitment and site selection.
The Development Assistant’s effectiveness stems from AstraZeneca’s robust data infrastructure, which transforms curated data sources—from Electronic Laboratory Notebooks (ELNs) to Laboratory Information Management Systems (LIMS)—into FAIR (Findable, Accessible, Interoperable, Reusable) data products that support scalable, multimodal AI applications. By enabling natural language access to this data, AstraZeneca enhances efficiency and collaboration, allowing teams to focus on higher-value innovations.
Originally launched as a proof of concept in mid-2024, the Development Assistant reached a production-ready Minimum Viable Product (MVP) within six months, adhering to stringent cybersecurity and AI governance requirements. Its multi-agent architecture is designed for scalability and agility, prepared to accommodate rising user demand and data complexity. AstraZeneca aims to expand the platform to over 1,000 users in 2025, integrate more diverse data sources, and foster seamless collaboration across domains. With the Development Assistant, AstraZeneca is pioneering the future of AI-driven clinical development.
For further insights, check out this excellent resource.
Faculty: Achieving Early Success with AI Agents
Following AstraZeneca, Jason Lee, Chief Technology Officer at Faculty, offered a practical framework for organizations aiming to become successful early adopters of AI agents. He outlined three essential pillars for agent effectiveness: role definition, grounding in business context, and performance management.
Firstly, defining the agent’s role is vital. Lee suggests using the Observe, Understand, Decide, Act (OUDA) model to structure agent tasks as decision loops. This ensures that agents are tailored for specific business needs, rather than serving as generic assistants. By focusing agents on discrete, high-impact decisions, organizations can achieve targeted outcomes while minimizing risk and complexity.
Secondly, embedding agents within a rich business context is crucial. Faculty refers to this as a “private world model”—a digital simulation of an organization’s internal policies, processes, and constraints. This model allows agents to operate with alignment and realism, facilitating safer testing and lowering the risk of misbehavior prior to full-scale deployment.
Lastly, performance management should encompass more than just speed and accuracy. Lee emphasizes the need for both quantitative and qualitative evaluations, including trustworthiness and interpretability.
He introduced four agent archetypes, reflecting increasing levels of autonomy:
- Scout: Information discovery
- Analyst: Scenario analysis and recommendations
- Operator: Execution with human oversight
- Autopilot: Monitored autonomy within defined boundaries
His advice? Implement agents in phased deployments that deliver immediate value while paving the way for larger systemic transformations.
Novartis: Enhancing Clinical Trials with an Adaptive AI Strategy
Noah Green, Director of Digital Innovation, Strategy, Program & Portfolio Operations at Novartis, discussed the company’s strategic integration of AI and data science throughout its R&D pipeline to expedite drug development timelines.
Novartis has initiated three key programs to shorten development times:
- Fast-to-IND: Reduces Investigational New Drug submission time by 12 months across therapeutic areas;
- Enhanced Operations: Saves 1-2 years through improved efficiency and innovative trial designs;
- AI-Enabled R&D: Cuts cycle times by over 6 months using predictive modeling and AI simulations throughout the R&D lifecycle.
Collectively, these initiatives aim to decrease overall drug development time by as much as 19 months.
Central to this transformation is Novartis’s adaptive AI strategy, which avoids a one-size-fits-all approach. Instead, the company employs targeted AI capabilities tailored to each phase of development. Key clinical trial applications include protocol design, site selection, clinical operations optimization, document generation, decision support systems, and more.
For a deeper dive into this topic, you can read another insightful blog post here or explore insights from experts at Chvnci.
SEO Metadata:
Keywords: AWS Life Sciences Symposium, Clinical Trials, AI in Pharmaceuticals, AstraZeneca, Faculty, Novartis, AI Agents, Drug Development
Leave a Reply