Highlights from the 2025 AWS Life Sciences Symposium’s Clinical Trials track

On May 6, more than 1,000 leaders from over 400 life sciences organizations convened in New York City for the seventh annual Amazon Web Services (AWS) Life Sciences Symposium. Centered on the theme “Building for Breakthroughs: AI-powered Innovations Transforming the Pharmaceutical Value Chain,” the event featured 26 sessions with 39 expert speakers, all focused on the practical, real-world application of AI across the pharma discovery-to-delivery lifecycle.

One thing became unmistakably clear: generative AI is no longer a distant promise—it’s already transforming the way the pharmaceutical industry discovers, develops, and delivers new therapies. This impact was particularly evident during the Clinical Trials Breakout Track, where industry leaders showcased how AI is streamlining key stages of the clinical trial process—from designing protocols and selecting sites to enrolling patients and submitting data to regulators.

Clinical Trials at a Crossroads

Pharmaceutical clinical trials are overdue for reinvention. On average, it takes 6–7 years and up to $2.6 billion to bring a new therapy to market. A major bottleneck is patient recruitment, which accounts for nearly a third of both the time and cost—yet 80% of trials fail to meet enrollment targets, and 85% struggle to retain participants. As trials grow more complex and globally dispersed, traditional manual processes are proving unsustainable.

Artificial intelligence offers a powerful path forward. By tapping into organizational data and real-world evidence, AI can revolutionize clinical research—from optimizing trial design and predicting recruitment outcomes to intelligently matching patients and enabling real-time monitoring through automated data capture and anomaly detection. The outcome: faster, more adaptive, and data-driven trials.

The stakes are high—and the moment for action is now. With regulatory bodies like the FDA beginning to issue clearer guidance on AI in clinical development, early adopters have a unique opportunity to shape industry standards and lead the next wave of innovation. Those who act quickly can realize substantial gains in speed, cost efficiency, data quality, and patient inclusion.

That sense of urgency—and optimism—was palpable during the Clinical Trials Breakout Track, where pioneering organizations shared how they’re already using AI to transform clinical research. Here are the key takeaways

AstraZeneca: Accelerating clinical trials with AI Agents

AstraZeneca opened the track by showcasing its Development Assistant. This generative AI-powered agent, built on AWS, simplifies access to clinical data and accelerates decision-making. Vaishali Goyal, Senior Director of Patient Safety, explained how the tool allows clinical operations teams to query structured and unstructured data using natural language, providing real-time, evidence-based insights.

AstraZeneca built the platform on Amazon Bedrock, integrating retrieval-augmented generation (RAG) with text-to-SQL capabilities. This combination rapidly surfaces insights from AstraZeneca’s extensive data landscape. Each response includes traceable source information, reducing manual effort while ensuring transparency and trust, as it helps address critical industry challenges, like patient recruitment and site selection.

The Development Assistant’s effectiveness stems from AstraZeneca’s strong data foundation. The company transforms curated data sources—from Electronic Laboratory Notebooks (ELNs) and Laboratory Information Management Systems (LIMS) to clinical systems—into FAIR (Findable, Accessible, Interoperable, Reusable) data products. These products fuel scalable, multimodal AI applications. By enabling natural language access to this data, AstraZeneca drives greater efficiency and collaboration, allowing its teams to focus on higher-value innovation.

Originally launched as a proof of concept in mid-2024, the Development Assistant reached a production-ready Minimum Viable Product (MVP) in six months, clearing rigorous cybersecurity and AI governance requirements. Its multi-agent architecture supports scalability and agility, designed to handle increasing user demand and data complexity. AstraZeneca plans to scale the platform to over 1,000 users in 2025, expand integration with richer data sources, and enable seamless collaboration across domains. With the Development Assistant, AstraZeneca is laying the foundation for the next generation of AI-powered clinical development.

Watch the session recording | Access the presentation

Faculty: Building early success with AI agents

Following AstraZeneca, Andy Brookes, Chief Technology Officer at Faculty, presented a practical framework for organizations to become successful early adopters of AI agents. He outlined three core pillars for agent effectiveness: role definition, business context grounding, and performance management.

First, defining the agent’s job is essential. Brookes recommends using the Observe, Understand, Decide, Act (OUDA) model to structure agent tasks as decision loops. This approach ensures agents are purpose-built for specific business needs and autonomy levels, rather than acting as generic assistants. By focusing agents on discrete, high-impact decisions, organizations can drive targeted outcomes while minimizing complexity and risk.

Second, embedding agents within a rich business context is crucial. Faculty calls this a “private world model”—a digital simulation of an organization’s internal policies, processes, and constraints. This model enables agents to operate with alignment and realism, supporting safer testing and reducing the risk of misbehavior before full-scale deployment.

Third, performance management must extend beyond speed and accuracy. Brookes emphasizes evaluating agents through both quantitative and qualitative measures, including trustworthiness and interpretability.

Brookes introduced four agent archetypes representing increasing levels of autonomy:

• Scout: Information discovery
• Analyst: Scenario analysis and recommendations
  Operator: Execution with human oversight
• Autopilot: Monitored autonomy within defined boundaries

His recommendation? Adopting agents in phased deployments that deliver immediate value while building toward larger, systemic transformation.

Watch the session recording | Access the presentation

Novartis: Accelerating clinical trials with an adaptive AI strategy

Noah Hoh, Director of Digital Innovation, Strategy, Program & Portfolio Operations at Novartis, shared how the company is strategically integrating AI and data science across its R&D pipeline to accelerate drug development timelines.

Novartis has implemented three core initiatives to reduce 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 6+ months using predictive modeling and AI simulations throughout the R&D lifecycle.

Together, these initiatives will reduce total drug development time by up to 19 months.

Central to this transformation is Novartis’s adaptive AI strategy, which avoids a one-size-fits-all model. Instead, the company employs a targeted AI strategy, matching specific capabilities to each development phase. Key clinical trial use cases include protocol design, site selection, clinical operations optimization, document generation, decision support systems, and more.

The Intelligent Decision System (IDS), built on AWS, exemplifies this approach. IDS uses digital twins to simulate clinical workflows, allowing teams to test strategies and forecast outcomes before implementation. This reduces risk and increases operational efficiency.

Novartis has adopted a phased implementation approach, delivering immediate benefits while building toward comprehensive AI integration across all R&D operations. With a pragmatic, phased approach, Novartis is ensuring its teams see real benefits from AI today—while laying a strong foundation for a fully AI-enabled future. The strategy is clear: deliver measurable gains now, and build toward long-term, systemic transformation.

Watch the session recording | Access the presentation

Merck: Building a modern clinical data platform that scales

In the race to harness generative AI, one truth remains constant: AI is only as effective as the data that powers it. Clinical trials currently face challenges with fragmented systems and siloed data, creating complex architectures that impede innovation and regulatory compliance.

Merck is transforming its clinical trials through a new AWS-based modern clinical data platform powered by Veeva’s Development Data Platform. Merck’s Executive Director, Gopi Gopinath, and AVP, Clinical and RWE IT, Marjorie Waters, hared how this transformation, streamlines clinical operations through a unified, metadata-driven architecture. The platform consolidates fragmented clinical, operational, regulatory, and safety data into a single source of truth, enabling faster trials while maintaining data integrity.

The platform integrates data across internal systems and external partners into one unified repository, and offers key capabilities, like: Study Data Tabulation Model (SDTM) standardization; automated third-party data ingestion; enhanced security features including data masking and role-based access; centralized metadata management; and, self-service analytics tools. With integrated analytics and visualization tools, the platform offers crucial insights for trial planning, patient enrollment, site performance, and risk-based monitoring—reducing operational overhead and enabling proactive decision-making.

The new architecture also establishes the foundation for advanced AI applications, like using generative AI for protocol and report writing, implementing predictive modeling for trial success, developing simulation tools for protocol optimization, and creating synthetic control arms.

By adopting a cloud-native, platform-based approach, Merck is not only modernizing its trial infrastructure, but also building the digital foundation necessary to fully realize the promise of generative AI in clinical development.

Watch the session recording | Access the presentation

Unlocking Real-World Data and Evidence for Faster, Data-Driven Clinical Trial Insight

Real-world data (RWD) drives modern healthcare decisions: 85% of Food and Drug Administration (FDA) approvals from 2019-2021 relied on real-world evidence (RWE). However, this valuable data remains scattered across healthcare providers, insurance companies, and medical registries. Researchers often spend months collecting and organizing data before analysis can begin. Current data silos, privacy requirements, and complex database queries limit the potential of artificial intelligence (AI) tools for using this valuable treasure trove.

AWS’ WorldWide Real World Data Lead, Praveen Haridas, and Datavant’s President and GM, Arnaub Chatterjee, presented a solution to this challenge. Their new platform, Datavant Connect, built on AWS Clean Rooms, enables researchers to analyze linked patient data without exposing protected health information (PHI). The platform reduces the traditional four-month discovery process to two weeks while maintaining data owner control. It includes built-in Health Insurance Portability and Accountability Act (HIPAA) compliance and governance features. All of this means that privacy isn’t traded for speed.

Pharma leaders are already capitalizing on these capabilities. Boehringer Ingelheim’s Executive Director and Global Head of RWE Center of Excellence, Paul Petraro, shared how the organization uses the platform to scale their RWE strategies to better understand treatment effectiveness and economic outcomes, using beta blocker effectiveness assessment as an example.

But that’s not all. To make insights even more accessible, AWS introduced intelligent agents that allow researchers to query complex datasets using natural language, removing the barrier of coding expertise, democratizing access to RWD. These agents provide auditable, role-specific experiences integrated with medical literature, clinical data, and regulatory protocols—all while maintaining strict compliance and data security.

This came to life as Lilly’s Senior Director, RWD, Greg Cunningham, showcased their Real World Data (RWD) Insights Agent, built on AWS, which slashes insight generation from days to minutes. The “virtual analyst” allows non-technical users to query complex datasets using natural language, eliminating the need for SQL expertise. Built on Amazon Bedrock, the system uses multiple AI agents to manage metadata discovery and cohort definitions while maintaining audit trails and compliance. And, it can scale quickly to onboard new data sources and adjust to organizational needs—all within a responsible, human-in-the-loop framework.

The presentation demonstrated AWS’s commitment to transforming healthcare data analysis, enabling faster evidence generation and improved patient outcomes through collaborative innovation with market leaders. Attendees got a glimpse of how we are collaborating with partners like Datavant to make it easy for organizations to discovery, evaluate, and subscribe to RWD from broad data producer network. They also got to a first-hand view into how AWS can help organizations derive insights from subscribed RWD via user centric tools, includes cloud-native, agentic AI applications that allows non-technical users to analyze complex healthcare data via natural language queries.

Watch the session recording | Access the presentation | Read more

Building for tomorrow’s success, today

With approximately 70% of all Life Sciences (LS) investments in generative artificial intelligence (AI) focused on clinical trials, the need for speed, precision, and scalability has never been greater.

As new use cases in clinical trials continue to emerge, one thing remains constant: the need for a scalable, secure data foundation. This foundation enables organizations to adapt, evolve, and keep building—regardless of future technological advancements.

We extend our gratitude to all partners, customers, and attendees for being part of this journey. While we’re proud of our shared accomplishments, we’re even more excited about the future ahead.

Connect with an AWS Representative or explore our AWS Partner network today to unlock the new era of life sciences innovation.

Further reading:

1. 7th Annual AWS Life Sciences Symposium: Keynote highlights | Watch the keynote recap
2. Accelerating Life Sciences Innovation with Agentic AI on AWS
3. Highlights from the 2025 AWS Life Sciences Symposium’s Manufacturing Track
4. Highlights from the 2025 AWS Life Sciences Symposium’s Commercialization Track 
5. Highlights from the 2025 AWS Life Sciences Symposium’s Drug Discovery Track