Using Amazon Bedrock Agents to Accelerate Decisions Across Drug Development at AstraZeneca

Drug discovery and development is a long and complex process for the biopharmaceutical industry, with clinical trials serving as a critical stage for demonstrating the safety and efficacy of new therapies—especially now, as medicines become more complex, personalized, and driven by new therapeutic modalities.

However, the global nature of clinical trial programs for a company like AstraZeneca can result in disparate systems, creating bottlenecks in analysis and decision-making. Clinical teams spent hours manually compiling answers to questions like “Where are the highest-performing trial sites?” In an environment where speed is essential, faster access to clinical insights became urgent.

To overcome these challenges, AstraZeneca collaborated with the AWS team to build AI agents for different research and development (R&D) functions, including clinical, regulatory, patient safety, and quality. The teams developed this solution using Amazon Bedrock Agents, which uses multi-agent architecture to automate multistep tasks by seamlessly connecting with company systems, APIs, and data sources. The resulting application—Development Assistant—combines text-to-SQL generation with retrieval-augmented generation, unifying both structured and unstructured data sources to generate insights.

When users ask natural language questions like “What are the top five countries with the most clinical trial sites?” Development Assistant interprets the query, augments it with domain-specific terminology, generates SQL code, runs it across AstraZeneca’s data sources, and returns insights within seconds. The system also provides full transparency, showing exactly which data tables were accessed and how results were generated, helping users across clinical development and operations make informed, trusted decisions at speed.

A core driver to Development Assistant’s success is AstraZeneca’s Drug Development Data Platform (3DP). The 3DP ingests data from a variety of source systems—such as clinical, regulatory, quality, and safety IT product platforms—and aligns it with common vocabularies to produce findable, accessible, interoperable, and reusable data products. These standardized, connected datasets serve as the backbone for the Development Assistant. Built on this foundation, Development Assistant unlocks seamless access to AstraZeneca’s complex data through natural language queries, driving significant cost savings while accelerating insights and letting data scientists focus on high-value analysis rather than data preparation. Enhanced with controlled vocabularies, the application delivers accurate, context-aware responses tailored to AstraZeneca’s R&D environment. Clinical teams can ask questions and receive actionable answers with transparent source references.

AstraZeneca recognized that relying on a single AI agent would create performance bottlenecks and less accurate responses, especially for complex queries. To address this, the team implemented a multi-agent architecture using Amazon Bedrock Agents, introducing a supervisor agent that routes natural language queries to the appropriate specialized subagents. For example, this can include a terminology agent for decoding pharmaceutical acronyms, a clinical agent for trial-related data, a regulatory agent for compliance queries, and a database agent for technical operations.

By deploying domain-specific agents, AstraZeneca has tailored the system to deliver more precise, context-aware insights while maintaining performance at scale. This modular approach facilitates comprehensive, trusted answers and lets the platform expand and adapt as use cases grow across the R&D value chain. With built-in guardrails and a rigorous, production-ready setup, the multi-agent system not only improves reliability and transparency but also helps Development Assistant scale safely and effectively to meet evolving scientific needs. “Amazon Bedrock provided a multi-agent architecture that essentially gives us the flexibility and scalability we were lacking before,” says Vaishali Goyal, senior director of R&D IT at AstraZeneca. “Now, when a clinical user submits a prompt to Development Assistant, the supervisor agent guides it to the right expert agent for the most relevant and accurate response. Insights that once took hours are now available in minutes.”

Working alongside the AWS team, AstraZeneca moved Development Assistant from concept to production in only 6 months, including cybersecurity and AI governance checks, and is scaling the solution to over 1,000 users. “Development Assistant is breaking down traditional silos between clinical, regulatory, and safety domains and delivering comprehensive cross-functional analysis that was previously time intensive or impossible to achieve,” says Goyal.

Next, AstraZeneca plans to expand Development Assistant to other research domains beyond clinical trials, creating seamless interoperability across all R&D functions, from early discovery through regulatory approval. “We have the opportunity to accelerate our medicines to patients much quicker,” says Cassie Gregson, global vice president of R&D IT at AstraZeneca. “Our collaboration with the AWS team is key to continuing our journey of simplifying, automating, and transforming the way we work across the development pipeline for our people, clinical trial teams, and patients. For patients, this innovation translates to something even more powerful—the potential to bring life-changing medicines to market faster.”