How EY Uses Elasticsearch to Power RAG for Finance Clients
EY, one of the world’s largest professional services networks, built a generative AI platform for financial institutions using Elasticsearch’s Relevance Engine (ESRE) at the core. The solution enables banks to extract structured insights from ESG reports, financial statements, and compliance documents using retrieval-augmented generation. It achieved 10–15% accuracy gains over baseline document extraction and ran 3x faster than standard RAG implementations.
Impact
10–15%
Accuracy improvement in document extraction
3x
Speed improvement over Native RAG
Challenge
EY’s financial services clients needed to extract structured insights from large volumes of unstructured documents—ESG reports, financial statements, compliance filings—but manual analysis was slow and existing RAG approaches lacked the speed and accuracy required for production deployment.
Solution
EY built a generative AI platform on Elasticsearch’s ESRE, using vector embeddings for large-scale document retrieval, enhanced chunking and indexing for accuracy, and integration with LlamaIndex and LangChain for end-to-end RAG pipeline orchestration.
Tools & Technologies
What Leaders Say
“Elastic’s cutting-edge work in search and retrieval attracted us. We integrated these capabilities into our AI stack for Retrieval Augmented Generation (RAG), which improves accuracy and accelerates retrieval of insights from unstructured data.”
“Imagine extracting 14 key variables from a 40-page PDF or comparing information across multiple reports from different years. This is where we save clients significant time and resources.”
“With Elastic, we can simultaneously promote responsible AI and innovation. It means our clients can adhere to regulations, act faster, and mitigate internal risks.”
Sign up to read complete case studies, access detailed metrics, and unlock all use cases.
Full Story
Banks face a growing burden of unstructured data. Regulatory filings, ESG commitments, capital adequacy reports, and multi-year financial statements contain critical information that drives compliance decisions—yet most of it cannot be queried at scale. EY, which provides advisory and technology services to financial institutions globally, saw this as an opportunity to build a production-grade generative AI platform that could give banking clients an analytical edge.
Before EY’s solution, financial teams were spending significant time manually reviewing dense documents to extract key variables. Extracting 14 data points from a 40-page PDF, or comparing figures across reports from different years, required analyst hours that added up quickly. Vendor integrations were slow to ramp up, and homegrown AI components introduced maintenance risk.
EY selected Elasticsearch as the foundation for its RAG architecture, specifically for the ESRE (Elasticsearch Relevance Engine) suite, which combines machine learning models, a vector database, and advanced search and retrieval. ESRE’s embedding model enabled EY to generate and store large-scale vector representations at speed, forming the retrieval backbone for the generative AI pipeline. The platform also integrated seamlessly with LlamaIndex for LLM application orchestration and LangChain for workflow chaining, giving EY flexibility across its AI stack.
The results were measurable from the start. Accuracy across document extraction methods improved by 10–15% compared to previous approaches. Elastic’s RAG setup ran 3x faster than standard Native RAG configurations. EY’s solution handled diverse document formats at scale and could expand across organizational units without re-engineering the core architecture. For one of EY’s flagship use cases—ESG reporting—the platform allowed banks to streamline reporting on internal metrics and supply chain ESG data, meeting regulatory standards with less manual effort.
The deployment set EY up as an active builder of responsible generative AI infrastructure for the financial sector. With Elastic handling the search and retrieval complexity, EY can focus on the domain logic and client delivery rather than maintaining foundational components. The team sees the platform as a durable foundation for expanding generative AI capabilities as financial regulations and client needs continue to evolve.
Amazon Bedrock