IBM enriches conversational search with Elastic

Delivering accurate, scalable, and secure AI-driven assistance

IBM’s clients needed AI-powered conversational assistants capable of providing accurate, real-time responses grounded in their proprietary data. To achieve this, IBM required a robust search solution that could scale to handle diverse data formats such as text, images, and videos. The solution also had to support secure access, seamless integration with large language models (LLMs), and efficient data processing and retrieval across multiple enterprise data sources.

Enriching experiences with RAG

By integrating Elasticsearch with IBM watsonx Assistant and IBM watsonx Discovery, IBM transformed the way its customers build conversational AI experiences. This partnership enabled IBM to deliver retrieval augmented generation (RAG) capabilities, allowing customers to enhance their chosen LLMs with their own proprietary data and improve chat experiences with precise, contextually relevant information. The integration empowers users to create AI assistants with advanced conversational search features.

Innovations like our vector database and ELSER model further streamlined vector search and semantic search, providing IBM’s customers with tools for faster, more accurate, and context-aware AI experiences.

Combining search and AI

Through the precision of search and intelligence of AI, IBM watsonx Discovery integrates seamlessly with Elasticsearch, giving customers the ability to perform semantic, federated, and vector search with their proprietary data. This powerful combination supports k-nearest neighbors (kNN) and approximate nearest neighbors (ANN) search, flexible multi-cloud model management, and Elastic’s open inference API, enhancing IBM foundation models like Slate and Granite.

Our solution also includes advanced features such as scalar quantization for efficiency gains and reranking models for optimized search results. With these capabilities, IBM’s clients can deliver AI-driven conversational assistants that combine business-specific data with LLMs, offering unparalleled speed, precision, and scalability.

IBM chose Elastic because our vector database has the ability to store text, image, and video embeddings while supporting secure and native hybrid search capabilities. Our extensive catalog of integrations allows IBM to unify data from various sources, including third-party, ensuring seamless retrieval across disparate data types.

With the world’s most downloaded vector database, Elastic is constantly investing in innovative enhancements — like scalar quantization, which delivers up to 8x to 32x efficiency gains for developers building AI-enabled apps. Elastic’s continuous investment in cutting-edge search technologies ensures that businesses using IBM watsonx Assistant and IBM watsonx Discovery remain at the forefront of AI-powered conversational experiences.