Coveo has announced the general availability of its Passage Retrieval API, designed to enhance the accuracy and security of generative AI applications for enterprises by reducing hallucinations.

The new API aims to improve information retrieval from enterprise systems, helping organizations deliver more precise data to large language models and AI applications, including Salesforce Agentforce, Microsoft Copilot, Amazon Bedrock Agents, and SAP Joule.

"While large language models hold immense potential, their success hinges entirely on the quality of the data they access," said Laurent Simoneau, Co-Founder, President and CTO at Coveo.

"Many organizations are adopting Retrieval-Augmented Generation (RAG) and quickly realizing that the hardest part - the 'R' - lies in retrieving precise and relevant information from scattered enterprise systems."

The platform features a unified index that connects data from various enterprise systems including CRMs, ERPs, ITSMs, and CMSs through prebuilt connectors. It employs hybrid semantic and lexical indexing, along with machine learning models for automatic relevance tuning.

Early results have shown promise, with one Fortune 500 software company reporting a 22% improvement in article and passage retrieval accuracy within two weeks of implementation, leading to a 73% improvement in their internal GenAI chatbot's answer accuracy.

The API supports multiple use cases, including custom GenAI experiences, chatbots and virtual agents, internal knowledge management, and customer-facing support. It is compatible with various AI platforms and LLM models, including Google Gemini, Anthropic Claude, and OpenAI GPT-4.

https://www.coveo.com/en/lp/passage-retrieval-api