Kensho

Kensho functions as the core artificial intelligence and innovation hub for S&P Global, specializing in converting messy, unstructured data into clean financial intelligence. The platform is engineered specifically for institutional investors, quantitative analysts, commercial banks, and software engineers who build enterprise-grade generative AI applications. By combining domain-specific machine learning models with deep financial datasets, it bridges the gap between raw public market documentation and automated, data-driven trading strategies.

The platform divides its enterprise data access across two distinct retrieval frameworks depending on the complexity of the task. For precise corporate research where the exact endpoint is known, it relies on Deterministic Retrieval to pull exact metrics like specific earnings release dates or structural line items. For open-ended, complex financial analysis, it runs an Adaptive Retrieval engine that dynamically evaluates an ambiguous natural language prompt, automatically routes queries across multiple structured datasets simultaneously, and builds a comprehensive cited response for the end user.

The Kensho LLM-ready API is a developer framework that allows large language models like ChatGPT, Gemini, and Claude to pull verified S&P Global financial data using natural language queries. Rather than building complicated custom API wrappers for every dataset, developers can use this single endpoint to let an AI agent seamlessly look up historical financials, multi-entity margin comparisons, and market transaction histories. To simplify integration, Kensho provides a dedicated Python SDK alongside built-in Model Context Protocol server support, meaning enterprise AI agents can securely execute data function calls straight out of the box.

The platform features an advanced toolkit designed to automate complex corporate research and data ingestion pipelines:

Kensho Extract: Uses deep learning to parse unstructured PDFs and presentations, transforming complex tables, charts, and text sections into machine-readable JSON or CSV files for RAG pipelines.

Kensho Link & NERD: Automatically maps vague or fragmented company, person, and location references found in text directly to verified global identifiers and S&P Capital IQ Pro profiles.

Kensho Scribe: A specialized voice-to-text engine optimized for financial audio that boasts a significantly lower word-error rate than general speech engines when transcribing earnings calls filled with industry jargon and heavy accents.