Source: S&P Global Press Room — 2026-08-12
Summary
S&P Global expanded its partnership with Microsoft to bring its capital markets data, earnings call transcripts, private company financials, and global securities data into Microsoft 365 Copilot and Copilot Cowork's agentic workflows — a major scope increase from a prior integration that covered only S&P's Energy AI Ready Data. The connection runs through the S&P Global AI Data Portal's Deterministic Retrieval solution, branded the Kensho LLM-ready API, designed to return cited and verifiable results rather than free-form generated text. New surfaces include a Copilot in Excel connector for financial modeling and a plugin for Copilot Cowork's agentic experiences, both sitting under S&P's newly formed Kensho Data Platforms unit.
Key Takeaways
- The expansion covers S&P Global's full proprietary dataset — capital markets data, earnings call transcripts, private company financials, and global securities data — inside Microsoft 365 Copilot, up from a prior integration limited to Energy AI Ready Data alone.
- Retrieval runs through the S&P Global AI Data Portal's Deterministic Retrieval solution, also marketed as the Kensho LLM-ready API, engineered to hand Copilot cited, source-traceable answers instead of generated text that could hallucinate figures.
- New product surfaces include a Copilot in Excel connector built for financial modeling and reporting, plus a plugin that feeds S&P Global data into agentic workflows inside Copilot Cowork.
- Named use cases include company research, financial analysis, transcript intelligence, peer benchmarking, and competitive analysis, all accessible without leaving the Microsoft 365 environment.
- The integration sits under S&P Global's newly formed Kensho Data Platforms unit, the same group that runs the Capital IQ Pro terminal, signaling a dedicated push to package S&P's data for AI-native interfaces.
Reel Script
Hook: An AI assistant that confidently invents a company's quarterly revenue number is a liability, not a feature, in finance. S&P Global just wired its entire licensed dataset into Microsoft 365 Copilot through an API built specifically to cite its source for every figure it returns.
Core Concept: Most generative AI tools answer financial questions by predicting plausible-sounding text, which is exactly how they end up inventing numbers that were never in any filing. S&P Global's answer is "deterministic retrieval," marketed as the Kensho LLM-ready API: instead of letting a language model guess at a figure, the system looks up the actual licensed data point — a specific line in an earnings transcript, a securities price, a private company's reported revenue — and hands the model that exact value along with a citation back to the source. The AI still writes the sentence around the number, but it isn't allowed to make the number up. That distinction matters enormously in finance, where an analyst using a wrong EBITDA figure in a model isn't a minor glitch, it's a bad investment decision. Think of it as the difference between an assistant who remembers facts from memory and one who always looks up the actual document before answering, then hands you the page number.
Hands-On: The mechanics show up in two new product surfaces. Inside Copilot in Excel, an analyst can now pull S&P Global data — capital markets pricing, earnings call transcripts, private company financials, global securities data — directly into a financial model or report, with each figure traceable back to its licensed source rather than pasted in from memory. Inside Copilot Cowork, the same data becomes available to agentic workflows, meaning an autonomous agent chaining together multiple steps of a task (say, pulling comparable-company financials, then building a peer-benchmarking table) can call the Kensho API mid-task instead of hallucinating placeholder numbers to keep the workflow moving. This is a significant expansion in scope: the prior version of this Microsoft tie-up only piped in S&P Global's Energy AI Ready Data, a single vertical dataset. The new version opens the full licensed catalog, spanning named use cases like company research, transcript intelligence, peer benchmarking, and competitive analysis. The integration sits under S&P Global's newly created Kensho Data Platforms unit, the same group that runs the Capital IQ Pro terminal — signaling S&P is treating "package our data for AI copilots" as its own product line rather than a one-off partnership.
Takeaway: Citations aren't a nice-to-have in financial AI, they're the whole product — nobody trusts a number they can't trace back to a filing. This is the most credible version of "AI for finance data" on the market right now. If your team lives in Excel and Copilot, this is worth testing before your next earnings season.