S&P Global has launched Adaptive Retrieval, a service allowing customer AI agents and large language models to access and assemble licensed S&P Global data through natural language queries. It sits alongside the existing Deterministic Retrieval method, and both are now offered through a single interface, the S&P Global AI Data Portal.
Deterministic Retrieval, built on the Kensho LLM-ready API and available to customers since 2025, provides API-driven access through direct, structured queries – suited to focused tasks such as researching a specific company or analysing earnings call transcripts. Adaptive Retrieval handles requests spanning multiple datasets at once, aimed at multi-step work including in-depth research and report generation. Customers can use either method, or both.“The use of AI in financial services is rapidly accelerating and evolving, from tightly controlled workflows to fully autonomous, multi-agent systems,” says Sally Moore, Chief Client Officer and Co-Head of Market Intelligence. “With Deterministic and Adaptive Retrieval now available together, we’re ensuring that S&P Global’s trusted data is accessible across that full range of workflows, so customers can access data the way they need it today and adapt as their architectures evolve.”
Deterministic queries produce predictable request patterns: a known dataset, a known field, a known consumption record. Natural language retrieval that assembles across sources shifts that. S&P Global’s framing emphasises that its data is already cited, structured and ready for AI systems to consume, positioning citation and auditability as the differentiators – a response to the requirement that agent-consumed data be traceable back to source.
Continuity with the Kensho Reorganisation
The launch is a product expression of the structural move S&P Global made three weeks ago. The 6 July reorganisation split Market Intelligence into two verticals, Kensho Data & Platforms and Enterprise Solutions, elevating Kensho from embedded AI engine to named, client-facing layer. Sally Moore co-heads the first of those verticals alongside her expanded Chief Client Officer role; Bhavesh Dayalji took the new role of Head of Kensho Data & Intelligence. The AI Data Portal is the first significant client-facing product to carry the new structure’s logic.
In March, S&P Global described opening Capital IQ Pro to third-party AI agents through Model Context Protocol (MCP) servers and a single access point built with Kensho, combining client-owned content, S&P reference data and GenAI querying in one entitlements architecture.
MCP servers exposing licensed content into third-party AI workspaces have become table stakes among the major vendors, leaving entitlements, metering and governance as the contested ground. Central tool registries, proxied access and controller/worker hierarchies have emerged as the governance patterns firms are reaching for, alongside operational problems including context bloat and a lack of synchronisation between MCP servers inside a single organisation.
Dayalji places the retrieval layer at the start of a longer build. “For S&P Global, the data retrieval layer is only the beginning. Cited, verifiable S&P Global data provides the trusted foundation on which higher-value AI-native experiences can be built,” he says. “Now, S&P Global data flows directly into the tools and platforms where customers work through financial skills and plugins, and MCP apps that allow customers to visualize, explore, and interact with S&P Global data inside AI applications. This work continues as we develop additional workflow solutions and AI-native experiences that put trusted data at the center of how customers work with AI and multi-agent systems.”
The Competitive Frame
S&P Global describes itself as the first to offer both retrieval methods together,. LSEG has pursued exposure of its licensed data through MCP servers into whichever AI workspace clients happen to use, under its “LSEG Everywhere” positioning. Bloomberg has taken the opposite tack with ASKB, betting on Terminal-native context and the depth of proprietary content integration as the moat. S&P Global’s position sits between the two: proprietary data as the asset, but delivered outward into client architectures rather than defended inside a terminal.
Subscribe to our newsletter
