How S&P Global Uses Databricks Genie and MCP to Make Structured Data Conversational

From Data to Dialogue: Reimagining the Structured Data Experience

The challenge S&P Global Energy faces is one shared by nearly every data-intensive enterprise: vast structured data assets hold enormous value, yet the barrier to discovering and consuming that data remains frustratingly high for customers. Traditional querying methods typically require users to understand database schemas, master specific query languages, or rely on dedicated analytics teams.

According to source materials, S&P Global's core objective is to fundamentally improve how customers discover and consume data — making data assets conversational rather than forcing users to hunt through reports and dashboards.

The underlying logic is straightforward: when natural language becomes the interface between people and data, accessibility increases dramatically. Users no longer need to be SQL experts — they simply ask questions in everyday language, and the system interprets the intent, retrieves the relevant data, and delivers an answer.

The Role of Genie Agents and MCP

The technical foundation of this solution rests on two pillars: Databricks Genie Agents and MCP (Model Context Protocol). Genie is Databricks' conversational analytics capability, allowing users to interact with structured data in natural language while the system translates those queries into database operations and returns results.

Why MCP Is Needed

MCP, as a standardized protocol between models and external data and tools, addresses the problem of "context access." Large language models have no inherent knowledge of an enterprise's internal data structures or business semantics. MCP provides a regulated channel that allows models to access structured data assets in a safe, controlled manner.

Combining Genie Agents with MCP means S&P Global can connect its massive energy data assets to a conversational AI experience. A user's question is parsed by an agent, routed through MCP to the underlying data, and ultimately returned as a structured analytical response in conversational form.

Background: What Is MCP? MCP (Model Context Protocol) was introduced and open-sourced by Anthropic in late 2024, with the goal of establishing a unified standard for interactions between large language models and external tools and data sources. Before MCP, every AI application that needed to connect to external data had to build a custom integration layer for each data source — resulting in high integration costs and poor maintainability. MCP defines a standardized client-server communication protocol: the model (client) sends requests to an MCP server through a unified interface, and the server translates those requests into operations against specific data systems. The key advantage of this design is that data providers only need to implement an MCP server once to be callable by any MCP-compatible AI application. The protocol layer also supports built-in access control and auditing, preventing models from being exposed directly to raw data systems. For data-as-a-product companies like S&P Global, MCP opens up data accessibility while providing the necessary security guardrails.

The Value of an Agent-Based Architecture

Choosing an "agent" approach rather than a simple Q&A interface reflects a more mature design philosophy. Agents can handle multi-step tasks, call on multiple data sources, and reason and synthesize information when needed. This is especially important for the complex analytical demands of the energy sector — a single question may span multiple data tables, multiple time dimensions, and intersecting metrics.

Background: What Is Databricks Genie? Databricks Genie is a natural language data analytics feature introduced by Databricks in 2024, built into its AI/BI product line. Its core mechanism automatically translates a user's natural language question into a SQL query, leveraging metadata in Databricks Unity Catalog (table schemas, column annotations, business term definitions) for semantic alignment, then executes the query against the underlying data lake or data warehouse and presents results conversationally. A key design priority for Genie is trustworthiness — when query intent is unclear, it proactively asks for clarification and surfaces the generated SQL for user verification, rather than silently returning a potentially incorrect number. Compared to generic Text-to-SQL approaches, Genie is deeply integrated with Databricks' data governance layer, making the query generation process aware of data lineage, column-level permissions, and business semantics — particularly critical for heavily regulated industries like finance and energy.

Industry Significance of Conversational Structured Data

S&P Global's implementation is a compelling example of enterprise AI adoption today. A growing number of institutions with vast structured data assets are experimenting with large language models and standardized protocols to lower the barrier to data consumption.

The value of this model goes beyond efficiency gains — it's about unlocking the long-tail value of data. Data combinations and insights that were previously overlooked due to high query barriers can now be reached by everyday business users through natural language. For companies where data is the product (like S&P Global), this has direct implications for customer experience and competitive differentiation.

That said, making structured data conversational is not without its challenges. Query accuracy, correctness of semantic mapping, data permission management, and the trustworthiness of model-generated results all require continuous refinement in real-world deployments. Protocols like MCP exist precisely to provide the standardization and control needed to address these challenges.

Summary

By leveraging Databricks Genie Agents and MCP, S&P Global Energy has transformed static structured data assets into a conversational interactive experience — representing a clear direction in the evolution of enterprise data platforms. As natural language becomes the primary interface for data, how data is discovered and consumed will be fundamentally redefined.

Given that the source materials share limited technical specifics, this article focuses on interpreting the approach and its broader industry significance. As more enterprises adopt similar agent-based, protocol-driven architectures, conversational access to structured data is poised to become the dominant paradigm for data consumption.