Snowflake: Can AI Agents Unlock Fragmented Financial Data?

Snowflake has announced Cortex AI for Financial Services, a suite of AI capabilities designed to help financial institutions unify their data ecosystems and deploy AI models, applications and agents. 

The company also unveiled a managed Model Context Protocol Server, now in public preview, which enables organisations to connect proprietary data with third-party data in Snowflake from partners including FactSet, MSCI, Nasdaq eVestment and The Associated Press.

The MCP Server allows customers to connect their data with applications and agent platforms including Anthropic, CrewAI, Cursor, Devin by Cognition, Salesforce’s Agentforce and Windsurf. Financial services firms can use these connections to build context-rich AI agents and applications whilst maintaining security and compliance controls.

“The financial services industry has long been a leader in embracing new technology, and AI is no exception. However, the industry faces unique challenges in navigating fragmented data, robust compliance requirements, and the need for airtight security and governance,” says Baris Gultekin, VP of AI at Snowflake.

“By bringing AI directly to where their data already lives and enabling secure interoperability with remote agents, Snowflake is making it easier for highly-regulated industries like financial services to power business-critical use cases and tap into a unified ecosystem of best-of-breed data, AI and apps.”

Snowflake Cortex AI for Financial Services enables market analysis and fraud detection

Cortex AI for Financial Services is designed to accelerate tasks including market analysis, quantitative research, fraud detection, customer support and claims management. The ecosystem includes data from providers such as CB Insights, Cotality, Deutsche Börse, MSCI and Nasdaq eVestment through Sharing of Semantic Views, which will be generally available soon. Unstructured data publishers including CB Insights, FactSet, Investopedia, The Associated Press and The Washington Post are available through Cortex Knowledge Extensions, now generally available.

The suite includes Snowflake Data Science Agent, which automates data cleaning, feature engineering, model prototyping and validation. This allows teams to move from raw data to production-ready models for quantitative research, fraud detection, customer 360 and underwriting workflows. Snowflake Cortex AISQL adds functions for AI-powered extraction and transcription, enabling users to process documents, audio and images at scale for workflows including customer service, investment analytics and claims management.

Snowflake Intelligence, in public preview, offers business users a conversational interface to gain insights using natural language from data stored in Snowflake, as well as third-party data, applications and agents. This provides access to insights from both structured tables and unstructured documents without technical overhead.

Snowflake MCP Server connects Cortex Analyst and Cortex Search to external AI agents

The introduction of Snowflake MCP Server addresses challenges in connecting AI agents to enterprise systems. Previously, teams needed to create customised solutions for each integration, which slowed AI adoption. MCP provides a standardised way for large language models to integrate with data, APIs and services.

Snowflake MCP Server connects Cortex Analyst and Cortex Search to external AI agents through a standards-based MCP interface, unifying structured and unstructured data retrieval. This eliminates the need for custom integrations and accelerates the delivery of context-rich AI applications and agents. Remote agents can now connect with Snowflake data and third-party data shares from Snowflake Marketplace through Cortex Knowledge Extensions.

The MCP Server can connect with agentic applications and platforms including Anthropic, Augment Code, Amazon Bedrock AgentCore, CrewAI, Cursor, Devin by Cognition, Glean, Kumo, Mistral, Salesforce's Agentforce, UiPath, Windsurf, Workday and WRITER.

“Enterprises are moving from AI pilots to production, but until now, securely connecting AI to proprietary data has been a critical barrier,” says Jonathan Pelosi, Head of Industry, Financial Services at Anthropic. “Our partnership with Snowflake helps solve this by using MCP to connect each organisation's governed data directly to Claude. Customers can now use Claude's advanced reasoning on both structured analytics and unstructured documents via Cortex Analyst and Cortex Search, while maintaining enterprise security standards. With Claude and Snowflake, our joint customers are turning proprietary data into competitive advantage.”

Financial institutions and technology providers adopt Snowflake MCP Server

The managed MCP Server addresses barriers that have prevented enterprises from moving AI from pilots to production. Platform providers see the development as enabling deeper integration across the technology stack. “We’re excited to expand our partnership with Snowflake to deliver agent interoperability through protocols like MCP,” says Gary Lerhaupt, VP, Product Architecture, Agentforce at Salesforce. “This will enable deeper cross-platform connectivity and power more intelligent agentic experiences within Agentforce. Customers will easily be able to discover and connect to Snowflake's MCP server directly within Salesforce AgentExchange, accelerating the development and deployment of AI agents.”

The shift towards multi-agent architectures requires secure access to enterprise data at scale. The next wave of enterprise AI hinges upon orchestrating collaborative crews of specialised agents to automate complex processes,” comments João Moura, Co-Founder and CEO of CrewAI. “For these agentic workflows to succeed in the enterprise, they must be grounded in secure, high-quality data. Snowflake’s launch of a managed MCP Server provides the essential, secure pipeline for our agent crews to access, analyse and act upon governed data within the AI Data Cloud.”

Financial services firms are already deploying these capabilities to analyse unstructured data. Ian Macomber, Head of Analytics at Ramp, describes how the company processes customer feedback: “At Ramp, our mission is to help businesses save time and money. A critical part of that is deeply understanding their needs from the thousands of pieces of feedback we receive.

“With Snowflake Cortex AI, we can securely tap into and analyse our unstructured customer data, allowing teams across Ramp to ask questions in plain English and get instant answers. Snowflake makes it easy to democratise data-driven decision-making, which enables us to innovate faster and build the best possible platform for our customers.”