The collaboration is increasingly centred on Snowflake Cortex Agents, which combine Anthropic’s Claude models with Snowflake’s enterprise data, governance and security infrastructure. The approach allows organisations to build agents that retrieve information, reason across structured and unstructured data and take actions through connected business applications without requiring sensitive information to be moved outside the Snowflake environment.
The partnership has grown from a multi-year agreement valued at $200 million announced in December 2025. That agreement made Claude models available across Snowflake deployments on Amazon Web Services, Google Cloud and Microsoft Azure, while establishing a joint effort to take agent-based AI products to large enterprises. More than 12,600 Snowflake customers were covered by the expanded arrangement when it was announced.
Thousands of Snowflake customers are processing trillions of Claude tokens each month through Cortex AI, highlighting the rapid increase in enterprise demand for generative AI integrated directly with corporate information. Snowflake’s internal testing has also recorded accuracy above 90% on complex text-to-SQL tasks using Claude, an important capability for business users seeking answers from databases through natural-language questions rather than manually written queries.
The companies are targeting a broader transition from AI systems that merely answer questions towards software agents able to perform operational work. Cortex Agents can orchestrate multi-stage processes involving enterprise databases, external applications and specialised tools. Tasks can include analysing sales trends, generating reports, opening workflow tickets, updating customer records and performing calculations before presenting results to users.
Snowflake has added support for Model Context Protocol connections, providing a standardised mechanism for agents to interact with platforms including Jira, Confluence, GitHub, Salesforce, Google Workspace and Slack. The company is also developing reusable agent skills that allow organisations to package tasks such as forecasting and reporting into modules that can be deployed across departments.
Customer deployments indicate that the technology is spreading beyond conventional data analysis. United Rentals has been using natural-language interfaces across more than 1,600 locations to examine operational performance, while semiconductor manufacturer Wolfspeed has deployed dozens of AI agents across manufacturing, quality, supply chain and finance functions. Intercom has also used Claude through Snowflake Cortex AI as part of its Fin customer-service agent operations.
Governance has become a central element of the strategy as companies confront the security implications of giving AI agents permission to interact with business systems. Snowflake’s architecture applies existing access policies, audit controls and data-isolation rules to agent activity. Versioning allows developers to promote tested agent configurations into production or roll them back, while resource budgets are designed to limit spending at agent level.
That focus has expanded further with Cortex AI Gateway, introduced in July. The platform is intended to provide central oversight of how first-party and third-party agents access models, applications, tools and data. It can track token consumption, enforce spending limits and retain records of agent actions while routing requests between approved AI models according to cost, performance and availability requirements.
Cortex AI Gateway also reflects the increasingly multi-model nature of enterprise AI. Snowflake has relationships with Anthropic and OpenAI and is positioning its data platform as a control layer through which customers can choose models while retaining common governance. The company says the gateway supports more than 100 MCP servers and is being integrated with security and identity providers including Okta, SailPoint, 1Password, Aembit and Saviynt.
The shift places data-platform companies at the centre of competition over enterprise AI. Businesses have already spent heavily consolidating corporate information into cloud platforms and are increasingly reluctant to duplicate that data inside separate AI systems. Bringing frontier models to governed datasets offers an alternative in which AI applications operate closer to existing information and security controls.
Anthropic benefits from that distribution by placing Claude inside environments where large companies already conduct analytics and manage sensitive information. Snowflake, meanwhile, gains access to advanced reasoning models without relying on a single AI provider, allowing customers to build applications around different models as their capabilities, pricing and performance change.
The next stage is likely to depend less on whether AI agents can demonstrate impressive reasoning and more on whether organisations can operate them predictably at scale. Enterprises are demanding detailed permissions, audit trails, cost controls and mechanisms for isolating information between departments before allowing autonomous software to act on critical systems.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
