From experimental chatbots to governed AI agents
Snowflake’s plan to acquire Natoma and expand its AWS Snowflake partnership marks a shift from experimental chatbots toward governed AI agents that can securely act across enterprise systems, reflecting a wider move to prioritize AI agent governance, control, and observability over raw model performance. In the early generative AI boom, most projects focused on text generation and question answering. Now, enterprises want autonomous agents that trigger workflows, send emails, touch SaaS tools, and modify records without breaking security policies. Snowflake’s CEO Sridhar Ramaswamy describes this as entering “the era of the agentic enterprise,” where intelligence must reliably turn into action. That ambition exposes a governance gap: powerful agents without clear identity, permissions, or policy enforcement can become “rogue agents” that leak data or bypass controls. Natoma’s technology targets this gap, positioning Snowflake as an infrastructure provider for safer, production-scale agent deployments.
Natoma and the rise of the agentic control plane
Natoma builds a gateway for Model Context Protocol (MCP) servers, acting as a policy and identity layer for AI agents calling external tools such as Slack, Jira, email, and internal APIs. Every tool call can be tied to a verified identity, checked against access policies, and logged for audit, so teams know who asked an agent to do what and whether it was allowed. Ramaswamy describes Natoma as a critical piece of Snowflake’s “agentic control plane,” extending Snowflake Intelligence and Cortex Code (Coco) from data queries into daily business applications. Natoma’s founders wrote that “enterprises need a trusted control plane for the agentic era” so AI agents remain grounded in governed data and policies. This directly addresses fears around shadow AI and rogue agents, turning governance from a patchwork of app-level settings into a unified gateway that sits between agents and the broader technology stack.
Sixth Snowflake acquisition signals consolidation around AI governance
The Natoma acquisition is Snowflake’s sixth major deal since June 2025, underscoring a consolidation strategy that centers the company as an infrastructure layer for agentic AI. Earlier Snowflake acquisitions added pieces around databases, migration, data discovery, observability, and AI-powered data pipelines, but Natoma zeroes in on AI agent governance and identity across tools. The deal, which would bring 20 employees to Snowflake, moves the control plane beyond data and development workflows into the applications where work happens. This helps Snowflake present a single interface where agents can query data, reason over context, and trigger governed actions. For enterprises, that means fewer point solutions and fewer integration gaps where rogue agents might slip through. Snowflake is not only selling AI models and copilots; it is building a fabric that coordinates how those agents operate, who they represent, and what they are allowed to do.
Expanded AWS deal underlines enterprise cloud expansion for agentic AI
In parallel with the Natoma news, Snowflake announced an expanded AWS Snowflake partnership that includes a five-year, USD 6 billion (approx. RM27.6 billion) commitment to AWS infrastructure. This enterprise cloud expansion signals the scale of compute and data demands tied to agentic AI workloads. The collaboration centers on Graviton-powered compute and AI infrastructure, tighter integrations for generative and agentic AI, and joint migration programs. Matt Garman, CEO of AWS, said, “Enterprises are rapidly moving from experimenting with AI to putting intelligent agents to work that drive real business outcomes.” Keeping AI agents close to governed enterprise data is a shared priority: instead of shipping sensitive data to external AI systems, Snowflake and AWS push architectures where foundation models and agents run adjacent to existing data platforms, with unified policies, monitoring, and identity controls.
What the Natoma move reveals about enterprise AI priorities
Taken together, the Natoma acquisition and expanded AWS deal show that enterprises now rank orchestration, control, and safety as highly as model accuracy. Snowflake cites internal research indicating that 96 percent of organizations still face barriers scaling AI across the business, with governance and integration high on the list. AI agent governance is becoming a first-class infrastructure concern, similar to identity and access management for humans. By tying MCP-based tool access to a central control plane, Snowflake aims to become the infrastructure layer where autonomous agents are deployed, monitored, and constrained. For buyers, that approach promises fewer rogue agents, fewer shadow AI deployments, and more confidence that “intelligence into action” does not mean “insight into incident.” As agentic AI spreads, vendors that combine powerful models with clear, enforceable guardrails are likely to win enterprise trust.