Web Directions Year round learning for product, design and engineering professionals

Treating Infrastructure as Data: Building an AI-Native Control Plane

The way we manage cloud infrastructure has been fundamentally static. You write Infrastructure-as-Code declarations, engineers review and approve them, and automated systems deploy them. The human remains in the decision-making loop. But what if AI agents could directly query, understand, and modify infrastructure the way they handle data—without the approval bottleneck for every action?

This possibility hinges on a crucial shift in how we model infrastructure itself. Instead of treating cloud resources as opaque objects that require specialized knowledge to manipulate, what if infrastructure were exposed as queryable data? What if an agent could ask "show me all databases across my cloud accounts that lack encryption" the same way it queries a relational database? This is the paradigm that StackQL proposes, and it opens unexpected possibilities for autonomous infrastructure management.

Most cloud management tools today are built around specific vendor APIs or proprietary abstractions. They're designed for human operators and assume that understanding a resource requires domain expertise. An engineer needs to know that AWS calls something "security groups" while Azure calls it "network security groups," and that they have different configuration models. The complexity compounds across multi-cloud environments, where operational knowledge must span multiple vendor domains.

Treating infrastructure as data inverts this model. If cloud resources can be exposed through a unified query interface—where the same SQL-like syntax works across AWS, Azure, GCP, and other platforms—then AI agents can reason about infrastructure without learning vendor-specific semantics. An agent doesn't need to understand the details of how an RDS cluster differs from an Azure database; it understands data, schemas, and relationships.

The MCP server pattern amplifies this capability. By exposing a unified control plane through the Model Context Protocol, agents gain a standardized way to interact with infrastructure. The agent becomes a tool user that can perform multi-cloud resource management tasks: provisioning new resources, querying current state, detecting configuration drift, and applying remediation—all from a single coherent interface.

This matters now because the operational burden of multi-cloud environments is becoming unsustainable for human-only management. Teams are fragmenting their attention across vendor consoles and tools. Configuration inconsistencies accumulate. Compliance requirements vary across clouds. What was once a strategic advantage—using the best service from each cloud provider—has become an operational tax. AI agents, if given the right abstraction layer, can navigate this complexity in ways humans can't easily match.

The practical impact is significant. Consider a compliance scenario where your infrastructure must meet specific encryption standards. Rather than having compliance engineers manually audit each cloud account, an agent with access to a unified data model can systematically query all resources, identify gaps, and generate remediation plans. A misconfigured security group in a forgotten AWS account becomes discoverable and actionable.

There's also a subtle but important shift in how infrastructure decisions get made. When infrastructure is opaque and requires expert knowledge to modify, change is slow and risky. When infrastructure becomes data that agents can query and understand, the nature of operations changes. Agents can propose optimizations, detect inefficiencies, and suggest resource consolidations. Humans oversee and approve, but the agent becomes the investigator rather than the human.

The maturity question remains open: how much autonomous infrastructure management should actually be autonomous? Most organizations will initially use agents as sophisticated auditing and recommendation engines rather than fully autonomous managers. But the architecture that enables comprehensive autonomous management is the same architecture that enables informed, collaborative human-agent infrastructure operations.

Jeffrey Aven is demonstrating these architectural patterns and real multi-cloud management applications at AI Engineer Melbourne 2026, June 3-4 in Melbourne, Australia.