Capability shipped. Authorization is the open invoice.

A formal analyst category for AI agent management now exists, and IBM's placement signals that enterprise buyers have a structured evaluation frame for agent governance tooling. For Identity Control Surface planning, the criteria Gartner applies, multiagent orchestration, runtime guardrails, observability, and risk management, map directly to the control requirements that enterprise identity teams must specify before scaling agent deployments. The category itself validates that Compiled Corporation investment in agent infrastructure is mainstream procurement, not exploratory spend.

The three-plane architecture makes the Decision Surface explicit: design, runtime, and control are separated rather than collapsed into a single product layer. The journey ledger, which lets processes wait and resume, introduces stateful agent behavior at scale, raising Identity Control Surface questions about session continuity and authorization persistence across paused states. Cisco's attachment of AI Defense and Splunk observability to the control plane indicates that vendor-native monitoring is becoming a standard feature expectation, not an add-on.

A protocol that lets services verify payment before delivering resources moves agent-to-service authorization from a governance question to a transactional primitive. For Identity Control Surface, this means payment authority is now scoped at the session level, and the pre-funding model introduces a new pattern for bounding agent spend. For Compiled Corporation design, the ability to aggregate multiple requests under one session changes how firms will need to audit and attribute agent-initiated transactions across workflows.

This is the governance counter-argument to the Stripe Machine Payment Protocol announcement. Ashurst's framing clarifies that Identity Control Surface work in agentic payments requires two distinct layers: capability (can the agent technically invoke a payment) and authorization (has the customer mandate scoped that authority). Firms deploying agents with payment access who conflate these layers carry legal and regulatory exposure. The failure-tracing requirement Ashurst identifies maps to the observability gap that agent management platforms are beginning to address.

When thousands of engineers each run their own AI coding agents, spend and governance fragment fast. Cognition's product surfaces a Compiled Corporation control problem: agent usage that scales with headcount requires organizational-level attribution and policy enforcement, not individual billing accounts. For Identity Control Surface architects, this is a concrete case where non-human identity governance, scoped by team, project, and budget, becomes an operational requirement before costs become unmanageable.

A real estate operator running AI across a geographically distributed employee base is a Compiled Corporation deployment at operational scale, spanning maintenance coordination, tenant communication, and back-office workflows. The combination of ChatGPT Enterprise for general staff and Codex for technical work indicates a two-tier access model, which raises Decision Surface questions about how Heimstaden governs which tasks each tier can act on autonomously. The 2026-W43 index scores Workforce AI Access at 73, and cases like this illustrate what broad internal deployment looks like in practice.