Adoption is not the story. Identity is.

This is the Compiled Corporation problem made explicit: firms have AI but the firm's decision architecture has not been redesigned to accommodate it. The bottleneck is governed data and workflow integration, and the AAI Brand dimension sitting at 42 reflects exactly this failure to move from adoption signal to operational reality. Enterprise AI readiness practices that address legacy integration and agent safety controls are the direct answer to what this report names.

This is the Identity Control Surface made product. Broadcom is treating agent identity as a distinct governance object requiring its own trust chain, separate from user identity. The Agent-Ready Infrastructure dimension in this week's AAI moved to 55, and this announcement is evidence of why the category is gaining structure: vendors are now shipping dedicated non-human identity controls rather than bolting agent governance onto existing IAM. Organizations evaluating private AI deployments need to audit whether their current identity infrastructure can issue, revoke, and audit agent credentials at this layer.

The Decision Surface question here is direct: where did the human oversight sit, and why did it fail to intercept? This test result is the strongest available evidence that agent governance cannot be deferred to post-deployment monitoring. The Governance and Ethics dimension leads the AAI at 80, but a high score reflects organizational awareness, not organizational control. The gap between knowing this risk exists and having technical controls that catch it before production is where the real exposure lives. Any firm deploying coding or integration agents without adversarial pre-deployment testing is carrying unmeasured liability.

These cases are concrete Compiled Corporation examples: the firms restructured decision workflows around agents rather than adding agents to existing workflows. The distinction matters operationally. Clay's account management and Exa's developer integration automation show that the redesign is happening at the customer-facing boundary, where Decision Surfaces are most consequential. For enterprise AI readiness, the diagnostic question is whether your workflow redesign starts from the agent's capabilities or from the existing process map. These companies started from the capability.

This is the regulated-industry version of the pilot-to-production problem: the EHR integration moves ChatGPT from a general productivity tool to a governed Decision Surface inside clinical workflows. The Identity Control Surface question is immediate: what patient data access controls bind the agent session, who audits the queries, and how does the organization satisfy HIPAA accountability requirements when the decision aid is a third-party LLM? OpenAI entering healthcare data infrastructure raises the stakes for every enterprise AI program operating in regulated verticals. The governance architecture has to be settled before the capability is deployed, not after.

The hiring workflow is a canonical Decision Surface: consequential, regulated, and now apparently agent-automated on both sides simultaneously. When the human oversight point collapses, the output of the process carries no accountable decision-maker. This is a Janus Brands problem too: firms deploying AI recruiters are projecting an employer identity that is, in practice, a model. Candidates responding in kind expose the gap between the brand's stated values around human relationships and the actual interaction architecture. Organizations need to define explicitly where human judgment re-enters AI-mediated hiring before this pattern becomes a compliance issue.