What moved the register
This week the index splits cleanly along one fault line: the supply of agentic infrastructure is racing ahead of the enterprise's ability to govern, measure, and trust it. Three organization dimensions fell, Scaling Maturity, Governance & Ethics, and ROI Impact. Two product dimensions rose, the AI Interaction Layer and Agentic AI Deployment, and one product dimension fell, AI UX Maturity. One brand dimension rose, Agent-Ready Infrastructure. All three category composites and the overall score held.
Start with supply. AWS reported that AgentCore customers deployed 2x more agents in Q3 than the prior three quarters combined, across tens of thousands of customers, and that its Strands tooling cuts agent task cost by 32%, per About Amazon. IBM shipped cross-vendor agent lifecycle management. Alphabet's CEO described agents that book, form-fill, schedule and purchase. SAP told developers to build the Autonomous Enterprise. The Decision Surface is migrating toward the agent taking the action, with a 32% task-cost reduction and cross-vendor tooling now in the market. Agent-ready infrastructure, the interaction layer, and agentic deployment each moved up.
Now the demand side. The same week produced a dense wall of counter-evidence. Cloudera found 89% of enterprises have scaled back AI deployments over token costs, per IT Pro. MIT Technology Review Insights reported fewer than one in five organizations extract significant value and 38% are still mostly piloting, per MIT Technology Review. Deloitte named it an ROI reckoning. Scaling maturity and ROI impact both stepped down.
Governance is where the two sides collide, and it is the quarter's Identity Control Surface story. The Redwood survey found 77% of firms report fallout from ungoverned AI agents with only 58% enforcing a governance policy, per CFO Dive. Delinea found 84% reporting an AI incident and only 11% saying controls keep pace with deployment, per Redmondmag. WIRED documented a contractor's personal Claude agent reaching a secure Pentagon system. Read across these findings, the index reads the recurring gap as non-human identity: who owns the agent, who approved it, and what it can touch. Governance stepped down, and AI UX maturity fell alongside it as organizations reported agent actions they could not trace or explain.
The reading for the week: capability is now abundant and governance is scarce. Governance tooling shipped this week from AvePoint, and MIT Technology Review argued for real-time agent telemetry. The gap is between tooling and practice, and between deployment volume and the identity architecture that would make deployment safe to scale. The surveys report that controls lag deployment, with only 11% of organizations saying their controls keep pace. That is the work: before an agent touches a financial transaction, it needs an identity, an owner, a permission boundary, and an audit record. The index will reward the enterprises that build that layer ahead of the next deployment.