Agent Control Planes Need a Robust Standards Stack — and They Don't Have One Yet
A new Forrester analysis surveying 47 agentic AI vendors lands on a sobering finding: 79% of participants recognise agent control planes as a distinct product category, yet the foundational standards stack those planes depend on is nowhere near complete. The piece maps a three-layer enterprise model — build, orchestrate, control — and then traces exactly where the gaps lie: OpenTelemetry's generative AI conventions are still in draft, W3C agent identity specifications remain unratified, and the competing A2A and ANP protocol proposals have yet to converge.
The implications for teams building production agentic systems today are real and immediate. Instrumentation without finalized telemetry conventions means observability dashboards are fragile and vendor-specific. Without a settled identity layer, cross-organization agent handoffs rely on ad-hoc trust arrangements rather than verifiable credentials. And governance — the audit trails, access controls, and policy enforcement that enterprise compliance teams require — can't be bolt-on if the underlying protocols don't support it natively.
The analysis isn't pessimistic so much as precise. It identifies what is actively in progress, which working groups are closest to producing usable drafts, and what realistic timelines look like. For architects deciding right now which orchestration stack to standardize on, knowing exactly where the holes are is more useful than waiting for a standards body to announce completion. The standards are coming — but the vendors will ship before them.