Editor’s note (Oct 29, 2026): This article has been corrected. The original version carried a paragraph about crawler-metadata date disagreements – an observation about our own retrieval tooling, not a fact about the world – which has been removed. An inference about which parties benefit from the metering-before-pricing sequence has been replaced with the observable fact: both releases publish zero pricing figures for this integration. A claim that commercial terms “remain a bilateral conversation” has been cut, since neither release addresses commercial terms at all.
ServiceNow and OpenAI have deepened their partnership, introducing a mechanism that allows users to execute governed workflows directly within ChatGPT. While our October pricing survey mapped the industry-wide pattern of vendors leaving enterprise-agent pricing unpublished, this latest development serves as a concrete instance where the technical plumbing for governed access is deployed while the rate card remains entirely dark. There are zero pricing figures, tiers, or per-seat costs listed in ServiceNow’s releases for this integration.
The core of this integration relies on the ServiceNow Action Fabric, which connects OpenAI agents to enterprise workflows by layering ChatGPT’s reasoning capabilities over ServiceNow’s existing business rules, permissions, and context. A new ServiceNow MCP plugin, available in the OpenAI marketplace today, enables users to transition from simple inquiries to taking governed actions within a single conversation. This shift aims to move AI interactions from passive information retrieval to active, secure task completion within the enterprise environment.
For the teams that administer those workflows, the practical change is narrow and concrete: ServiceNow’s existing permissions and business rules still decide what the agent may do, now exercised inside the chat window rather than beside it. What is not carried into that window is any indication of what the capability costs.
To manage this complexity, ServiceNow has launched the Control Tower Alliance, a group of eleven founding members including Anthropic, Amazon Web Services, Cisco, HiddenLayer, Kong, NVIDIA, Okta, OpenAI, Stripe, WitnessAI, and Zenity. The alliance is designed to address the infrastructure of AI adoption, spanning security, governance, value measurement, and token metering. These elements are orchestrated through the ServiceNow Super Intelligence Control Tower, which acts as a centralized plane to discover, secure, govern, observe, and measure enterprise AI deployments.
Despite the focus on financial oversight, the actual cost of the integration itself is absent from both releases. Instead, the conversation around money is redirected toward the alliance’s capabilities. Scott Woody of Metronome, a Stripe product, noted that pricing is ‘increasingly tied to usage and outcomes’ and that the alliance helps businesses implement ‘reliable billing, real-time visibility into usage, and controls on spending.’ Similarly, Dan Graves of WitnessAI said its integration brings ‘new levels of AI policy control, usage analytics, and cost controls’ without ‘breaking regulations or budgets.’
The asymmetry is visible without overreading it. The alliance is building the machinery of price visibility — token tracking, spend controls, value measurement — while the price those instruments would measure stays unpublished for the integration in front of it. Both releases publish zero pricing figures, tiers, or per-seat costs for this integration. What the releases do not tell us is who, if anyone, benefits from that sequence — they name no buyers, no negotiated rates and no competitive effect, so this account does not claim one. What remains is the observable order: the metering apparatus ships first, the price stays dark.
That arrangement also reads as a signal about how enterprise AI procurement is settling. The governance layer is becoming the thing that is standardized, inspected, and shared across an eleven-member alliance, while neither release addresses the commercial terms at all. Procurement in the governance era starts to look like this: the controls get a public apparatus and the price stays behind the contact-sales ceiling, each arrangement judged on its own terms.
The roadmap makes the sequencing starker. The plumbing ships now — the MCP plugin is in the marketplace today, with enhanced capabilities expected by the end of 2026 — while voice capabilities for the Super Intelligence Control Tower are not expected until Q1 2027, with OpenAI-powered security capabilities slated for early access in January 2027 and general availability in Q2 2027, and the integration of Codex into the AI Workflow Factory expected later in 2027. The governance apparatus arrives in phases while the pricing decision stays open, and neither release commits to a phase for it. The question the staggered timeline leaves behind is which phase the pricing decision lands in — or whether a published rate arrives at all before every capability does.
The industry is assembling a sophisticated metering apparatus — incorporating value measurement, token tracking, and spend controls — directly into the platform. As the governance era matures, the contact-sales ceiling persists. The tools to measure and control enterprise AI spending are being built, but the price of the integration itself remains unpublished.
