OpenAI Just Gave Enterprise Admins the AI Spend Controls They've Been Asking For

What Happened

On 18 June 2026, OpenAI announced new credit usage analytics and updated spend controls for ChatGPT Enterprise. Admins can now see a full breakdown of AI credit consumption by user, product, and model, and set layered spending limits that reflect how different teams actually work.

This is live today for all ChatGPT Enterprise workspaces. No rollout window, no phased release. If you're an enterprise admin, you can open the Global Admin Console right now and start using it.

When It Happened and What the Timeline Looks Like

OpenAI published the announcement on 18 June 2026. Earlier this year (the exact date isn't specified in the release), they had already introduced granular credit usage limits for custom roles, which allowed admins to restrict advanced model access by user type. Today's update builds directly on that foundation, adding workspace-level defaults, group-level limits, individual overrides, and the analytics layer to make sense of all of it.

Think of it as two phases: the first gave you the dials, and the second gives you the dashboard to know which dials to turn.

How It Works: The Mechanics Matter Here

The new capabilities sit inside the Global Admin Console and operate across four levels of control. Understanding how they stack is important before you configure anything.

Table 1: ChatGPT Enterprise Spend Control Hierarchy (Source: OpenAI, June 2026)

Level	Set By	Scope	Override Possible?
Workspace default	Admin	All users	Yes, by group or individual
Group limit	Admin	Defined team or role	Yes, by individual
Individual override	Admin	Named user	N/A
User-requested increase	Employee (request only)	Self	Pending admin approval

That last row is worth noting. Employees can now see their own credit usage in their workspace settings, request more capacity when they hit a limit, and attach context about what they're working on. Admins get an informed request rather than a support ticket or a blocked user. Power users keep moving; costs stay predictable.

What the Analytics Actually Show You

The new analytics view surfaces credit trends over time, identifies top users, and breaks spend down by user, product (ChatGPT vs. Codex), and model. That distinction between products matters more than it might seem. An organisation that has rolled out Codex to engineering but only uses ChatGPT more broadly will now be able to see exactly where credit consumption is sitting and whether adoption is genuinely broad or concentrated in a small cohort.

For teams that want to push the data further, the same figures are accessible via a unified Cost API, so you can pull it into your own BI tools or finance dashboards rather than staying inside OpenAI's console.

What This Actually Means for Your Business

We've spoken with growth leaders and founders who've been scaling ChatGPT Enterprise across their organisations and facing a common problem: they could see the monthly invoice but couldn't attribute it. Which teams were driving usage? Was the spend concentrated in a handful of power users or genuinely distributed? Was Codex delivering productivity gains in engineering, or was it idle after the initial rollout enthusiasm?

Those are no longer open questions. The analytics layer now makes AI spend legible in the same way cloud infrastructure spend became legible once tagging and cost allocation tools matured. That's a meaningful shift. It moves AI from a line item you manage reactively to an investment you can govern proactively.

There's also a softer implication for adoption strategy. The employee-facing credit view and request workflow removes a common friction point: the all-or-nothing problem where admins either gave everyone full access or risked blocking legitimate high-value work. Now the system supports nuance, which means you can roll out more confidently without the fear that one team's heavy usage will blow the budget for everyone else.

The Question Worth Asking Before You Configure Anything

The risk with tools like these is that teams spend more time managing the controls than thinking about the outcomes they're trying to drive. Before you open the console and start setting limits, it's worth being clear on what "valuable usage" looks like in your context. Credit consumption isn't inherently good or bad; context is everything. An engineer running deep Codex sessions on a product-critical problem is a very different signal from a team running exploratory prompts with no clear output.

We'd encourage finance and growth leaders to use the new analytics as a diagnostic, not just a cost tool. The usage patterns that emerge in the first few weeks will tell you a great deal about where AI is genuinely embedded in workflows and where adoption is still surface-level.

If you're thinking about how to build that kind of AI visibility into your broader growth and content strategy, our advisory work on GEO and AI search readiness covers how organisations are structuring those decisions right now. And for a broader view of how enterprise AI adoption is reshaping content and discovery, see our latest insights from the Surge45 team.

The Action to Take This Week

If you're a ChatGPT Enterprise admin, log into the Global Admin Console, navigate to the Billing tab, and review the usage breakdown before you change any settings. Let the data inform the configuration, not the other way around. If you're a founder or growth leader without direct admin access, ask your IT or operations lead for a usage report this week. The numbers will likely surprise you, and they'll give you something concrete to act on.