AI spending is no longer something you can apologize for at the end of the quarter. For enterprise teams that have moved past the pilot stage, ChatGPT costs can rival SaaS licenses, cloud bills, and — in some organisations — whole team headcounts. OpenAI acknowledged that reality on June 18, 2026, when it shipped new usage analytics and spend controls for ChatGPT Enterprise.

The update is practical, not flashy. But for any business owner or IT leader who has stared at an unexpected AI invoice and wondered where the credits went, it matters.

What OpenAI Shipped

The core of the update is a unified view inside the Global Admin Console. Admins can now see credit consumption across both ChatGPT and Codex — the AI coding platform — broken down by user, product, and model. That last part is important: not all models cost the same, and knowing which teams are running which models at what volume is the first step toward cost hygiene.

The same data is also available via a Cost API, so organisations can pull it into their own finance tools, dashboards, or BI platforms for deeper analysis. That integration point is a signal that OpenAI expects mature enterprise customers to own their cost data, not just view it in a vendor portal.

On the controls side, admins can now:

• Set a default credit limit for the entire workspace
• Configure group-specific limits for different teams or business units
• Create individual overrides for people who need higher capacity
• Approve or decline employee requests for additional credits — with employees able to attach context about what they are working on

Employees can see their own credit usage against their available budget and request more when they need it. That two-way visibility is sensible. It avoids the situation where an employee hits a hard wall mid-task and has no way to explain why they need more, or where an admin approves a blanket increase without knowing the business case.

Why This Is the Right Move at the Right Time

Enterprise AI deployments in 2026 have a cost problem — not because AI is overpriced, but because most organisations did not build the financial infrastructure to manage it before they rolled it out.

The typical pattern goes like this: a pilot gets approved, usage spreads faster than anyone expected, the bill comes in, and finance asks questions that IT cannot answer. Who used what? Which team is driving the cost? Is this usage generating value, or is it noise?

OpenAI’s update does not answer the “is this valuable?” question — that requires connecting spend data to business outcomes, which is still a human judgment. But it does answer the “who spent what” question, which is the prerequisite for everything else.

Forrester’s enterprise AI governance research consistently identifies cost visibility as one of the top barriers to scaling AI deployments. CIOs who cannot report clearly on AI spend find it harder to justify further investment, even when the results are good. This update removes one of the most common friction points in that conversation.

What This Means for Business

If you are running ChatGPT Enterprise across your organisation and you do not yet have cost governance in place, this is the practical starting point:

Map your usage by team before setting limits. Run the analytics for 30 days first. Understand which teams are high-volume users, which models they prefer, and what that maps to in terms of business output. Setting blanket limits before you understand usage patterns typically punishes your most productive users.

Use the Cost API to connect AI spend to your existing reporting. If your organisation already has a finance dashboard or BI tool, getting AI costs flowing into it puts AI on the same footing as your cloud or SaaS spend. That normalisation helps with budget planning and makes AI investment conversations with leadership far easier.

Build a credits request workflow. The new system supports employee credit requests with context notes. Use that. Teams that can articulate why they need more capacity — “deploying a new agent for client onboarding, expected to save X hours per week” — are much more likely to get approvals and to build internal trust in how AI resources are managed.

Treat the Cost API as a BI data source. Data teams and analytics leaders at EDNA Learn-trained organisations are well-positioned here. The same skills that make a Power BI or Python dashboard useful for financial reporting apply directly to AI cost analysis.

The broader signal is that enterprise AI has matured past the “try it and see” stage. Organisations that treat AI spend the same way they treat cloud spend — with budgets, owners, and accountability — are the ones that will scale AI sustainably. OpenAI just made that a bit easier to do.

If you are evaluating how to build AI agent infrastructure with proper governance baked in from the start, that is exactly what we build at Enterprise DNA. Talk to us about how to structure your AI operations.