Variable spend needs standing visibility
- What did OpenAI cost us this period, and what changed?
- Which models and projects drive that consumption?
- Which teams own the workloads behind the invoice?
- How does API consumption compare with our seat-licensed AI?
- Are we paying twice for capability employees already have in an assistant license?
What OpenAI provides — and where the boundary sits
The OpenAI platform reports organization usage and cost with model and project breakdowns. That is the authoritative record of what you were charged, and Midgentic reads it directly rather than estimating from token counts.
The gap is context. The platform has no view of your assistant licenses, your developer AI seats, or the manually tracked tools departments buy — and therefore cannot tell you what share of AI investment this represents or whether it overlaps something you already fund.
What Midgentic actually ingests from OpenAI
- Metered API spend
- Organization cost reporting, normalized for currency and minor units.
- Model-level usage
- Consumption by model, so cost changes can be traced to specific workloads.
- Project attribution
- Usage split by project, mapping consumption to teams and initiatives.
- Usage trends
- A retained historical series rather than a single billing snapshot.
- ChatGPT Enterprise (manual / CSV)
- Seats, license terms, cost, and exported usage entered or imported, and clearly labeled as such.
- Value assumptions (entered by you)
- Documented assumptions behind estimated value, held separately from measured outcomes.
No prompt or completion content is collected.
The questions this data answers
- Consumption spend
- Actual metered cost per period in your reporting currency.
- Cost drivers
- Model and project breakdowns that explain movement rather than just reporting it.
- Build versus buy
- Internally built AI workloads compared with licensed products covering similar ground.
- Portfolio share
- How much of total AI investment is variable consumption.
- Full OpenAI relationship
- API consumption alongside manually tracked ChatGPT Enterprise licensing, each with its source shown.
Consumption is one input, not the whole picture
OpenAI consumption becomes useful to an executive only when it sits next to licensed platforms and manually tracked tools. That is the work of spend intelligence and AI visibility, and it is why the Anthropic API Platform is modeled the same way.
Common questions
What OpenAI data does Midgentic collect?
Usage and cost reporting for your OpenAI organization: metered spend, usage by model, and project-level attribution from the platform's administrative reporting endpoints.
Does Midgentic connect to ChatGPT Enterprise?
Not automatically today. The automated connector covers the OpenAI API Platform. ChatGPT Enterprise is supported as a manual and CSV provider — license terms, seats, cost, and exported usage can be tracked in the same portfolio, always labeled with its source.
Does OpenAI spend sync automatically?
Yes for the API Platform, because consumption is metered and reported by the platform. Seat-licensed products such as ChatGPT Enterprise use license terms you enter.
Does Midgentic read prompts or completions?
No. Only usage and cost reporting is collected. Request and response content is never accessed.
Is the OpenAI API connector generally available?
It is implemented and in beta, enabled per workspace during onboarding while it is validated with early customers.