Quarter-End Reviews Can Flag Mismatches Before Board Materials Go Out
OpenAI presents ChatGPT Work as a quarter-end control that compares NetSuite data with financial models, executive memos and board materials, then issues a “no-go” when the figures do not match. In the demonstration, it flags two critical discrepancies—including updated June actuals and an outdated forecast in a revenue memo—traces them across the affected files, and directs teams to correct them before rerunning the review.
Two critical mismatches produce a no-go result
ChatGPT Work frames quarter-end review as a publication decision: “NO-GO.” It compares finance materials pulled from Google Drive—a revenue memo for executives, a board package, and a financial model—with figures posted in NetSuite. Across thousands of rows, multiple tabs, and several document types, the review identifies two critical issues and names the affected reports and numbers.
One issue is a June update to actuals in NetSuite. The other is an outdated forecast flowing through the revenue memo relative to the model. A changed figure can exist in the underlying system while an earlier version persists in executive prose or board materials. The review is designed to test that chain of related documents against the financial system, rather than inspect any one spreadsheet or presentation alone.
The no-go finding makes the decision explicit: the materials are not ready to send until the mismatches have been addressed.
One changed actual can persist across the model, memo, and deck
A stale number is not necessarily confined to the document in which it originated. It can be reflected in a forecast model, repeated in an executive memo, and turned into a chart in the board package. ChatGPT Work is shown tracing an identified discrepancy to each of those surfaces.
The financial model appears as a Google Sheets file titled “OpenAI FY26 Governed Finance Model.” Through the browser embedded in chat, the reviewer can open the file and use its editing or commenting controls to ask the responsible team to fix the issue. The purpose is not to have ChatGPT make the correction autonomously; it is to direct people to the relevant file and number.
The workflow also identifies affected board materials. The “OpenAI FY26-Q2 Board Package,” shown in Google Slides with a waterfall chart, is one of the documents that can carry the error forward. The reviewer can move among the tabs and materials implicated by the finding rather than reconstruct the document trail manually.
ChatGPT Work presents an “Ordered correction sequence” to guide that process. In practical terms, the sequence is: identify the mismatch, open the affected material, send the team a correction request, and work through the issues one by one. The workflow narrows the search, but the team remains responsible for resolving the source data or document error.
The 20-page “FY26 H2 Revenue Memo” shows why prose belongs in the review. Dense narrative makes it easy for an old figure to survive even after the model changes. ChatGPT Work highlights the relevant figures in red, directing the reviewer to the passages that need revision instead of requiring a manual search through the entire memo.
Corrections are followed by a new readiness decision
Once the team has made the requested changes, the review can be rerun. The user says that a fresh answer on whether the materials are ready to send can be returned in minutes.
That repeatability matters because changing a number in the model does not establish that the same value has been updated everywhere else. The rerun checks the revised materials against NetSuite again and can still return a negative readiness result if inconsistencies remain. The practical value described is confidence in the numbers being released, with the initial no-go decision serving as the alternative when the documents do not yet agree.
The same cross-check can extend beyond the immediate quarter-end package. The example board package is 34 pages, but it may sit inside a broader board deck running to hundreds of pages. In that setting, the workflow is presented as a way not only to cross-check figures, but also to surface conflicting themes across the documentation.