Most AI demos stop at chat. In this one I connect Codex directly to Google Drive and let it work inside a real Google Sheets workbook - reading the data, analyzing it like an analyst, updating rows, editing formulas, and even making a business recommendation across multiple sheets.
To keep it realistic, I built a fake outdoor-retail company called Harbour & Pine Outfitters with the kind of workbook a small operations team actually uses every day: a live KPI dashboard, an Orders tab, an Inventory Pipeline, Vendor Codes with weighted-savings formulas, and a change log.
Then I hand it to Codex and watch it work:
• Analyze the Orders tab to surface top products by units sold and the fulfillment issues that need attention
• Add and update rows directly in the sheet while keeping the structure and formulas intact
• Bump a vendor's win probability and recalculate the weighted savings (real spreadsheet formulas, not plain text)
• Pull context across Orders + Inventory + Vendor Codes to recommend the next business decision on its own
• Write the next-step recommendations into the Dashboard and log everything in a change log
The point isn't the fake company it's the workflow. The same approach works for sales reports, project trackers, finance sheets, inventory, or client sheets. This is AI as an actual operations teammate, not a chatbot.
⚠️ One practical tip from the video: don't point Codex at production sheets directly. Work on a copy you can't be 100% sure every edit is correct.
0:00 What Codex + Google Workspace can do
0:40 The demo workbook: Harbour & Pine Outfitters
1:54 Codex analyzes the Orders tab
3:20 Adding a new order row in Sheets
5:15 Updating the Inventory Pipeline urgency
5:51 Vendor Codes: editing weighted-savings formulas
7:33 A cross-sheet business recommendation
9:20 Writing next steps into the Dashboard
10:56 Change log & wrap-up