ChatGPT and Excel: turn invoices, orders and PDFs into useful reports
Use the files on your Windows PC to build a checkable report instead of asking AI to invent a table without source data.
Excel is still the operational center of many small businesses: invoices, costs, inventory, deadlines and customer lists often end up in spreadsheets. LocalCommander lets ChatGPT work with the actual files on your Windows PC instead of relying on copied snippets.
The useful part is not simply generating a pretty table. A good workflow records assumptions, counts source rows, separates uncertain records and saves the result to a new file rather than overwriting the originals.
Process 1: build an invoice register from PDFs
Start by listing the files in the target folder. Determine which PDFs contain readable text and which are scans or otherwise ambiguous. Extract only fields that can be supported by the source: document number, dates, supplier and amounts.
Put unclear documents into a review queue instead of guessing values. Link each output row to its source file where possible.
Review the supported PDFs in Invoices/2026 and create a new Excel workbook with document number, date, supplier, due date and amounts. Link each row to the source file. Put unreadable or uncertain documents into a separate review sheet. Do not send or book anything.
Process 2: combine several sales exports
When orders, product costs and delivery data live in separate exports, first identify a reliable shared key. If there is no trustworthy SKU or order ID, report the mismatch instead of joining records by similar names.
After validation, calculate useful measures such as gross margin, returns, shipping costs or monthly revenue and keep control totals in the output.
In the Reports folder I have orders.xlsx, costs.csv and shipping.xlsx. Identify reliable shared IDs, report duplicates or missing keys, calculate margin by SKU and month, and save the result as a separate XLSX with control totals and a manual-review sheet. Do not overwrite the originals.
Process 3: proposals from an Excel price list and a Word template
A price list and an approved DOCX template can be combined into a draft proposal with selected products, calculations and terms.
Never invent rates, discounts or missing commercial terms. Mark missing fields for review and inspect the final document before it is sent to a customer.
Process 4: deadlines and missing fields
A project spreadsheet becomes much more useful when empty dates, overdue rows and missing owners are surfaced in a separate priority report.
If the workbook contains contact details, prepare reminder drafts first. Sending messages is a separate action and should not be assumed.
Combine spreadsheet work with browser work
A stronger workflow crosses surfaces: read a price list, compare it with a signed-in store panel, prepare a list of differences and update only the approved fields after review.
That does not require a dedicated connector for every web app, but it does require an authorized session and a browser interface that can be handled reliably.
How to verify the output
Record source row counts, rejected rows, join rules, control totals and output paths. Compare a sample of results against the original documents.
For accounting or financial records, apply your own data-handling rules and have the responsible person review the results. LocalCommander does not replace an accountant or submit tax filings on its own.
Good to know: Support for scans, formulas and special document formats depends on the file and available tools. The agent should report missing or unreadable data instead of guessing.
Turn the guide into a real task.
Connect your Windows PC, choose a low-risk task and verify the result. Start on the Free plan.