Convert an AIA G702/G703 pay application PDF to Excel

The short answer: there are three ways to get an AIA G702/G703 pay application out of a PDF and into Excel: retype it, run it through a generic PDF-to-Excel converter, or use a tool that extracts the data and verifies it against the form’s own arithmetic. Retyping a 40-line G703 takes 20 to 45 minutes and adds keying errors. A generic converter or OCR tool returns cells with no guarantee any of them are right. Extraction with verification recomputes every identity the form must satisfy, per line G = D + E + F, H = G / C, I = C - G, the sum of Column G equal to G702 Line 4, and Lines 5 through 9 of the certificate, so a misread digit breaks an equation and gets flagged instead of exported. Scans work too: the same arithmetic catches what OCR misreads. For a receiver certifying payment from the spreadsheet, the verification is the product.

Option 1: retype the AIA G702/G703 by hand

A 40-line G703 continuation sheet is 40 rows across nine columns, close to 360 figures, plus the nine payment lines and the change-order block on the G702. Keying that from a PDF takes 20 to 45 minutes per application, and manual entry introduces the exact transposition errors you are trying to catch. There is a second cost that matters more to a receiver: once you retype it, you own the keying mistake. If the spreadsheet you certified from says $84,000 where the application said $48,000, that error is now yours. And on the fifth of the month, when a dozen subcontractors bill at once, minutes per document become hours per draw.

Option 2: run the AIA G702/G703 through a generic PDF-to-Excel converter

Generic converters, spreadsheet “import PDF” features, and construction OCR tools all do the same thing: they read the page and hand you cells. On a clean native PDF the grid mostly survives. On a scan or a phone photo the digits wobble, on a custom “G702-style” layout the columns drift, and on a multi-page G703, where the grand totals print only on the final page, the total row lands wherever the parser left it.

The deeper problem is not read quality. A converter is a format tool, not a verifier. It does not know that Column G must equal D + E + F, or that the G703’s grand total must reappear as Line 4 on the G702. A misread digit that turns $84,120 into $34,120 produces a spreadsheet that looks finished and is wrong by $50,000, and nothing in the file will tell you. Extraction without verification does not remove the checking work. It moves it downstream to you.

Option 3: extract and verify the AIA G702/G703 in one step

A pay application is the rare document that can prove its own extraction, because its numbers are over-determined: the same figure is constrained by a row formula, a column total, and a cross-form tie. PayAppCheck uses that redundancy. It extracts the G702 data and every G703 line from a native or scanned PDF, maps custom layouts back to the standard columns, then recomputes every identity the form must satisfy. A misread digit breaks at least one equation, so it surfaces as a flag with the expected figure, the stated figure, and the dollar delta, instead of passing silently into your spreadsheet. You resolve the flags in a review grid beside the source document, then export. For the line-by-line mechanics, see the worked schedule of values example. If what you need is a blank workbook to bill from rather than a way to read inbound ones, take the free template, an AIA-style workbook with the same formulas built in.

What extraction alone cannot tell you about an AIA G702/G703

These are the identities a converter never checks and a verifier recomputes on every document:

• On every G703 line, Total Completed and Stored to Date must satisfy G = D + E + F.
• Percent complete is H = G / C, and any line over 100 percent is billed beyond its scheduled value.
• Balance to finish is I = C - G on every line.
• Column C must sum to the Contract Sum to Date, G702 Line 3, and Line 3 = Line 1 + Line 2.
• Column G must sum to G702 Line 4. This is the single most important check on the document.
• Retainage is 5a = rate x completed work and 5b = rate x stored material, and Line 5 = 5a + 5b.
• Line 6 = Line 4 - Line 5, Line 8 = Line 6 - Line 7, and Line 9 = Line 3 - Line 6.
• Line 7 must equal Line 6 of the previous application, which ties this month’s billing to last month’s.

Together these catch two error classes with one mechanism: digits the OCR misread, and errors that were in the application before it was ever scanned. A converter with perfect read accuracy still passes a wrong document straight through. For how verification separates receiver-side tools from submission platforms, see pay application software compared.

AIA G703 to Excel, CSV, or JSON: what the export looks like

The Excel export is a workbook with two sheets. The Data sheet holds the full schedule of values, columns A through I, alongside the G702 summary lines. The Flags sheet lists every check that ran, with the expected figure, the stated figure, and the delta, so the next person in the chain can see exactly what was verified and what was corrected. CSV carries the same line items for import into an accounting system. JSON carries the whole reconciled structure, flags included. For volume, the extraction API takes a POST with the PDF and returns the reconciled JSON, with a webhook when processing completes, so a lender or GC handling dozens of draws per month can skip the UI entirely.

Questions people ask