How to Batch Process 30 Subcontractor Invoices from Different Formats
into One Project Cost Sheet

The Draw Cycle Creates the Batch — Whether You're Ready or Not

In most industries, a batch is something you choose to create. You let invoices accumulate for a week, then process them together because it's more efficient. In construction, you don't choose. The subcontract language sets a single monthly deadline — typically the 20th or 25th — and every sub responds to the same date. The result is a flood of paperwork that arrives in a 48-hour window, every month, for every active project.

This isn't a technology problem that needs solving. It's a structural feature of how the industry pays for work. Under AIA A201 §9.3, the contractor submits one consolidated payment application to the owner each month. That consolidated application is only as clean as the data feeding into it — which means 30 individual subcontractor invoices need to become one set of numbers, accurately, before the GC's own draw deadline.

For a GC running three projects, that's 90 invoices converging on the same three-day window. The project accountant isn't facing a data entry problem. They're facing a throughput problem: 90 data sets need to move from PDF to spreadsheet in the time it would normally take to process 15.

Why Single-Invoice Extraction Doesn't Survive the 25th

If you process invoices one at a time — upload, extract, download, repeat — each one takes 3–5 minutes even with AI assistance. For 30 invoices, that's 2.5 hours of screen time. But the real cost isn't the 2.5 hours. It's the downstream work the per-invoice approach creates:

Thirty separate downloads. Each extraction produces its own Excel file. After processing the last invoice, you have 30 files with identical column structures and different data. Someone has to open each one, copy the data row, and paste it into a master sheet — or write formulas to consolidate them. That consolidation step takes as long as the extractions themselves, and it's where merge errors creep in: a row pasted into the wrong project tab, a column shifted by one, a file missed entirely.

Per-file configuration drift. When you define extraction columns one invoice at a time, the 23rd invoice inevitably gets a slightly different column set than the 5th — not because you changed your mind, but because you're tired and the drywall sub's invoice format confused you about whether "Materials Stored" should be a separate column. By the end of the batch, your columns don't match across files, and consolidation becomes a manual reconciliation exercise.

You can't spot cross-invoice patterns. Processing invoices individually means you never see the full picture until consolidation is done. You finish at 6 PM, open the master sheet, and only then notice that two subs billed for the same cost code at rates that don't match — or that a change order amount that appeared on three invoices doesn't add up to the approved CO total. By the time you catch it, the draw deadline is tomorrow morning.

Define Columns Once, Extract from Every Sub

The alternative is to reverse the workflow: define your output schema first, then feed every subcontractor invoice through the same extraction pipeline. For a step-by-step guide on setting up the field extraction itself — including how column-name extraction finds values by meaning rather than page position — see our walkthrough on subcontractor invoice data extraction. Here, the focus is on what changes when you multiply the input by 30.

You type your column headers once:

Then you upload all 30 invoices at once — the concrete sub's AIA-style PDF, the HVAC sub's company letterhead, the electrician's QuickBooks-generated invoice, the painter's handwritten bill. The extraction engine processes them in parallel, locating each column's value in each document regardless of where it appears. You download one file: a spreadsheet with 30 rows, each representing one subcontractor invoice, with identical columns.

The operational difference is that the batch approach eliminates the consolidation step entirely. You don't merge 30 files. You verify one file. And because all 30 rows were extracted with the same column definitions, you can sort by Job #, filter by Cost Code, or subtotal by subcontractor immediately — no reformatting, no column alignment, no "wait, which file had the plumbing line items?"

If your subcontractor invoice volume is growing beyond what batch processing alone can handle, the scaling framework in our guide to scaling invoice processing without adding headcount covers the organizational side of the problem — from process design to team structure at different volume thresholds.

Retainage Variance: When Every Sub Has a Different Withholding Rate

In a batch of 30 subcontractor invoices, retainage is never a single number you can apply across the board. One sub's contract withholds 10%. Another withholds 5%. A third has completed their scope and is billing at 0% retainage — they've already earned back what was withheld. A fourth is on a public works project where state law caps retainage differently than the GC's standard subcontract.

If you're processing invoices individually, the retainage calculation happens in your head — or more likely, in a formula you type into each row's cell after extraction. For 30 invoices, that's 30 manual formula entries. Get one wrong, and the sub's Net Due is off by $4,200, which you'll discover when they call to ask why their check was short.

The batch approach handles this with computed columns: a column that doesn't just extract a value from the document — it calculates one during extraction. You define a column called "Net Due" with the logic Total Billed × (1 − Retainage %). The AI reads the Total Billed and Retainage % from each invoice, performs the calculation, and populates Net Due automatically — per invoice, per subcontract, per the actual numbers on the page.

The result is a column where every row's Net Due reflects that specific subcontract's retainage terms — not a blanket 10% guess applied to the whole batch. For a deeper look at how computed columns handle multi-step calculations across different document types, see our introduction to computed columns in document extraction.

From Batch Output to Draw Request Package

The spreadsheet you download is not the final deliverable. It's the input to your draw request package — the consolidated payment application you submit to the owner, lender, or architect for approval. Getting from the extracted data to a submission-ready package involves three steps that the batch output makes faster:

1. Sort and subtotal by job. If your batch included invoices from multiple projects, the first view your PM or owner's rep needs is a per-job summary. Since every row carries a Job # column, a single sort groups all invoices by project. A subtotal on Total Billed and Net Due per job gives you the draw totals for each project's payment application — numbers that used to require summing across 30 separate files.

2. Cross-reference against the schedule of values. Each subcontractor's billed amount should align with their line in the project's schedule of values. With all sub data in one sheet, you can VLOOKUP the SOV line amount against the extracted invoice total. Discrepancies stand out immediately — a sub who billed 45% complete when the SOV says 40%, or a sub who forgot to deduct previous payments from their current draw.

3. Build the lien waiver checklist. Every draw package requires conditional lien waivers from each subcontractor receiving payment. The extracted Net Due column tells you exactly how much each sub is owed this period — which is exactly the amount their conditional waiver should reflect. Mismatched waiver amounts are the single most common reason draw packages get rejected. Having verified Net Due figures in one place turns the waiver reconciliation from a full-day task into a side-by-side column comparison.

When One Invoice Fails in the Batch

In a batch of 30, something will go wrong on at least one invoice. A handwritten field the AI reads as $5,800 instead of $3,800. A scanned document where the retainage line is partially obscured. A sub who sent last month's invoice by mistake and the numbers don't match the current draw period. The question isn't whether these happen — it's whether the batch workflow handles them without derailing the entire process.

The practical approach is partial reprocessing, not full restart. The batch output gives you a 30-row spreadsheet. You spot-check dollar fields — Total Billed, Net Due — against the source documents. One row has a discrepancy. You reprocess only that invoice (not the whole batch), copy the corrected row over the bad one, and move on. The other 29 rows stay untouched.

This is the operational advantage of a merged-output batch workflow: errors are isolated to individual rows and can be fixed without affecting anything else. There's no domino effect where re-extracting invoice #17 forces you to re-merge files and rebuild the master sheet. The structure stays intact. One row gets corrected.

Frequently Asked Questions

What happens if a subcontractor submits a revised invoice after I've already run the batch?

Process the revised invoice separately with the same column set — the extraction will match the same fields. Download the single-invoice result, replace that subcontractor's row in your master batch spreadsheet, and re-sort. The batch output structure doesn't need to be rebuilt. This is the same partial reprocessing workflow described above, applied to a late revision rather than an extraction error.

Can I batch process invoices that span multiple projects?

Yes. Include a Job # column in your extraction schema. Each invoice will be assigned the job number found on that document. After extraction, sort the batch output by Job # to group invoices by project, then subtotal per project for your draw package. The column definitions stay the same regardless of how many projects the invoices represent — which is the key to running one batch across your entire active project portfolio instead of one batch per project.

How does batch processing handle mixed file types — PDFs, scans, and phone photos?

The extraction engine handles all of them in the same batch. A subcontractor's emailed PDF, a scanned paper invoice from the job trailer, and a phone photo of a handwritten bill can all be uploaded together. The AI processes each document type with the same column-name matching logic. The only practical consideration is image quality — a blurry phone photo won't extract as cleanly as a crisp PDF, which is why the verification pass focuses on low-confidence documents rather than checking every cell.

Do I need to separate AIA G702 forms from standard subcontractor invoices in the batch?

No. An AIA G702 pay application uses different field labels than a standard invoice — "Total Completed & Stored to Date" instead of "Total Billed," "Less Previous Certificates" instead of "Previous Payments" — but the AI matches the meaning, not the label text. Upload G702s and standard invoices together. The column-name extraction finds the value that corresponds to each column regardless of what the document calls that field. For G702-specific extraction details, see our AIA G702 data extraction guide.

Does this replace the need for a construction ERP?

No, and it doesn't try to. This solves the data-capture layer — getting subcontractor invoice data off the page and into a structured spreadsheet. It doesn't replace your accounting system's approval routing, three-way matching, payment processing, or project-level WIP reporting. For a small to mid-size GC running QuickBooks or Sage 100, the spreadsheet-first batch approach replaces the manual data entry that feeds those systems. For a larger GC on Viewpoint Vista or CMiC, it replaces the step between "invoice received" and "data imported" — which, in many firms, is still a human reading a PDF and typing numbers into an ERP screen.