Indian purchase order extraction

A purchase order arrives with a table already in it: line items, HSN or SAC codes, quantities, and prices, plus the vendor’s GSTIN and delivery terms. Retyping that table into a tracking sheet is pure overhead, and it is exactly where PO-to-invoice mismatches get born. Dynamite Docs extracts the PO into line-item rows, PO number, vendor GSTIN, HSN/SAC, quantities, and prices, so matching against invoices and e-way bills starts from clean data.

Try this workflow free or view pricing.

The PO is a table waiting to be retyped

A purchase order’s value lives in its line items, and every line carries HSN/SAC, quantity, and price details that all have to be transcribed correctly.
POs come in every vendor’s layout, Word exports, billing-software PDFs, scans, and each one places the vendor GSTIN and delivery terms differently.
Line-level mismatches are the ones that hurt: a wrong quantity or a missed line surfaces when the invoice arrives and the two documents will not tie out.
The PO is the reference point for the whole purchase: it links to the invoice, the e-way bill, and the delivery record, and that chain only works when the PO number and GSTINs match across documents.

How Dynamite Docs handles Indian purchase orders

01 — Drop the PO batch in

Upload POs from your vendors, Word exports, billing-software PDFs, and scans. Clean digital files parse deterministically; scans are read by the vision model you choose.

02 — Header and lines are inferred

PO number and date, vendor GSTIN, and the line-item table, HSN/SAC, description, quantity, unit price, and totals, are pulled from each layout automatically. No templates to build per vendor.

03 — Review line-level confidence

Confidence scores flag the lines worth a glance, usually a quantity or a price on a dense PO. Corrections become patterns, so the next PO from the same vendor extracts cleanly.

04 — Export for matching

Rows export to Excel, CSV, or JSON, or push to Google Sheets, ready to be matched against invoices and e-way bills in your three-way matching workflow.

Document types this workflow handles

Purchase order PDFs — billing-software exports parsed deterministically
Word / email PO attachments — text extracted from the file
Scanned or photographed POs — read by the vision model you choose
PO batches with invoices — matched against the invoices they generate

What gets extracted from an Indian purchase order

Inferred per layout. A typical PO run yields the header and line-item fields, each confidence-scored:

PO number — PO-2026-0417
PO date — 2026-07-30
Vendor GSTIN — 27AABCS1429B1ZP
Vendor — Kaveri Engineering Co.
Line item — Rotary hammer — 16 mm
HSN / SAC code — 8467
Quantity — 6
Unit price — ₹31,000
Line total — ₹1,86,000
Delivery terms — Delivered to Pune, freight prepaid

A realistic example

A purchase order extracts to line-item rows like these, ready for matching against the invoice and e-way bill when they arrive:

PO #	Date	Vendor	GSTIN	Line item	HSN/SAC	Qty	Unit price	Line total
PO-2026-0417	2026-07-30	Kaveri Engineering Co.	33AABCC3456C1ZD	Rotary hammer — 16 mm	8467	6	₹31,000	₹1,86,000
PO-2026-0417	2026-07-30	Kaveri Engineering Co.	33AABCC3456C1ZD	Chisel set	8207	12	₹1,250	₹15,000
PO-2026-0418	2026-07-31	Shri Balaji Traders	27AABCS1429B1ZP	A4 copier paper (ream)	4820	40	₹600	₹24,000

The anchor of the purchase chain

The PO is the anchor of the purchase chain. It sets out what was ordered and on what terms, the vendor GSTIN, the line items with their HSN or SAC codes, quantities and prices, and the delivery details, and it is the document the invoice and the e-way bill are both measured against. The whole chain holds only if the PO number and the GSTINs match across every document, which is why the header fields matter as much as the line-item table.

Line-item extraction is where the value sits. A PO with forty lines is a table that would otherwise be retyped, and a mistake in any single quantity or price becomes a mismatch the moment the invoice arrives. Because the line-item table is inferred from the layout rather than matched against a template, a PO from a new vendor still extracts as rows, HSN/SAC, description, quantity, unit price, and line total in one pass.

Three-way matching is the payoff. When the invoice and the e-way bill are extracted into the same workspace, the PO number ties the documents together and the line totals line up for comparison. The review narrows to the differences that matter, a quantity that changed, a price that does not match the agreed terms, instead of a pile of PDFs checked by eye.

Corrections compound across vendors. The first PO from a new supplier takes a glance at the line items; once the layout is remembered, the same vendor’s POs extract cleanly and the review shrinks to the fields that genuinely changed. For a purchasing desk that raises POs against several suppliers a day, that is the difference between keying tables and reviewing rows.

Why bring-your-own-AI matters for purchase orders

POs carry pricing and vendor details that belong to your business, not to a third-party training set. Bring your own keys, lock vendor data to no-training providers, or run fully local with Ollama when the terms should not leave your machines.

Clean digital POs, including most billing-software exports, parse deterministically for free, identical results every run. Scans and photos are read by the vision model you pick, so you control quality and per-page cost on the image-heavy part of the work.

The line-item table is the value: HSN/SAC, description, quantity, unit price, and line total come out per line, so the three-way match against invoices and e-way bills runs on rows. Vendor-layout corrections compound, and the extracted items stay yours.

Questions, answered

Does it extract the vendor GSTIN from the PO?

Yes. The vendor’s GSTIN comes out as a confidence-scored header field, alongside PO number, date, and delivery details, so the PO can be matched against the invoice and e-way bill for the same supplier.

Does it extract HSN/SAC codes and quantities per line?

Yes. Each line item is extracted with its HSN or SAC code, description, quantity, unit price, and line total, ready to compare against the invoice line by line.

Can it handle scanned or photographed POs?

Yes. Scans and photos are read by the vision model you choose, with the same header and line-item schema inferred as for clean digital files.

Does it do three-way matching for me?

No. Dynamite Docs extracts and structures the documents; matching decisions remain with your team. By keeping PO, invoice, and e-way bill rows in the same workspace, it gives you the data to match on, not the decision.