A manufacturer I worked with runs plants in four states. Every one of those plants holds its own GSTIN, because that’s how GST works, one PAN, but a separate registration for every state where the business has a warehouse, office, or fixed place of operation. An invoice that lands in their AP inbox isn’t just a bill to be read and paid. It’s a document that has to be matched to the correct state registration, checked for the right place-of-supply treatment, and validated before it can ever touch their input tax credit.
OCR doesn’t know any of that. It reads text. It doesn’t know that using the wrong state’s GSTIN on an invoice risks penalties and denial of ITC, or that a transfer between two branches in different states is itself a taxable supply that needs its own invoice and IGST treatment. That’s the gap “touchless” processing is actually meant to close, and it’s also where a lot of AP automation quietly stops short.
Why OCR was never the hard part
Extracting a vendor name, an amount, and a GSTIN off a scanned invoice is table stakes at this point. The harder problem for a multi-state enterprise is context: which registration does this invoice belong to, is that GSTIN actually valid for the state the goods or services were supplied in, and does the invoice even qualify for straight-through processing once GST rules are applied, not just OCR confidence.
Since e-invoicing became mandatory for every business with turnover above ₹5 crore, that context has real teeth. An invoice without a valid IRN can’t be used to claim input tax credit at all. And on the IRP side, the validation isn’t cosmetic, it checks the GSTIN, invoice uniqueness, HSN codes, and tax calculations before an IRN is even issued. A single mismatched digit in a supplier’s GSTIN, or an invoice logged against the wrong state entity, can silently break ITC eligibility for that transaction. Nobody catches that by reading the PDF faster.
What "touchless" actually has to check for a multi-state business
For a company holding GSTINs across several states, real touchless processing means the system is doing work no OCR tool was built to do:
- Matching each incoming invoice to the correct GSTIN and state registration, not just detecting that a GSTIN exists.
- Flagging place-of-supply mismatches before they become ITC denials.
- Recognising inter-branch transfers as taxable supplies requiring their own invoicing and IGST treatment, rather than processing them like a routine vendor bill.
- Confirming the invoice carries a valid IRN before it’s eligible for autonomous processing at all
Each state registration also files its own returns. So the same invoice that gets processed touchless also has to land in the correct state-wise reconciliation, GSTR-1, GSTR-3B, ITC ledgers, without someone manually sorting invoices by state after the fact.
The CFO question worth asking
If your AP platform can extract fields from an invoice but can’t tell you whether that invoice belongs against the Maharashtra registration or the Gujarat one, it isn’t touchless, it’s fast data entry with an ITC risk built in. For a business with three GSTINs, that risk is manageable with a sharp finance team. For a business with fifteen, across a footprint most Indian enterprises are now operating at, it stops being something a human can catch invoice by invoice, and starts being something the system has to be built to catch by design.
That’s the actual bar for touchless processing in India: not whether it reads the invoice, but whether it understands which entity, which state, and which return that invoice belongs to before a human ever sees it.
