Reads the formats clients actually send
Digital PDFs, scanned copies, photographed bills, e-invoices with IRN and QR code, and Excel or CSV sales and purchase registers.
Practice Engine / Document Intake & Coding
Every invoice read once, coded the way that client is always coded.
PDFs, scanned bills, phone photos, forwarded emails and spreadsheet exports come in. Structured entries go out, mapped to each client’s own chart of accounts, with a confidence score beside every field.
Intake · Lahartara Motors- MHS-1452.pdf Read
- Photo of fuel bill Read
- Sales register Sep.xlsx Reading
- Forwarded email, 3 PDFs Queued
SupplierMahadev Hardware Stores99%
GSTIN09AAGFM3318L1ZK98%
Invoice no.MHS/145299%
HSN731891%
Taxable₹4,860.0099%
AccountConsumables72%
Low confidence on account. Sent to review with two suggestions.
Capabilities
What Document Intake & Coding handles
Codes into each client’s chart
Entries map to the ledgers that client already uses, not a generic template. A trader’s purchase ledger and a clinic’s consumables ledger stay distinct.
Learns per client, not per firm
When a reviewer corrects a coding, the rule is remembered for that client and that supplier. It is not silently applied to anyone else.
Confidence on every field
Supplier, GSTIN, invoice number, HSN or SAC, taxable value and tax heads each carry a score. Low scores go to review instead of into the books.
Duplicate and sequence checks
The same invoice uploaded twice, or forwarded by both the client and their accountant, is caught before it is posted twice.
Line-level GST split
Mixed-rate invoices are split by line, with CGST and SGST or IGST worked out from the place of supply and checked against the invoice totals.
Workflow
From upload to a coded entry
- 01
Arrives
Through the client portal, a forwarding address for the client, or a bulk upload by your staff.
- 02
Read
Each document is classified (invoice, credit note, debit note, expense bill) and its fields extracted.
- 03
Coded
Ledger, party, tax heads and narration are proposed using that client’s history.
- 04
Routed
High-confidence entries join the review batch. Anything uncertain is held with the reason shown.
In detail
What comes in, and what is read from it
Where the professional stays in charge
Coding is a proposal. Nothing is posted as final until someone at your firm has reviewed the batch. Corrections you make become rules for that client; the engine does not decide policy on its own.
| Document | Typical source | Fields read |
|---|---|---|
| Tax invoice (B2B purchase) | PDF from supplier, email forward | Supplier, GSTIN, invoice no. and date, HSN, taxable value, tax heads, total |
| E-invoice | PDF with IRN and signed QR | All of the above plus IRN and acknowledgement details |
| Expense bill | Phone photo, scan | Vendor, date, amount, tax where shown, nature of expense |
| Credit and debit notes | PDF, scan | Original invoice reference, reason, tax reversal |
| Sales and purchase registers | Excel or CSV export | Row-level entries mapped by column |
Questions
Asked by firms evaluating this module
What happens to a document the engine cannot read well?
It is held in review with the unreadable fields marked. Your staff can key the missing values or request a clearer copy from the client through the portal in one step.
Does it use our chart of accounts or its own?
Yours, per client. During onboarding the existing ledger list for each client is imported, and coding proposals only ever point to ledgers that exist for that client.
How does the learning work if two staff code differently?
The latest reviewed decision for a client and supplier pair becomes the proposal next time. Conflicting corrections are surfaced to the reviewer instead of being averaged.
Can it handle handwritten bills?
Clear handwritten amounts and dates are often readable, but confidence is usually lower and most will route to review. The system says so rather than guessing.
Practice Engine
Works with the rest of the engine
Review slip · awaiting approval
Your client list, prepared and waiting for one signature.
Bring a month of real documents from two or three clients. We will run them through the engine with you and show you the review queue they produce.