Tally automation with AI — bills in, vouchers booked.
AI reads the bill, proposes the voucher, drops a draft in Tally, and your accountant approves. We do not replace Tally. We stop the retyping.
The bill arrives as a PDF, a photo, or a WhatsApp forward. Someone in accounts retypes it into Tally — party, GSTIN, HSN, qty, rate — and hopes the challan matches. At month-end the pile is still there.
Where the hours actually go
PDF, photo and WhatsApp bills. Challan mismatch. GST typos. A month-end backlog because entry never kept up with the inbox.
Capture → Read → Match → Book draft → Approve
The bill as it already arrives. AI pulls the fields. Challan and party against your books. A voucher waiting in Tally. Nothing is live until someone who knows the ledger says yes.
Munshi — productised Tally bill booking
Munshi is ZeroOne D.O.T.S AI's productised version of this workflow. It reads bills and vouchers and books them into Tally for approval, private on hardware you own. Munshi on the products page → · Why it stays private →
Who this is for
| Kind of MSME | What the pile looks like | What Munshi takes first |
|---|---|---|
| Textile / job-work | Dyeing, printing, embroidery bills against grey challans | Job-work purchase bills and GST lines |
| Trading | Purchase bills on WhatsApp, sales already in Tally | Inward bills + party match |
| Manufacturing | RM bills, job-work out, freight | One vendor or one bill type, then expand |
| Services | Vendor invoices and retainers retyped every month | Recurring vendor bills |
Textile units in Surat see this most loudly — we wrote that up as five workflows that actually pay.
One bill type, then the next
Pick one vendor or one inward series. Run samples with your accountant in the room. Write the rules. Draft + approve. Then expand. Ledgers stay on-prem — same rule as owning the model.
Can it read Gujarati or Hindi stamps and bill text?
Yes. A large share of the bills we see from Gujarat traders and job-workers have Gujarati or Hindi stamps, item names and notes. Those stay in the draft for your accountant — they are not skipped just because they are not in English.
What if the AI misreads a bill?
Misreads stay a draft. Nothing posts to Tally until your accountant approves. The point is to stop retyping, not to remove the person who knows when a GST rate looks wrong.
Do I need new hardware?
Usually no. Ordinary office machines are enough for bill reading and draft booking. The workflow is private by default — ledgers stay on-prem.
Is this only for textile units?
No. Textile is where we see a lot of job-work bills, but trading, manufacturing and services offices have the same pile: PDF bills, WhatsApp photos, challans that do not match, and a month-end backlog.