PaperDataReads
Companies, groups, the ledger master and the daybook, month by month, over Tally's own XML interface. Read-only, always.
TDS errors, caught
before you file
A manual scrutiny samples. This one reads every voucher for the year.
PaperData TDS Audit · Q2 FY 2025-26What this is
A full year of vouchers, read the way a partner would read them.
A TDS scrutiny means reading the books entry by entry, deciding what attracted TDS and under which section. On a decent sized client that is two or three days of work, and nobody has time to read every voucher, so it gets sampled. The exposure hides in the entries nobody reached.
PaperData TDS connects to Tally and pulls the raw daybook itself: every voucher for the year, 100% of them. It uses AI to judge what each payment actually was and which section applies, then checks what was really deducted. It is not a filing utility, and it is not a calculator you feed a prepared list to.
Judges· AI
Which ledgers can attract TDS at all, and what a payment really was, decided by AI from the ledger names and voucher narrations rather than a keyword list.
Reports
Required against deducted, entry by entry, with the shortfall in rupees and the reasoning behind every call written into the export.
Why this is not another TDS tool
Other tools check the TDS you already found. This one finds the TDS you missed.
Every other tool in this category starts from a list your team has already built by hand. Building that list, deciding which expenses attract TDS and under which section, is the actual work. That is the part this does for you. It also does not depend on the books being tidy. Where there is no clean TDS ledger, or the deduction sits scattered across entries, it still works out what was actually deducted.
Conventional TDS software
- You hand it a list of parties and payments you have already identified as liable.
- It computes TDS on the entries you gave it, and helps you file the return.
- Deciding which expenses attract TDS, and under which section, is still manual.
- Anything the ledger scrutiny missed is invisible to it, because it was never in the list.
- It reconciles against the TDS ledger in the books, so where the books have no clean TDS ledger it has nothing to work from.
PaperData TDS
- It connects to Tally and pulls the raw daybook itself: every voucher for the year, unfiltered.
- It works out which ledgers can attract TDS at all, and which parties behind them are liable.
- It determines the applicable section per entry, then checks what was actually deducted.
- The entries nobody spotted are exactly the ones it is built to surface.
- It establishes what was actually deducted even where no dedicated TDS ledger exists, or the deduction sits scattered across entries.
The scrutiny
What a partner does over three days, in one pass.
This is the ledger scrutiny itself, not a calculator bolted onto the end of it. Steps marked AI are where judgement is applied to language rather than numbers.
- 01
Pull the raw books
Company list, group hierarchy, the full ledger master with PAN and GSTIN, and the daybook fetched month by month and de-duplicated by voucher GUID. No export, no template, no prepared list, just the same data a partner would open Tally to read.
- 02
Narrow to what can attract TDSAI
Bank, GST, income-tax, stock, equity and round-off ledgers are ruled out, and every remaining expense ledger name is judged individually by AI to decide whether it belongs in a TDS audit at all. An exempt catalogue handles the ledgers that never attract deduction.
- 03
Work out which parties are liable
Nothing is supplied. The creditors and parties behind the entries are identified from the books themselves. GSTIN structure, then PAN structure, then the name determine whether each is an individual or a company, which sets the rate. Parties with no PAN are separated for 206AA.
- 04
Decide the section, then check the reasoningAI
The rule engine derives a section from the nature of the expense. That call is then reviewed by AI against the ledger names, the TDS ledger used and the actual voucher narrations, and can be overridden. Every override is written into the export with its confidence and the reason for it, so you can audit the machine.
- 05
Track thresholds across the whole year
Aggregates run per party for the full period, not per invoice, with dated rates, the shared 194I pot, and 194Q's excess-only basis. A vendor who crosses ₹1,00,000 in the eleventh month is caught in the eleventh month.
- 06
Compare against what was deducted
Required versus deducted, entry by entry, resolving to missing, short, excess or in order, with the shortfall in rupees and a remark explaining the verdict. What was actually deducted is established from the books themselves, including where there is no dedicated TDS ledger, or the deduction sits scattered across entries and settlements rather than sitting under one head.
Every classification the AI overrides is written into the Excel export with its confidence and its stated reason, so the judgement is reviewable rather than taken on faith.
Where the judgement is
The arithmetic is the easy part.
Applying 10% to ₹58,120 is not what takes a firm three days. Working out that a ledger called “Professional Charges” was actually a works contract, and that the next entry under the same head genuinely was professional fees, is what takes three days. That is a reading problem, and it is exactly where the AI is pointed.
“Professional Charges”
Ledger name in the books
Keyword
Matches on the word professional. Books it to 194J at 10%.
Read it
The narration describes fabrication and installation at a plant. That is a works contract under 194C at 2%, and the classifier is given the narration precisely so it can say so.
At stake
Eight percentage points, on every entry under that head.
“Insurance”
Ledger name in the books
Keyword
Reads as an exempt payment and is dropped from the audit.
Read it
Premium paid attracts nothing; commission received is 194D. Identical ledger names, opposite treatment. Separating them needs the surrounding context, not the label.
At stake
An entire section missed, or a clean payment wrongly flagged.
“Misc. Exp / Delhi Office”
Ledger name in the books
Keyword
Matches no keyword list anywhere. Silently skipped.
Read it
Every expense ledger name is judged on meaning rather than checked against a fixed vocabulary, because no two firms name their ledgers the same way.
At stake
The entries nobody thought to look at.
Each of those calls is recorded: what the rule engine derived, what the AI classifier concluded, how confident it was, and why. You are reviewing a judgement, not accepting an output.
What it catches
Four ways a TDS entry goes wrong.
Every flag carries the section it was tested against, the rate that should have applied, and the shortfall in rupees.
Missing
No deduction where one was due
The threshold was crossed during the year and nothing was deducted. Aggregates are tracked per party across the whole year, not per invoice, so a vendor who creeps past ₹1,00,000 in the eleventh month still gets caught.
Short
Deducted, but not enough
2% applied where 10% was due, or the wrong payee type assumed. Required TDS is recomputed per entry from the section rule, the rate in force on that date, and whether the party is an individual or a company.
Wrong section
Booked under the wrong head
Professional fees sitting under 194C. The section implied by the TDS ledger is compared against the section the rule engine derives from the expense, and any mismatch is flagged with both.
206AA
Payments without a PAN
Parties with no PAN on the ledger master are flagged, with the higher 206AA rate shown alongside what was actually applied. The report carries a dedicated 206AA chapter.
Coverage
Twenty sections, and the awkward parts of them.
You know the rates. What is worth saying is how the edges are handled: a rate that changed mid-year, two sections drawing on one threshold, and a section that charges only on the excess.
- 194CContractors
- 194JProfessional & technical
- 194QPurchase of goods
- 194I(a)Rent (plant & machinery)
- 194I(b)Rent (land & building)
- 194HCommission & brokerage
- 194AInterest
- 194DInsurance commission
- 194RBenefits & perquisites
- 194TPartner payments
- 194MCertain individual/HUF payments
- 194NCash withdrawals
- 194OE-commerce participants
- 194SVirtual digital assets
- 194IAImmovable property
- 194IBRent by individual/HUF
- 194ICJoint development
- 195Non-residents
- 206C(1H)TCS on sale
- 192Salary (identified, not computed)
Rates that changed mid-year
Rules carry dated rate and threshold periods, so an entry from April is measured against the rate in force in April, not today's. 194H moving from 5% to 2% is handled as a date range, not a flag day.
The 194I dual pot
194I(a) and 194I(b) share one threshold. The Finance Act 2025 treatment is implemented as both a ₹6,00,000 full-year cumulative pot and a ₹50,000 calendar-month pot, applied together.
194Q on the excess only
Where most sections charge on the full amount once the threshold is crossed, 194Q charges only on the excess over ₹50,00,000. That is computed separately rather than bolted onto the standard path.
Front-loaded on crossing
When a party crosses an annual threshold, the full cumulative requirement lands on the crossing entry rather than being spread, which is what the department expects to see.
What you get out
A working paper, not a black box.
Every figure on screen traces back to a voucher, and every export carries the reasoning with it.
- Every voucher for the year checked, not a sample. Ten thousand entries or more in one pass.
- Required against deducted, entry by entry, with the shortfall stated in rupees.
- Each entry gets a verdict: missing, short, excess or in order, with the section and rate it was tested against.
- Exposure under 40(a)(ia) quantified, so the cost of leaving it unfixed is a number the partner can see.
- Every figure traces to its voucher, with the reasoning exported alongside, so the reviewer approves a judgement rather than accepts an output.
Delivered as
Excel workbook
29 columns per entry: base amount, rule-derived section, rate, TDS required, deducted, shortfall, running balance, status and remark, plus a summary sheet. Per party or all parties at once.
PDF report
Executive summary, section-wise chapters, a 206AA chapter, and a methodology and disclaimers section. Carries your firm name and logo as letterhead.
CSV
The same entry-level detail as the workbook, for anyone who wants it in their own tooling.
Your client's data
Where the data goes.
The honest version, because this is the question that decides whether a firm installs anything at all.
The app runs on your PC
PaperData TDS is a desktop application. The audit engine, your voucher data, and every report it produces stay on the machine you installed it on.
Tally access is read-only
It reads companies, groups, the ledger master and the daybook over Tally's own XML interface on localhost. It does not write to Tally, alter vouchers, or post entries. Corrections are yours to make.
Classification uses hosted AI
Deciding what a payment was is done by hosted models from OpenAI and Anthropic, and it is part of how the audit works rather than an optional extra. Expense ledger names, party names and voucher narrations are sent for that judgement. Amounts are not part of the classification request.
The assistant sees more
When you ask the in-app assistant a question about an audit, it is given entry-level detail (party, voucher, section, status and the amounts) so it can answer. That is the only step where figures leave the machine, and only for the questions you ask.
Who built this
PaperData is built by Vishal Gupta, a graduate of IIT Delhi. The TDS engine was developed against live client books, with practising Chartered Accountants reviewing the classifications it makes, and it continues to be shaped by that feedback.
We are not tax advisors. The tool flags and explains. The CA decides.
See it on your books
Run it on a client's books this quarter.
Install it on the PC that runs Tally and run your first audit the same afternoon, or let us walk through a real audit on your data first. The installer is per-user, so it needs no administrator rights on a locked-down firm machine.
- A Windows PC
- Tally Prime or Tally.ERP 9, running
- Tally's XML/HTTP port (9000) enabled
- A licence key
PaperData-TDS-Setup.exe · 41 MB · Windows · v1.0.129
TDS is the first compliance we automated. ITR preparation, GST reconciliation and tax audit are what we are building next, on the same review-and-approve workflow. They will appear here when they work, and not before.