For CA firms running 10+ statutory audits a year
Every finding, traceable to source.
Lekhya tests purchases, revenue and bank transactions across the full population, links every exception to the document and page behind it, and builds the working papers as the work is done — Tally exports, GST data and scanned invoices included.
Open a sample engagement and work through it yourself — no account, nothing to install. Or take thirty minutes with a Chartered Accountant, not a sales representative.
Illustrative product interface showing an audit engagement with sample data
Areas
Revenue
2,417 transactions tested8 exceptionsPotential revenue cut-off issue
₹18,40,000
- Invoice date
- 29 Mar 2026
- Revenue entry
- 29 Mar 2026
- Dispatch date
- 03 Apr 2026
Goods were dispatched after the year end. Revenue appears to have been recognised in FY 2025–26 on an invoice raised before dispatch. Confirm the terms of transfer of control before concluding.
Evidence
- Invoice INV-2026-1184Page 1
- Dispatch document DC-4471Page 1
- General LedgerRow 18,921
Drafted · unreviewed · awaiting a named auditor
Illustrative interface with sample data for a fictional client. Select an area or an exception to explore.
The work
The work is repetitive. The deadline is not.
Between September and October, your seniors are not exercising judgment. They are matching invoices to goods receipts, ticking a sample of twenty-five against the ledger, chasing the same six documents from the same client, and rebuilding working papers the week before filing because nobody had time to write them as they went.
You test a fraction
A sample of 25 from a population of 1,200 leaves 98% of transactions unexamined — not because you chose to, but because there were not enough hours.
The file is reconstructed, not recorded
Working papers assembled after the fact are harder to defend under review and slower for a partner to sign.
You turn work away
Capacity in October is the binding constraint on what the firm can accept in June.
The shift
Lekhya performs the repetitive work. Your auditors perform the judgment.
The data comes in from Tally, the GST portal and the client’s documents. The testing runs across every transaction in the period, not a sample. What comes back is a queue of exceptions, each already linked to its evidence, waiting for a human to decide what it means.
Computed, not generated
Matching, footing, casting, reconciliation, variance and duplicate detection are deterministic software. They produce the same answer every time.
Read by a model
Extracting fields from a scanned invoice, classifying a document, drafting the narrative of a working paper — each marked as drafted, each awaiting review.
The model never decides a total. Arithmetic is the one thing software has always been able to do exactly, and we have not handed it to a language model.
How it works
Four stages, and you can inspect all of them.
- 01
Ingest
Connect the Tally company file or upload the export. Pull GSTR-2B. Add invoices, goods receipts and bank statements — scanned or digital. Lekhya reads them and tells you which ones it could not.
- 02
Structure
Ledgers mapped, vendors normalised across spelling variations, documents classified and fields extracted. You review the mapping once; it carries forward next year.
- 03
Test
Procedures run across the full population. Matching, reconciliation, duplicate detection, cut-off, variance and missing-document checks.
- 04
Review
Exceptions arrive in a queue. You accept, reject or annotate each one. The working paper is assembled as you go, and your sign-off is recorded against your name.
Traceability
Click any finding and follow it all the way down.
Most tools give you a number and ask you to trust it. Lekhya gives you the chain. A ₹18,40,000 cut-off exception expands into the procedure that raised it, the transaction it tested, the field extracted, the document that field came from, and the page it sits on — highlighted.
This is what makes the work defensible. When a reviewer asks where a figure came from, nobody reconstructs anything. When a peer reviewer asks the same question next year, the chain is still intact, and it exports with the file.
Finding
Potential revenue cut-off issue
₹18,40,000 recognised in FY 2025–26
Raised automatically when dispatch evidence dated after the year end was matched to an invoice recorded before it.
Procedure
R-04 · Revenue cut-off testing
Full population · 2,417 transactions
Compares invoice date, revenue posting date and dispatch date for every sales transaction within fifteen days of the year end.
Transaction
INV-2026-1184 · Bharat Components Ltd
Invoice 29 Mar 2026 · ₹18,40,000
Matched to ledger posting in the sales account on 29 March, and to dispatch document DC-4471 dated 3 April.
Extracted field
dispatch_date = 2026-04-03
Extracted from a scanned PDF · confidence: needs review
Read from the dispatch document by the extraction model and flagged for confirmation because the source is a scan rather than structured data.
Source document
DC-4471 · Dispatch document
Page 1 · received from client 12 May 2026
The page opens with the dispatch date highlighted in place. The chain is preserved on export and remains intact at next year's rollforward.
Step 1 of 5
Automated testing
Named procedures, not “AI-powered insights.”
Every procedure has a name, a defined population and a documented result. Nothing runs that you cannot describe to a reviewer.
On purchases
- Three-way matching across purchase order, goods receipt and invoice
- Invoice-to-ledger agreement, line by line
- Duplicate invoice detection across vendor, amount and date variants
- Missing goods receipt or missing approval
- Price and quantity variance against the order
- Cut-off testing either side of the year end
- Round-sum and just-below-threshold entries
- Payments without a matching invoice
On GST
GSTR-2B against the purchase register against the general ledger — with differences categorised rather than dumped as one list:
- TimingInvoice booked in a different period to the supplier's filing
- Supplier non-filingCredit claimed against an invoice not appearing in 2B
- Ineligible creditBlocked under section 17(5) but claimed in the books
- Rate mismatchTax rate differs between the invoice and the register
- GSTIN errorCredit recorded against an incorrect registration
Every procedure in both modules, and what happens when the client’s data is incomplete: the full product.
Risk
Test the whole population, then sample for judgment — not for capacity.
Sampling exists because humans cannot read 1,200 invoices. Software can. Lekhya runs the deterministic checks across every transaction in the period, which changes what sampling is for: you still select items for judgment-based procedures, but you are no longer using a sample of 25 to stand in for work you did not have time to do.
The practical difference is that the unusual item — the duplicate paid twice in March, the invoice with no goods receipt, the vendor who appears once at a suspiciously round figure — is found because it was examined, not because it happened to fall into the sample.
Coverage, drawn to scale
Manual sample — 25 of 1,284 transactions
Roughly 2% examined. The remainder is untested.
Lekhya — 1,284 of 1,284 transactions
Every item examined. Amber marks the exceptions raised for review.
Working papers
The file is built while the work happens.
Every procedure Lekhya runs produces a working paper in your firm’s format, with tickmarks, cross-references, the population tested, the exceptions found and their resolution. The evidence links are live in the application and survive as page references when you export to PDF.
Preparer, reviewer and partner sign-offs are recorded against named users with timestamps. The full audit trail — every action, every change, every prior value — exports with the engagement file for peer review.
- Population
- 1,284
- Tested
- 1,284
- Exceptions
- 37
- Unresolved
- 12
| Ref | Vendor | Variance | Tick | Status |
|---|---|---|---|---|
| PT-0118 | Shreeji Traders | — | ✓ | Signed off |
| PT-0119 | Nandan Steel Co. | 10,800.00 | ⱡ | Awaiting review |
| PT-0120 | Kirti Enterprises | — | ◊ | Evidence missing |
Prepared by
A. Kulkarni · 14 May
Reviewed by
S. Rao · 17 May
Partner
Awaiting sign-off
Illustrative working paper with sample data.
Security
Your client’s data stays in India.
Hosted in India
All client data is stored and processed in Indian regions.
Never used for training
Your engagements do not train any model, ours or a third party's.
Access is scoped
Staff see only the engagements they are assigned to, and evidence links never grant access a user does not already have.
Everything is logged
Every action, with user, timestamp and prior value — exportable with the file.
We publish exactly which certifications we hold and which we do not. Formal third-party certifications — SOC 2, ISO 27001, ISO 42001 — are not currently claimed. We make no certification claims on this site until an audit is complete and the report can be produced on request.
Who it’s for
Built for firms where the partner still reviews the file.
5–15 audit staff
You run 20 to 60 statutory audits a year, mostly on Tally, mostly against the same two deadlines. Lekhya removes the vouching and matching so your seniors spend October reviewing rather than ticking.
15–50 audit staff
You have methodology, templates and a review hierarchy that work. Lekhya runs testing underneath them and produces papers in your format — with an audit trail built for peer review from the start.
Not for everyone. If you run fewer than ten audits a year, the arithmetic will not work for you yet, and we will tell you that on the call rather than after the invoice.
Rather than a quote
We have no pilot outcome to publish yet, so open the file instead.
A worked revenue cut-off procedure on a sample engagement — the population tested, the exception it raised, the invoice and dispatch document behind it, and the working paper it produced. Illustrative sample data, not a testimonial.
Start with one engagement you have already completed.
Give us a statutory audit you finished last year. We will run purchase testing against it and show you every exception we find — so you can compare our results against what your team actually found, on work whose answers you already know. Two weeks, fixed scope, fixed fee.
Thirty minutes to start. You will speak to a Chartered Accountant, not a sales representative.