From Tally export to filed-ready GSTR-2B in one run
A chartered accountancy firm handles end-to-end GST compliance for manufacturing, trading, and distribution clients, and every filing cycle meant rebuilding the same reconciliation by hand in Excel. FireAI built a ten-step pipeline that takes a Tally Day Book export as its only input and produces a validated, reconciled GSTR-2B, with every exception queued and every run traceable.
- 10 steps
- from intake to delivery
- 7
- GSTR-2B sections classified
- <10 min
- per reconciliation run
- Every run
- traceable to its ruleset
Trusted by 200+ orgs to boost business insights.
A chartered accountancy firm handling GST compliance for manufacturing, trading, and distribution clients across India used FireAI to replace manual Excel reconciliation with a ten-step automated pipeline. A Tally Day Book export now becomes a validated, filed-ready GSTR-2B in under ten minutes, with exceptions queued for review and every run traceable to its input files and ruleset version.
Every filing cycle meant rebuilding the same reconciliation by hand
GST reconciliation sounds mechanical: match the invoice in Tally to the invoice in GSTR-2B, check the GSTIN, verify the tax amounts, classify the ITC. In practice, a client with 200 purchase invoices in a month has far more than 200 rows to verify. Each invoice can carry CGST, SGST, or IGST, some fall under reverse charge, some come from ISD distributions or imports, and some carry GSTINs that were entered incorrectly months earlier. The cost of getting any row wrong is a disallowed ITC claim.
The team was accurate, but accuracy at that volume, done entirely by hand, consumed hours that should have gone into advisory and review. Every month started from scratch, with no run history, no audit trail, and no way to trace a filed output back to the exact input file and rules that produced it.
- Tally exports arrived in different formats per client, so every new engagement was a manual mapping exercise
- GSTR-2B classification was tagged row by row in Excel, and one wrong tag misclassified the ITC claim
- GSTIN mismatches between Tally and the portal surfaced late, sometimes only at the point of filing
- RCM journal entries had no automatic path to the correct GSTR-2B section
- Duplicates, negative values, and future-dated entries sat in the same working file as clean data
- No audit trail linked a filed output to the input file and ruleset version that produced it
Tally exports and Excel, rebuilt every month
Each month, the accounts team pulled the Day Book or purchase register from each client's Tally, pasted it into a working spreadsheet, and began classifying by hand.
GSTINs validated by eye
A supplier registered under a different legal name on the portal, or a transposition error in Tally, was only caught when the numbers refused to match, often late in the process.
Section tagging with no ruleset
Each row was tagged to a GSTR-2B section manually, with no versioning and no consistency check, so the same ISD transfer could be classified differently from one month to the next.
Exceptions flagged by colour-coding
Duplicates and anomalies were marked by colouring rows in the working file, which held up until the file grew large and finding every flagged row became its own task.
A ten-step pipeline from Tally export to filed-ready GSTR-2B
The pipeline does not replace the firm's reviewer. It does everything before the reviewer touches the file, so the reviewer is approving or querying, not assembling.
FireAI built an end-to-end automation that takes a Tally Day Book export as its only input and produces a clean, validated, reconciled GSTR-2B output, with every exception surfaced, every rule versioned, and every run traceable.
Intake with schema validation
The client drops the Tally export into a connector folder and a run triggers automatically. Sheet names, columns, data types, and row counts are checked against the expected template first, and a file that fails is rejected with a specific error, so no bad data flows through.
Cleaning, normalisation, and one field map
GSTINs are standardised, dates parsed, invoice formats corrected, and duplicates flagged into a clean canonical table. A fixed Tally-to-GSTR-2B field map feeds both the rules engine and the LLM vocabulary layer, so both share one dictionary.
A versioned business rule engine
Each row is classified into its GSTR-2B section (B2B, B2BA, CDNR, CDNRA, ISD, IMPG, or IMPGSEZ), tax splits are computed, ITC eligibility applied, and reverse charge flagged. All rules live in one versioned ruleset file.
Validation with a separate exceptions queue
GSTIN checksums, cross-footed tax amounts, duplicates, negative values, future-dated entries, and state mismatches are checked, and suspicious rows go to an exceptions queue rather than the output file. The reconciliation matches each Tally transaction against GSTR-2B portal data on date, GSTIN, particulars, and trade legal name.
Review checkpoint, then a traceable delivery
The reviewer sees the populated GSTR-2B template, the KPI dashboard, and the exceptions queue. Approvals go to delivery; rejections route back to the relevant pipeline step with notes. The run log records file hashes, the ruleset version, timestamps, and the approving reviewer.
Why they chose FireAI
- The pipeline reads the Tally export format the firm already uses, with no change to the client-side workflow
- A versioned ruleset means every month's filing traces back to the exact rules that produced it
- The exceptions queue separates clean data from uncertain data before the reviewer opens the file
- The LLM vocabulary layer lets a reviewer ask in plain language which suppliers had GSTIN mismatches this month
What changed when the pipeline went live
Reported as the capability delivered; client volume figures are being verified at the 90-day review.
The most immediate change was time. A reconciliation that previously took a team member the better part of a day to assemble, classify, and validate now runs in minutes, and the reviewer's job becomes judgement rather than assembly. The second change was confidence: because every output ties to a versioned ruleset and a run log, a classification questioned during an audit is answered from the log. The third was consistency. Clients whose Tally data used to produce different-looking reconciliations depending on who did the work now get the same structured output every time.
How the rollout went
Three phases, live within the agreed windowThe firm changed nothing about how it collects data. Clients export their Tally Day Book exactly as before, and FireAI built the intake and schema validation layer to accept that format directly. The pipeline went live with no disruption to any active client's monthly filing schedule.
- 1
Data mapping and schema definition
The Tally Day Book field structure was mapped against the GSTR-2B template, with every transformation documented in one field mapping config that serves both the rules engine and the LLM layer.
- 2
Rules engine build and validation
The rule engine was tested against the firm's actual Tally exports, covering RCM entries, ISD transfers, import invoices, credit notes, and amended invoices, and signed off before going live.
- 3
Dashboard and review workflow
The KPI dashboard and reviewer checkpoint were connected to the pipeline output, with the exceptions queue routing each flag type back to the pipeline step that owns it.
Key takeaways
- GST reconciliation at scale is a process problem, not a data problem; the data already lives in Tally
- Automation changes what the reviewer does: the team went from assembling reconciliations to approving them
- A versioned ruleset makes a filing defensible, because every output traces to the rules that produced it
- Separating the exceptions queue from the output is what makes the output trustworthy
- Once the field map and ruleset exist, onboarding a new client is a configuration step, not a new implementation
Who should consider FireAI?
CA firms reconciling GSTR-2B for five or more clients a month, and finance teams on Tally whose GST work takes more than two days per filing cycle.
Frequently asked questions
Stop rebuilding the same reconciliation every month
If your GST work still runs on Tally exports and Excel, FireAI can turn it into a validated, traceable pipeline where the reviewer approves instead of assembles. Book a demo on your own data.
Want results like Chartered accountancy & GST compliance?
Book a demo