Executive Summary

A fully automated AI-driven accounting engine that converts raw excel bank statements into:

✔ Clean and Structured ledgers and auto-classified Income & Expense Heads and creation of new ledgers.

✔ 44AD Computation Sheet (Comparison between Old and New Regime) in single view.

✔ Auto-prepared ITR-4 figures.

✔ Exportable Trading and Profit Loss Account and Balance Sheet PDFs and Excel Files.

The system removes 90% of the manual effort traditionally required for presumptive taxation cases under Section 44AD by automating bank narration interpretation, ledger grouping and tax computation.

Built using Python, AI models, Streamlit and Excel automation, it provides a smart, accurate and scalable DIY tax engine for Chartered Accountants.

Industry Context and Background

A majority of India’s MSME taxpayers file ITR-4 (44AD Presumptive Taxation). Out of 7.28 Cr ITR Filed in AY 2024-25, 1.88 CR ITR were filed under Presumptive Taxation Scheme (25.77% of Total ITR Filed for the year)

Despite its simplicity, CAs still spend significant time on:

1. Cleaning Raw Bank Statements and understanding Cryptic Narration Codes different for every bank.
2. Deriving Turnover from Bank Flows.
3. Preparing Final Accounts for compliance purposes.
4. Computing Presumptive Income Manually for Computation of Tax.

BUT no enterprise-grade tool exists that can:

🔹 Parse Bank statements in excel format of any bank (ICICI, BOB, HDFC, Axis, Kotak, SBI etc.) to creation of ledgers, mapping of ledgers and compute tax in a single go from start to end.

🔹 Generate financials instantly in single window for view of calculation of Net Profit and JV Adjustments.

🔹 Affordable for small CA firms without using two three different software’s and lot of manual efforts.

Most CAs still rely on Excel + manual review, which is slow, time consuming, error-prone and not scalable.

PART A - Problem Statement

“Bank statements contain the complete financial story — but extracting accounting meaning from them requires hours of manual work for computation of tax”

Key challenges:

1. No standard format across banks as narrations are messy and cryptic (UPI/IMPS/ACH/ECOM codes).
2. Personal vs business segregation is manual and every bank uses their own formats of Narrations.
3. Tally import requires clean ledgers and Ledgers must be created from scratch and manual accounting wastes time.
4. Turnover for 44AD must be reconstructed with manual punching of figures.
5. Compliance deadlines are strict and data to be processed needs a lot of time against the revenue generated for filing presumptive returns.

This leads to:

1. Hours wasted per client
2. Errors in turnover classification (especially UPI/IMPS credits)
3. Inconsistent accounting treatment
4. Difficulty scaling during peak season

Why Existing Solutions Fail

❌ Generic bank statement converters

Do NOT understand Indian taxation requirements, GST or CA-specific requirements.

❌ Tally import tools

Only map data—they do not compute tax or generate ITR-4 numbers. For that we need another software with manual plug in of figures.

❌ Not suitable for confidential financial data

Many solutions are cloud-based → privacy concerns for CA practice.

PART B - AI-Powered Solution Overview

A completely automated pipeline:

Bank Statement → Structured Dataset

Reads Excel/ CSV from all major banks and cleans date formats, debit/credit detection, and remarks.

AI Narration Classifier

Understands Indian narration patterns using rule-based + ML hybrid model:

UPI ID patterns, IMPS structure, POS/ECOM codes, Loan/EMI, Cash deposits/withdrawals

Ledger Mapping Engine

Auto-suggests ledger for each transaction based on predefined rules and option to change the mapping to the ledgers and helps in creation of new ledgers

Automated Final Accounts Generator

Creates:

1. Turnover based on bank transactions taking reference of Opening Trial Balance figures and bank transactions.
2. Calculation of Presumptive Income and Profit under 44AD (@ 6% / 8%).
3. Passing of Expense and Income Entries using JV Module and creating new ledgers.
4. Capital account movement of Home Loan, Interest Income, Capital Gains and other items of income.
5. Sundry balances of Final Accounts to be reported in ITR-4 and Cash/bank reconciliations.

ITR-4 Ready Output

Generates:

1. Profit u/s 44AD taking into account the Opening Balances, Bank Transactions and JV Entries.
2. Gross receipts and Balance Sheet Figures (ITR-4 Schedule BP).
3. Comparison of Tax as per Old Regime and New Regime taking into account the Finance Act of the relevant Assessment Years.

Output Formats

1. Excel and PDF downloadable Final Accounts

Why This Tool is Different

1. Built by a CA for CAs
2. Uses Indian narration intelligence
3. Fully offline (privacy safe)
4. Zero recurring cost
5. Customizable for any client/bank
6. AI + Rule-based hybrid → Highest accuracy
7. Includes 44AD Tax Computation + BS + P&L
8. First-of-its-kind: “Bank → Tax” pipeline in India

PART C -Technologies Used

Programming & Libraries

1. Python (Vibe Coding using Cursor.AI)
2. pandas
3. numpy
4. openpyxl
5. reportlab (PDF generation)
6. regex & rule-based narration parser (based on patterns identified from Chat GPT)
7. Streamlit (Interactive UI)

AI Layer

1. Prompt-engineered models (Chat GPT) for narration understanding
2. Reinforcement Learning via user-input feedback
3. Streamlit (Interactive UI)

Data Inputs

1. Opening Trial Balance for the year
2. Bank statements in excel and CSV formats (ICICI, BOB, Axis, Kotak, SBI, HDFC etc.)
3. JSON Files for Loading the Data and work last saved in case of data getting refreshed
4. Ledger master (List of Generic Master Ledgers normally used in predefined classification for accounts)
5. 44AD rule engine (YAML File with hardcoded rules for Tax Calculation)

Implementation Workflow

Step 1 – Upload Bank Statement and Opening Trial Balance

Handles:

1. Excel and CSV supported files

Automatically standardizes formats.

Step 2 – Narration Intelligence Engine

Classifies entries into:

1. Cash deposits
2. Capital account
3. Withdrawals
4. Bank charges
5. EMI/Loan
6. UPI/POS/IMPS

Step 3 – Ledger Suggestions with Instant Update

1. Per-row “Map Ledger” button
2. Updates instantly (no double clicks)
3. Saves mapping permanently
4. Retains user corrections

Step 4 – Final Output Generator

Produces:

1. Profit & Loss A/c and Balance Sheet (Preview on screen – for changes before PDF generation).
2. 44AD Computation as per New Regime and Old Regime and suggestion for selection of regime.
3. Detailed Computation Sheet with Computation of Gross Total Income, Net Income and Tax Break - Up.
4. PDF and Excel Export (Final Accounts)

Outcomes & Benefits

Quantitative Benefits

1. 80–90% reduction in time spent per client
2. 60% fewer errors in turnover recognition
3. 5x faster generation of P&L and Balance Sheet
4. 100% consistent classification logic

Qualitative Benefits

1. Better client insights
2. CA can handle more clients in less time
3. Improves compliance quality
4. Enhances digital practice model

Summary & Conclusion

This use case demonstrates how AI + Automation can transform the simplest yet most time-consuming compliance workload for CAs.

What earlier required hours of:

✔ Narration reading

✔ Classification

✔ Ledger mapping

✔ 44AD turnover computation

✔ Preparing financial statements

…now happens in minutes, with accuracy, speed and privacy.

“From Bank Statement to ITR-4 — Automated, Efficient & CA-Friendly.”