Overview

Audit on Agentic Mode is an autonomous audit workflow system designed for Chartered Accountants to eliminate repetitive manual checking of ledgers, vouchers, and financial statements. By leveraging AI-agents, the system ingests Trial Balance, Day Book of the current year, performs end-to-end ledger scrutiny, voucher scrutiny, risk identification, materiality calculations, Checklist for applicability of tax audit, internal audit, CARO 2020 and ICFR, and generates complete working papers automatically. This reduces human effort, accelerates audit execution, and ensures higher accuracy, transparency, and standardisation in audit documentation.

This system automates core audit procedures by combining deterministic rule-based engines with specialised AI agents.

It allows an auditor to upload raw financial data (Trial Balance & Day Book of the current year), and the system generates:

1. Ledger scrutiny results
2. Voucher scrutiny results
3. Company applicability Checklist report
4. AI-generated risk assessments
5. Fully formatted audit working papers

All outputs are produced in a structured, traceable and repeatable manner, reducing manual work and improving audit quality.

AI Tool(s) Used

1. Python-based AI Ledger & Voucher Rules Engine (Custom ML + Rule Engine)
2. Gemini 2.5 flash API Key (for data classification, risk reasoning, drafting audit observations, and working paper narrative)
3. Agentic Orchestrator (Autonomous execution of multi-step audit tasks)
4. Streamlit as Frontend
5. System automatically:
6. ├── Reads Excel/PDF files (openpyxl, pdfplumber)
7. ├── Validates data structure (checks for required columns)
8. ├── Cleans data (removes subtotals, blank rows)
9. ├── AI Classification Agent runs:
10. │  ├── Maps TDS sections to ledgers
11. │  ├── Categorizes ledgers (Asset/Liability/Revenue/Expense)
12. │  ├── Identifies party names and relationships
13. │  └── Flags potential risk indicators
14. ├── Executes 40 Ledger Scrutiny Rules (pre-coded)
15. ├── Executes 35 Voucher Scrutiny Rules (pre-coded)
16. ├── AI Risk Assessment Agent analyzes findings
17. ├── AI Report Generation Agent creates documentation
18. └── Compiles 4 output reports

1. 4 Reports Generated:
2. 1. Ledger_Scrutiny_Report.xlsx
3. 2. Voucher_Scrutiny_Report.xlsx
4. 3. Company Applicability Assessment Checklist
5. 4. Integrated_Audit_Working_Paper.docx

Primary Use Cases

1) Automated Ledger Scrutiny

Input: Trial Balance (Excel)

Process:

1. Applies more than 40 rule-based checks (unusual balances, negative entries, suspense, frequency anomalies).
2. Output: Ledger scrutiny report (Excel) + summary of triggers.

Business Value: Saves hours of manual review and ensures consistency of audit procedures.

2) Automated Voucher/Transaction Scrutiny

Input: Day Book (Excel – payments, receipts, journal vouchers)

Process:

1. Applies 35+ rules (duplicate vouchers, weekend transactions, round-number risks, missing narrations, outliers).
2. Output: Voucher scrutiny report (Excel).

Business Value: Quickly surfaces high-risk transactions for deeper audit testing.

3) Company Applicability & Compliance Checklist

Input: Revenue estimate + company type

Process:

1. Runs Companies Act thresholds.
2. Generates applicability summary (Tax Audit, CARO, Internal Audit, ICFR).
3. Output: Applicability Report (Docx).

Business Value: Helps auditors determine statutory compliances instantly.

5) AI-Driven Data Classification

Input: Ledger names, narratives, descriptions

Process:

AI Agents classify:

1. Ledger category (asset/liability/income/expense)
2. Sub-group
3. TDS section mapping
4. Keywords and materiality indicators
5. Output: Structured classification JSON + integrated into working paper.

Business Value: Removes manual sorting, speeds up planning and risk assessment.

6) AI-Driven Risk Assessment

Input: Rule violations from ledger and voucher scrutiny

Process:

AI Agent assigns:

1. Risk score 0–100
2. Priority level (Critical/High/Medium/Low)
3. Expected impact
4. Suggested audit procedures

Output: Structured risk report + working paper summary.

Business Value: Improves audit planning and helps allocate effort intelligently.

7) AI-Generated Audit Working Papers

Input: All structured findings + financial analysis + AI summaries

Process:

AI Agent prepares:

1. Executive summary
2. Key observations
3. Suggested audit procedures
4. Supporting evidence references
5. SA 230 compliant documentation

Output: Complete audit working paper (Docx).

Business Value: Cuts documentation time by 70–80% and standardizes the quality of audit evidence

System-Level Use Cases

A. Automated End-to-End Audit Analytics

A single click runs ingestion, analysis, rule engines, materiality calculations, risk assessment, Assertion matrix, applicability checks, AI agents and documentation.

All reports share the same run_id, ensuring version control and audit trail.

B. Error-Resilient Processing

If any module fails:

1. System retries
2. Produces fallback reports
3. Writes diagnostics and logs into run_log.json

Ensures the auditor never ends up with a blank output.

C. Clean, Traceable Output Management

1. Atomic writes prevent duplicates
2. File hashing ensures idempotency
3. All files stored under a structured folder with run-level traceability

“Multiple audit pain points — One AI Agentic Solution”

Why This Project Matters

This system brings together audit analytics, automation, and AI-driven judgment support to reduce effort and increase reliability.

It delivers:

1. Faster audits
2. Deeper risk insights
3. Stronger documentation
4. Consistent, repeatable procedures
5. Reduced human errors

This makes it highly relevant for CA firms, internal audit teams, and enterprise financial controls.