Financial AI Agent 'M&A Navi SmartDD' Achieves Over 99% Structuring Accuracy for PDF Financial Documents Including Ledgers with Proprietary Engine

M&A Navi Inc. (Headquarters: Tokyo, CEO: Yusuke Takita) has independently developed a technology based on years of research to convert PDF financial documents into structured data. This structuring engine serves as an innovative foundational technology that enables the use of generative AI in the highly accuracy-demanding field of finance.

In particular, the company's structuring engine autonomously converts large and complex financial documents—including balance sheets, income statements, hundreds-of-pages general ledgers, and high-density depreciation schedules—into highly accurate structured data through repeated self-validation. Benchmark testing using this engine achieved a precision rate of 99.2%. This high accuracy is based on rigorous measurement using over 200 verification rules.

This high accuracy is supported by an optimized OCR mechanism tailored to document characteristics, a rule-based verification system, and an orchestration mechanism that autonomously replans and retries until accuracy meets required standards by combining these components.

This technology serves as the core infrastructure of the financial AI agent 'M&A Navi SmartDD' and is expected to bring significant convenience and efficiency to financial analysis tasks for professionals supporting SMEs.

For financial analysis with AI, choose SmartDD—the verifiable solution.

Financial AI Agent 'SmartDD' Service Page

https://ma-navigator.com/smartdd

Benchmark Test Overview

Test Scope

Item

Details

Target

Financial statements (3 fiscal periods × 10 companies), account breakdowns, general ledgers, and full depreciation schedules

Selection Criteria

Comprehensive collection of financial statements exported from major domestic accounting software

Diversity Assurance

Selection prioritized format diversity due to differing layouts, notations, and chart of accounts across accounting software

Accuracy Definition

Structuring Accuracy (%) = (Total Verification Items − Error Count) / Total Verification Items × 100

Verification Details

Document Type

Verification Overview

Financial Statements (BS/PL/Cost of Goods Sold/Selling & Admin Expenses)

- Hierarchical consistency from asset/liability/equity breakdowns to subtotals and totals, plus balance sheet equilibrium

- Calculation integrity of five-tier profit (Gross Profit → Operating Profit → Ordinary Profit → Pre-tax Profit → Net Profit)

- Consistency of component totals and cross-checks with P&L

- Aggregation integrity of material, labor, and expense costs, cross-checked with P&L

Account Breakdowns

- Internal consistency of each breakdown item, cross-verified with corresponding BS/PL accounts

General Ledger

- Continuity of balances, debit-credit integrity, and reconciliation with BS/PL for ending balances and period cumulative totals

Depreciation Schedules

- Calculation integrity of acquisition cost, accumulated depreciation, and book value, cross-verified with corresponding BS accounts

Utilization of Structured Data

Structured data can be accessed via four interfaces:

1. SmartDD (https://ma-navigator.com/smartdd)

The 'Analysis Lab' feature of the financial AI agent SmartDD uses verified data as input to provide advanced financial analysis as standard functions, including detection of under-depreciation, customer aggregation, fraud risk detection, capital requirement forecasting for equipment investment, cash flow calculation, and cash flow statement creation.

2. Excel Add-in (alpha release)

Import structured financial data directly into Excel. Automates data entry without altering existing Excel-based analysis workflows.

3. RESTful API (alpha release)

Provides APIs for integration with internal CRM systems such as Salesforce and HubSpot. Enables programmatic execution from PDF input to structured data retrieval.

4. MCP (Model Context Protocol) (alpha release)

Offers an MCP server allowing AI agents like Claude to directly reference structured financial data. Enables seamless connection from data retrieval to analysis in AI-driven financial analysis and reporting workflows.

Future Outlook

The technology has demonstrated proven accuracy in financial document structuring. It can be applied to any business process involving financial documents.

Application Areas

Use Cases

Loan Underwriting, Credit Assessment, Financial Due Diligence

- Automate and standardize analysis and evaluation through automatic financial statement ingestion

- Fraud risk scoring

- Automated comparative time-series analysis of financial data

Tax Accounting

- Digitization of ledger data and automation of validation

Business Planning

- Enables AI-driven direct-method cash flow calculation and cash flow statement creation using ledger data

Moreover, the framework of 'AI structuring of unstructured documents + rule-based automated verification' is not limited to financial documents. The same technical foundation can be applied to any business domain where verification rules can be defined.

The company welcomes inquiries from organizations considering adoption.

[Company Overview]

"Empowering every entrepreneur with free M&A through technology."

Company Name: M&A Navi Inc.

Headquarters: 5F, 1-8-12 Kitashinagawa, Shinagawa-ku, Tokyo

CEO: Yusuke Takita

Corporate Website: https://corp.ma-navigator.com/

M&A Navi: https://ma-navigator.com/

M&A and Business Succession Column: https://ma-navigator.com/columns/