OCR software designed for tax documents

OCR makes it possible to automatically extract data from tax documents (bundles, VAT, bank statements), to simplify accounting, make declarations reliable and save time.

Koncile can process a wide range of tax forms, including forms 2065, 2035, as well as individual tax returns. Thanks to our template creation feature, it is also possible to process any other specific declaration by customizing the fields and adjusting the data extraction to meet your company’s precise needs.

Yes, integration is a key feature of Koncile's OCR solution.
Our technology is designed to be easily integrated with tax management software thanks to:

• A flexible API : It allows a secure and customizable connection with your tax software, accounting tool, ERP or any other system already in place.
• Pre-built connectors : We are actively developing direct integrations with the main tax software and automation platforms (Zapier, Make) to simplify implementation.

This integration allows a automatic transfer of extracted data (tax returns, Cerfa, declarations) to your tools, eliminating manual entries and reducing the risk of errors.

Koncile allows you to import scanned bank checks in PDF or image formats (PNG, JPEG, etc.) for automated reading and extraction of key information.

‍

Yes, with our OCR solution, you can automatically extract specific information from a bank statement, such as transactions, balances, dates, debits/credits, and much more. Our technology processes banking documents quickly and accurately. The extracted data can be easily integrated into your accounting or management systems, making it easier to track and analyze financial transactions.

Koncile integrates a trust scoring system which assesses the reliability of data extracted from various tax documents, such as tax returns, bundles, VAT receipts, bank statements.

This score is based on several criteria, including legibility of the document, the complexity of tax information to be extracted And the volume of data analyzed.

For example, a dense bundle or a poor-quality scanned document can affect the accuracy of the extraction. The algorithm takes into account both visual aspects (scan quality, text sharpness) and the semantic elements (consistency of amounts, identification of fiscal fields) to generate a overall trust score.

This score allows fiscal teams to quickly identify reliable data, to prioritize verifications and to secure reporting and compliance processes, including in the context of automated document processing via Invoice OCR.

Koncile features a state-of-the-art OCR (Optical Character Recognition) engine, specifically optimized for processing tax documents. Powered by machine learning, it delivers highly accurate data extraction even from documents of variable quality, including skewed scans, blurry images, unusual fonts, or complex layouts.

This engine goes beyond simple character recognition. It also interprets the structure of tax documents, such as declaration tables, numeric columns, or annexes, to ensure faithful and structured data extraction.

The integration of large language models (LLMs) further enhances reliability. Unlike traditional OCR engines, Koncile understands the tax-related context. It can validate, correct, or even infer missing or unclear information by interpreting the overall meaning of the document. This capability is crucial for extracting critical data such as declared amounts, tax references, or administrative identifiers with confidence and precision.

‍

Yes, Koncile offers advanced customization of tax data extraction to precisely match the needs of your department. With a simple and intuitive interface, you can define the fields to extract without requiring any technical expertise.

You can create custom fields such as "Tax ID number," "Declaration reference," "Amount due," or "Payment deadline," and assign each a specific data type (text, number, date, amount, tax identifier, etc.). This flexibility allows you to tailor the extraction to meet regulatory requirements and the expected formats from tax authorities.

You can also refine the results by setting up precise extraction rules. For example, you can specify that the VAT number always appears near a keyword like "intra-community VAT". This ensures more accurate data extraction and supports compliance with fiscal regulations.

1. Document Preprocessing (Image Optimization)
When a tax document is provided as an image (scan or photo), Koncile automatically applies image enhancements to improve text recognition. This includes straightening the document, removing visual noise, adjusting contrast, and sharpening the image. The goal is to produce a clear, high-quality image ready for analysis.

2. Advanced OCR (Reading and Structuring Information)
KonCile’s OCR engine then converts the image into digital text. Powered by machine learning, it goes beyond simple character recognition. It identifies the structure of tax documents, such as declaration tables, column-based financial data, and free-text sections. This ensures accurate reconstruction of the document in a structured, machine-readable format.

3. LLM Analysis (Contextual and Intelligent Extraction)
Large Language Models (LLMs) are then used to interpret the extracted text. These models understand tax-related context and can accurately identify key information such as amounts due, fiscal references, or VAT numbers, even when the data is poorly formatted or scattered across the document.

In summary: Koncile first enhances the document’s visual quality, then reads and structures the content with precision, and finally understands the meaning to extract the most relevant and reliable tax data.