AI OCR automation that turns raw documents into trusted data
ID documents
Invoices
Bank statements
Prescriptions
Insurance docs
Unique OCR technology for document analysis and fraud detection
Document automation workflows
Splitting, renaming and categorization, data enrichment & validation.
How our clients tackle complex document processing challenges
An AI OCR tool built for regulated environments
Founded by a former lawyer, Koncile was designed with compliance and data protection at its core. We are independently audited under SOC 2 and compliant with GDPR and HIPAA requirements.





Talk to an expert who knows your industry.
- Get an external expert opinion on your specific use case
- Benchmarks and results from hundreds of clients in your industry
- Help activating the right Koncile features for your workflow
Real life insights on document automation
All your questions on Koncile
Koncile is the AI platform for smart document extraction and fraud detection.
Our approach: we build specialized analysis agents with domain experts, so the rules and logic encoded in each agent reflect the reality of your business. Koncile extracts the fields with over 99% accuracy, runs the full pipeline (splitting, classification, routing, enrichment, matching) and detects anomalies before they enter your systems.
We have 10,000+ users at Servier, Epson, Cerfrance, Bilfinger, SpendHQ and Meelo. GDPR compliant and SOC 2 certified.
Three steps. Koncile cleans the document (deskewing, denoising, resolution enhancement). An OCR engine optimized by machine learning reads the text and reconstructs the structure (tables, columns, paragraphs). Large language models turn the raw text into structured data. Output is structured JSON, CSV, XLSX or Excel, ready to push into your ERP, CRM, HRIS or accounting tool. Every field comes with a confidence score.
Take a customer file: a single PDF bundling an ID, a pay stub, a bank statement and an insurance certificate.
Koncile splits the file, classifies each document, extracts the right fields, and matches everything against your customer database. It flags missing pieces, spots inconsistencies (a name that differs between the ID and the pay stub) so you can come back to the customer. It also gives an overall confidence score for the whole file.
Same logic for supplier onboarding, claim files, KYC and AP batches.
Forensic analysis catches the obvious: PDF layer scans, font inconsistencies, suspicious metadata, pixel anomalies.
But on its own this fraud detection is weak. Fraud sometimes leaves no visible traces.
This is where Koncile stands out. We have built contextual models for every major document type (ID documents, pay stubs, insurance documents, registration documents, bank statements, invoices, etc.), each one trained on the fraud patterns specific to its family and continuously updated with domain experts (accountants, lawyers and sector specialists). Every document gets a fraud score between 0 and 1, and high risk files trigger automatic alerts in your workflow.
Yes. An OCR trained on handwriting combined with LLM reasoning makes Koncile strong on tough inputs: blurry scans, phone photos, faded receipts, unusual fonts, prescriptions filled by hand.
On printed prescriptions, accuracy goes above 99%. On handwritten ones, the engine reaches 94.7%, well ahead of typical industry rates of 85 to 89%. Every extracted field comes with a confidence score, so you know which values are reliable and which deserve a quick human check before they flow downstream.
Three paths.
The REST API and Python SDK plug Koncile directly into your ERP, CRM, credit origination platform or supplier portal, with structured JSON responses and webhook support for asynchronous workflows. The MCP OCR server lets AI agents call Koncile's tools natively. Most teams ship to production in a day, and our team partners with you to design the right setup for your stack.
SOC 2 Type II certified, GDPR compliant. Koncile is audited annually (last audit completed in 2025) and HDS aligned for healthcare data. All your data stays in the European Union: documents are processed on AWS in eu-west-1 (Paris), with disaster recovery in eu-central-1 (Frankfurt). Nothing transits outside the EU. Data is encrypted at rest with AES 256 and in transit with TLS 1.3. Authentication uses OAuth 2.0 with JWT, Google SSO and 2FA (mandatory for admin and production access). We do not train AI models on your data, and you remain the sole data controller at all times. Enterprise customers can deploy on a private cloud. Full policy is here.




















