Klassif AI: IDP Software Vendor
On This Page
Belgian AI platform specializing in order confirmation processing with human-in-the-loop training and 93% accuracy claims.
Overview
Founded in 2017 by Tom Vermeulen in Leuven, Belgium, Klassif AI operates as a product of Raccoons. The platform targets order confirmation processing rather than general document processing, which is the core strategic bet: rather than competing across multiple document types like broad intelligent document processing (IDP) platforms, Klassif AI specializes in a single high-frequency workflow, processing incoming sales orders via email, PDF, or scan.
The platform claims 93% extraction accuracy against 60% for legacy optical character recognition (OCR) solutions, with an 87.3% average reduction in manual effort and 2.5x median return on investment over six-month payback periods. Van Marcke customer Stijn Hoegaerts reports "ten hours saved per day on average," a concrete outcome metric the company uses in enterprise positioning.
In February 2026, Antoine Smets, CEO of Klassif AI, announced the company's selection into BEyond Belgium, a year-long PulseFoundation acceleration program for ten Belgian tech scale-ups. Smets stated: "Being part of BEyond recognizes where we've come and accelerates where we're going. This program connects us with world-class mentors, industry leaders, and a network of founders who share our ambition to scale impact across enterprises." The selection places Klassif AI in an early-to-mid scaling phase rather than Series A maturity, but signals institutional recognition of its growth trajectory in the Belgian tech market.
How Klassif AI processes documents
Klassif AI ingests orders through Microsoft Outlook 365, Google Workspace, Dropbox, and direct ERP connections, then applies machine learning and natural language processing (NLP) to extract customer data, product codes, quantities, and pricing. Where confidence falls below threshold, the platform routes documents to human reviewers. Those corrections feed directly back into the model, enabling continuous improvement through normal employee workflows rather than dedicated retraining cycles.
The open-sourced react-pdf-ner-annotator handles PDF annotation using Tesseract.js for OCR, released after three internal development iterations. Model versioning supports production rollback with activity monitoring, allowing teams to revert to a prior model version if a new iteration degrades on edge cases. Typical deployment runs four to five weeks with no infrastructure changes required.
Use cases
Order confirmation processing
Manufacturing, logistics, chemicals, food industry, construction, and banking organizations deploy Klassif AI for order confirmation workflows. The platform processes incoming orders via email, PDF, or scanned documents, extracting customer data, product codes, quantities, and pricing. The company claims up to 93% productivity increases compared to manual handling, with Van Marcke's ten-hours-per-day saving as the most specific published customer outcome.
Invoice management
Organizations use Klassif AI to process supplier invoices and customer billing documents. Extracted data is validated before passing to ERP systems, reducing downstream correction cycles. The company claims 80% faster processing compared to traditional OCR solutions, though this figure is self-reported and carries no independent verification.
Technical specifications
| Feature | Specification |
|---|---|
| Core technology | Machine learning, NLP with human-in-the-loop training |
| Document types | PDF, email, scanned documents |
| Accuracy rate | 93% vs 60% traditional OCR (vendor-reported) |
| Manual effort reduction | 87.3% average (vendor-reported) |
| ROI | 2.5x median, 6-month payback (vendor-reported) |
| Implementation timeline | 4–5 weeks typical deployment |
| Infrastructure changes | None required |
| Open source component | react-pdf-ner-annotator (React, Tesseract.js) |
| Integrations | Microsoft Outlook 365, Google Workspace, Dropbox, ERP systems |
| Model versioning | Production rollback with activity monitoring |
| Parent company | Raccoons |
All accuracy and ROI figures above are vendor self-reported via klassif.ai. No independent third-party benchmark has verified these claims as of April 2026.
Resources
- Website
- Platform Overview
- Open Source Annotator
- BEyond Belgium Program Coverage
Company information
Headquarters: Leuven, Belgium
Founded: 2017
Founder: Tom Vermeulen
CEO: Antoine Smets
Parent: Raccoons
For vendors competing in adjacent document types, see DOConvert for supply chain purchase order automation and Paperbox for insurance mailroom processing. For a broader view of human-in-the-loop document processing approaches, the capability guide covers confidence thresholds, review interfaces, and accuracy optimization patterns.