PARIS and SAN FRANCISCO (BLOOM) – Retab, an innovative AI platform designed to automate large-scale document processing, has emerged from stealth mode on Wednesday, announcing a $3.5 million pre-seed funding and introducing what it touts as the “most potent document AI platform” available.

The company, created by engineers who previously developed in-house automation solutions for logistics, strives to overcome the challenges developers encounter when incorporating large language models (LLMs) into document-intensive processes.

“Retab is the developer-centric platform we always wished to have during development,” stated Louis de Benoist, co-founder and CEO. “While many build impressive demos that seem magical, they often fail when deployed in real-world applications.”

Leading the funding round were VentureFriends, Kima Ventures, and K5 Global, with contributions from Eric Schmidt (via StemAI), Olivier Pomel (CEO of Datadog), and Florian Douetteau (CEO of Dataiku). The investment will be used to expand infrastructure, enhance developer tools, and integrate with automation platforms such as Zapier, Dify, and n8n.

Retab differentiates itself from other AI tools as it is not an AI model per se but a control layer that manages processes ranging from prompt engineering to model selection and output verification. Developers specify the structure of data to be extracted, while Retab handles the logistics using its proprietary tools and advanced language models from top providers like OpenAI, Google, and Anthropic.

Key features include:

• Self-Optimizing Schemas: The system automatically adjusts and benchmarks extraction instructions before they go live.
• Model-Agnostic Routing: Retab chooses the best-performing model based on accuracy, speed or cost—reducing overhead up to 100x compared to other tools.
• k-LLM Consensus and Guided Reasoning: A system of step-by-step reasoning and multiple-model agreement acts as a built-in quality and safety layer.

Clients are already reaping benefits. A leading logistics company employed Retab to automate document processing, achieving a 99% accuracy rate with a quicker, more cost-effective model. Similarly, a financial firm utilized the platform to extract detailed metrics and risk indicators from lengthy quarterly reports—a task that previously took analysts several days.

“Retab is the OS for reliably extracting structured data,” said de Benoist. “It wraps the best models in a layer of logic that actually makes them usable with error handling and structured outputs.”

Florian Douetteau, CEO of Dataiku and one of the company’s investors, called Retab’s approach “crucial for scaling the AI-fication of the economy.”

Looking ahead, Retab plans to expand beyond PDFs and scanned documents into unstructured web data, enabling AI agents to access and interpret everything from invoices and contracts to customs manifests.

With just ten employees, the company is already positioning itself as a key infrastructure layer in the AI stack, bridging the gap between messy, real-world data and the clean, structured formats modern systems rely on.

For more information, visit www.retab.com.