A burgeoning Dutch startup aims to assist companies in extracting data from extensive volumes of intricate documents with a strong emphasis on accuracy and security. Recently, the startup, named Send AI, has secured support from Google’s Gradient Ventures to pursue this goal.
In the realm of document processing, Send AI is challenging established players like UiPath, Abbyy, Rossum, and Kofax by offering a customizable platform that enables companies to tailor AI models to their specific data-extraction requirements.
Traditionally, companies operating in highly regulated sectors such as insurance encounter a variety of document formats, ranging from PDFs and physical files to smartphone images taken from different angles and backgrounds. Processing such non-standard “unstructured” data can be complex for humans, let alone solely relying on machine-driven approaches, which can result in inaccuracies leading to claim rejections or administrative complications.
Unlike off-the-shelf document processing software designed for common document types across various industries, Send AI allows companies to train a computer vision model to identify specific documents and a separate language model to extract and validate pertinent data. Human intervention is integrated when uncertainties arise, ensuring each step is controlled and reviewed through a web interface.
The validation process ranges from simple checks like verifying numerical values to more intricate tasks such as cross-referencing registration numbers in databases for matches. Any uncertainties are flagged for human assessment.
Initially known as Autopilot and founded in Amsterdam in 2021, Send AI secured an initial \(100,000 investment from a university alumni fund. As it scales up its operations, the startup has raised an additional €2.2 million (\)2.4 million) in a pre-seed funding round co-led by Google’s Gradient Ventures and Keen Venture Partners, with contributions from angels associated with companies like DeepMind.
How Send AI Operates
Companies can utilize Send AI’s cloud-based software through APIs that streamline data ingestion from documents sent via email. Upon receiving the documents, Send AI enhances them visually before subjecting them to language models for classification and extraction.
Targeting primarily larger enterprises that grapple with document processing challenges, Send AI’s technology is beneficial for any business handling substantial document volumes.
Image Credits Send AI: Data extraction
In addition to the existing document-processing tools in the market, Send AI faces competition from startups leveraging advanced large language models (LLMs) like GPT-X from OpenAI. While these LLMs excel in tasks requiring subjective evaluations such as summarization or question-answering, Send AI’s focus on high accuracy across extensive document volumes sets it apart.
Send AI relies on smaller, open-source models that customers train by initially processing a small set of documents manually. Subsequent iterations involve processing new documents with human oversight for corrections.
Regarding pricing, Send AI adopts a credit-based system where customers pay per processing step, allowing flexibility based on the complexity of the task. This approach ensures cost-effectiveness and reliability, making it suitable for regulated industries like health insurance and government.
Data Control and Privacy
Send AI emphasizes customer control over their data, a key feature appealing to highly regulated industries. While cloud-based, Send AI’s approach differs from conventional LLMs by deploying isolated transformer models for each customer. This strategy enhances data privacy by preventing data blending across customers and enables precise data deletion upon request.
Despite operating in private beta mode, Send AI has already attracted notable clients such as insurance giant Axa. With plans to double its team size following the recent funding, the company is gearing up for a full commercial launch in the near future.
