Document AI Technology Emerges as a New Growth Engine in the Fintech Ecosystem

Unstructured documents such as insurance claims, loan review papers, and investment reports remain one of the last frontiers in digital transformation within the financial industry. Efficiently processing and utilizing this vast amount of document data has become a critical challenge that directly impacts both operational efficiency and customer service quality.

Lomin, a leading Document AI company, is accelerating digital transformation in the financial sector by offering AI technologies that automatically read and comprehend complex documents. Leveraging Vision-Language Model (VLM)-based AI Optical Character Recognition (OCR), Lomin enhances document recognition accuracy and automates workflows such as insurance claims, loan processing, and research analysis. The company is positioning itself as a key infrastructure player in the fintech landscape. Recently, Lomin has been expanding beyond its on-premise deployments and is transitioning toward a cloud-based Software as a Service (SaaS) model under its new initiative, "Lomin 2.0," to further scale its market presence.

The following is the full interview with Jihong Kang, CEO of Lomin.

Q: While Lomin does not provide financial services directly, it is considered a crucial infrastructure provider in the fintech industry. What explains this positioning?

“We specialize in processing document data. The financial sector, known for handling vast amounts of documents, demands high efficiency and data utilization. Because our solutions directly address these needs, we are often recognized as a foundational player in fintech and insurtech.

Many enterprises aim to automate repetitive document-related tasks. Our solution automatically classifies documents, extracts necessary information, and inputs it into systems—replacing traditional manual workflows. This not only reduces costs and processing time but also allows employees to focus on higher-value work. For end customers, the faster processing of insurance claims or loan approvals leads to significantly improved service quality.”

Q: What sets Lomin’s AI OCR apart from existing technologies? You mentioned that your system 'understands' documents—can you elaborate on that?

“Traditional OCR merely extracts text. Our solution goes a step further—it interprets meaning. For example, when the term ‘Lomin’ appears, our AI can determine whether it's a company name, an address, or something else based on context. Document classification itself is also part of this ‘understanding.’

Previously, we relied on natural language processing (NLP) to handle textual data. Now, we have implemented VLM, which enables the AI to understand both images and text simultaneously—providing contextual comprehension.

This advancement allows our system to accurately recognize even low-resolution images, faxes, and mobile-captured documents. Additionally, because our models require minimal training data—sometimes just a few dozen samples—we can deliver customized solutions rapidly and cost-effectively.”

Q: How are these technologies being applied in real-world financial settings?

“Our technology is widely adopted by insurance and securities companies.

For instance, Kyobo Life Insurance has implemented our solution in its insurance claim and review process. It now automatically classifies and recognizes dozens of document types, such as medical certificates and prescriptions—cutting claim processing times in half.

Mirae Asset Securities has rolled out our technology across its entire organization to automate document classification and data extraction. Another securities firm has built a ‘Textscope Document Parser’ that integrates analyst reports with internal Large Language Models (LLMs). This turns complex reports into actionable data, significantly enhancing research efficiency.”

Q: In the process of document automation, how does Lomin ensure data privacy? What principles should companies follow when navigating data utilization regulations?

“Our financial clients operate our solutions entirely within their internal networks, ensuring no data leaves their secure environment. During document processing, we apply personal information de-identification and data encryption measures.

From a legal standpoint, explicit customer consent is required to use data for model training. Large financial institutions strictly adhere to these requirements. That said, we hope regulations will become more flexible in allowing greater use of masked data—classified as general data—for broader innovation.”

For more details, please refer to the full article via the link below.