Korea Financial Investment (KFI), an AI-powered fintech company that analyzes distressed real estate assets, said on the 2nd that it had been selected to participate in the second-half 2026 Digital Finance Innovation Testbed (D-Testbed), organized by the Financial Services Commission and the Korea Fintech Center. Selected companies will spend 12 weeks refining their technology in an analytical environment based on real financial-sector data and will receive expert mentoring and an assessment of their business viability. KFI’s AI analysis engine, AIFF (AI-Fast Finder), automatically structures unstructured documents such as property registry records and appraisal reports, then uses AVM and IRR simulations and XAI risk analysis to calculate purchase prices, returns and risks. It processes an asset analysis that previously took about 24 hours in roughly 3 seconds; the company says its AI scoring accuracy is about 98% based on its own validation criteria.
In the wake of distress in real estate project finance and rental-deposit fraud, villas and officetels are flooding the court and public auction markets. But objective standards for determining the recovery value of non-apartment assets remain scarce. A startup using AI to address that gap is now being put to the test by financial regulators. Korea Financial Investment (KFI) said on the 2nd that it had been selected to participate in the “Second-Half 2026 Digital Finance Innovation Testbed (D-Testbed)”, organized by the Financial Services Commission and the Korea Fintech Center.
What is the D-Testbed? A 12-week trial using real financial-sector data
The D-Testbed is a program through which financial regulators identify promising fintech companies with innovative technologies and help them verify their technologies’ effectiveness and commercial viability using real financial data. Selected companies refine their technology over 12 weeks in an analytical environment that provides real data from the financial sector. They also receive mentoring from experts in law, investment, statistics and security, along with an assessment of their business viability.
During the test period, KFI plans to use real financial data to rigorously validate its valuation models for non-apartment nonperforming loans (NPLs) and court and public auction assets, and refine the engine to a level suitable for direct use in financial institutions’ review processes.
How AIFF works: One engine, from document structuring to investment decisions
Financial institution analysts must gather information themselves from more than 15 sources, including court auction records, property registry documents and appraisal reports, and calculate asset cash flows. According to the company’s analysis, reviewing a single asset takes about 24 hours, with roughly 75% of that time spent on repetitive data collection, entry and calculations.
KFI’s AI analysis engine, AIFF (AI-Fast Finder), handles the entire process in one system. It automatically structures unstructured documents—including property registry records, appraisal reports and auction property statements—using Document AI, OCR and LLM inference, then links creditor, collateral and debtor relationships in a knowledge graph. On that basis, AVM (automated valuation) and IRR simulations, Monte Carlo scenario analysis, and explainable AI (XAI) risk analysis calculate purchase prices, returns and risk scores, while also providing the rationale behind the results. The system is designed to “self-evolve,” improving in accuracy through reinforcement learning and fine-tuning as more data accumulates.
The analysis extends to filtering investment-eligible properties based on legal-rights risks, grading location competitiveness based on about 150 variables, calculating IRR, NPV and DSCR using cash-flow-based DCF analysis, estimating the probability of winning an auction, simulating bidding strategies and building an optimal portfolio. It automatically processes an asset analysis that previously took about 24 hours in roughly 3 seconds and reduces the number of assets requiring review by 70% to 80%.
Validation to date: Two patents and 98% AI accuracy; plans to expand as SaaS for financial institutions
KFI has registered two patents related to a digital securities trading system for the securitization of physical real estate assets and has filed a patent application for an AI real estate intelligence insight engine system. It has signed memorandums of understanding with an appraisal firm to validate valuation data and with an accounting firm to validate its cash-flow analysis engine and financial modeling. It is also conducting proofs of concept with major appraisal firms, accounting firms and construction companies. The company says its AI scoring accuracy is about 98% based on its own validation criteria. This year, it was also selected for the Ministry of Science and ICT’s K-Global ICT Support Program, the Korea Credit Guarantee Fund’s NEST program and the Federation of Korean Industries’ “Plus” project.
Based on the D-Testbed validation results, KFI plans to step up the provision of its analysis solutions as software-as-a-service (SaaS) and APIs to financial institutions, including banks, savings banks and asset managers, and later expand its market to individual investors and professional landlords.
“AIFF is an engine that uses data and AI to standardize distressed-asset reviews that used to take people days,” said Koh Byung-wook, CEO of Korea Financial Investment. “Through this D-Testbed, we’ll validate our technology using real financial data and set the standard for valuing distressed non-apartment assets in a way the financial sector can trust and use.”
────────────────────────────────────────
