A Multimodal Generative AI Framework for Early Warning of Corporate Credit Rating Transitions Using Financial Analytics, Annual Report Intelligence, and News Sentiment

Abstract

Corporate credit ratings play a critical role in lending decisions, investment analysis, portfolio management, and regulatory compliance. Traditional credit rating methodologies rely predominantly on structured financial information and expert judgement (Altman, 1968; Hand & Henley, 1997), while the rapidly growing volume of unstructured corporate disclosures and financial news remains underutilized. Recent advances in Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) (Devlin et al., 2019; Brown et al., 2020) provide new opportunities to automatically analyse qualitative information contained in annual reports and financial news and integrate it with conventional financial indicators.
This paper proposes a multimodal Artificial Intelligence framework for the early warning of corporate credit rating transitions by combining quantitative financial analytics with qualitative information extracted from annual reports and financial news. The proposed framework consists of three complementary modules: (i) quantitative financial analysis based on trend-weighted financial ratios using Exponential Moving Average (EMA) smoothing and sector-specific normalization, (ii) an Annual Report Intelligence Pipeline that extracts and summarizes six strategically important sections of annual reports and transforms them into sentiment-based and semantic features, and (iii) a News Intelligence Module that captures recent developments affecting corporate credit quality through dynamic news sentiment analysis. These heterogeneous information sources are integrated using a hybrid deep learning architecture to generate an explainable early warning score representing the probability of future credit rating transition.
The methodology is demonstrated through a case study of Craftsman Automation Limited, illustrating how qualitative disclosures and contemporary news complement traditional financial analysis in identifying changes in corporate credit quality. The proposed framework seeks to provide financial institutions, banks, NBFCs, and credit rating agencies with an explainable decision-support system capable of improving proactive credit surveillance, in line with recent calls for interpretable AI in credit risk management (de Lange et al., 2022; Castro Vieira et al., 2025).