BUSN30135 - Winter 2025
Identify non-recurring expenses and revenues not relevant to core operations of a business
Student team: Sandesh Puligundla, Pramod Nayak, Anatoly Bykhovsky, Bruno Graca Coelho
Abstract
We leverage Large Language Models (LLMs) to extract and analyze financial data from complex 10-K reports, bridging the growing gap between GAAP net income and core earnings. As financial disclosures become increasingly complex, this AI-powered workflow integrates financial knowledge and machine reasoning to identify abnormal financial adjustments and enhance earnings analysis. Rigorous testing using data from 18 companies demonstrated that optimized AI prompting significantly improves accuracy, though challenges remain in handling complex financial adjustments. Complexity matters, as with complex financial events the solution misses or miscategorizes them.