Join Our Work Group GitHub Repo
Finance: Research in Generative AI Use Cases and Risks
Participants
- Jenny Yu - Himalaya Quantitative Solutions
- Victor Bian and Zhou Li - HydroX AI
- Please join us!!
Recent Relevant Material
| Country | Title | Type | Abstract | Source | Note |
|---|---|---|---|---|---|
| USA | A survey of Generative AI Applications | Research Paper | This survey provides a comprehensive overview of over 350 generative AI applications across various domains including finance. | arXiv | Offers a structured taxonomy and concise descriptions. |
| International | Generative AI in Finance | Research Paper | This article discusses the impact of generative AI in finance focusing on efficiency cost savings and enhanced forecasting accuracy. | IMF | Examines the broad applications and systemic implications in finance. |
| USA | Advancements in Generative AI | Research Paper | This review paper covers GANs, GPT, Autoencoders, Diffusion Modelsm and Transformers, highlighting their advancements and applications in finance. | arXiv | Provides a detailed review of state-of-the-art generative AI models. |
| International | Generative AI: A systematic review using topic modeling | Research Paper | This systematic review explores the various models and applications of generative AI in finance. | ScienceDirect | Uses topic modeling to analyze trends and advancements. |
| USA | Generative AI in the Finance Function of the Future | Industry Report | Discusses how generative AI can transform core finance processes, business partnering, and risk mitigation. | BCG | Focuses on future integration and transformation. |
| USA | GenAI for finance: Practical steps to get started | Industry Report | Provides specific use cases and practical steps for implementing generative AI in finance. | Deloitte | Offers guidance on starting GenAI projects. |
| International | Generative AI in Financial Services | Industry Report | Explores how financial institutions are using GenAI for fraud detection, customer service, and compliance. | McKinsey | Details applications and challenges in financial services. |
| USA | The Impact of Generative AI in Finance | Industry Report | Examines the potential impact of generative AI on finance workflows and decision-making. | Deloitte | Discusses the role of data and AI in finance. |
| International | Supercharge your Finance workforce with GenAI | Industry Report | Highlights the opportunities for finance professionals to leverage GenAI for improved business intelligence and outcomes. | KPMG | Focuses on enhancing finance functions with GenAI. |
| International | Generative AI and LLM in financial risk modeling and applications | Research Paper | Discusses the use of large language models in financial risk modeling and applications. | European Journal of Finance | Emphasizes the role of LLMs in finance. |
| USA | Generative AI Special Topic Hub | Research Hub | Curated view into new research on generative AI applications in finance and other fields. | SSRN | Provides access to early-stage research. |
| USA | Why finance should lead the adoption of generative AI | Industry Report | Argues that finance functions are optimally positioned to lead GenAI adoption due to their data-centric nature. | KPMG | Discusses strategic initiatives for finance leaders. |
| USA | The state of AI in early 2024: Gen AI adoption spikes and starts to deliver value | Industry Report | Analyzes the adoption and business value of generative AI in 2024. | McKinsey | Highlights survey results and trends. |
| International | Stanford’s 2024 AI Index Tracks Generative AI and More | Industry Report | Tracks the growth, costs, and responsible use of generative AI. | IEEE Spectrum | Offers insights into the AI index and trends. |
| International | Advancements in Generative AI: A Comprehensive Review | Research Paper | Reviews the advancements in generative AI, focusing on models like GANs and GPT. | arXiv | Comprehensive review of generative AI advancements. |
| USA | Generative AI in Education and Its Implications | Research Paper | Examines the impact of generative AI in education and its potential applications in finance. | Springer | Highlights the interdisciplinary applications of GenAI. |
| International | GenerativeAI4Finance | Open-Source Project | Generative AI, Co-pilots For Use Cases in Finance. Includes foundational models, high-level reusable components, and examples of applications in finance. | GitHub - GenerativeAI4Finance | Includes various use cases such as Lending, Investment Banking, Embedded Finance, and more. |
| International | FinGPT | Open-Source Project | FinGPT: Open-Source Financial Large Language Models. Offers financial sentiment analysis and other financial applications. | GitHub - AI4Finance-Foundation/FinGPT | Focuses on financial sentiment analysis, training models with LoRA method. |
| International | GenAI-Showcase | Open-Source Project | Generative AI Use Cases Repository. Showcases applications in generative AI, including Retrieval-Augmented Generation (RAG), AI Agents, and industry-specific use cases. | GitHub - mongodb-developer/GenAI-Showcase | Integrates MongoDB with RAG pipelines and AI Agents for efficient data management. |
| International | FinGPT-all-branch | Open-Source Project | Data-Centric FinGPT. A full-stack framework for financial language models, focusing on real-time data curation and lightweight adaptation. | GitHub - Generative-Alpha/FinGPT-all-branch | Prioritizes real-time data updates and lightweight model adaptation. |
| International | AI4Finance Foundation | Open-Source Project | AI4Finance Foundation: Open-source financial LLMs and NLP models. Provides full pipelines for LLM training and fine-tuning in finance. | GitHub - AI4Finance Foundation | Supports various financial applications with real-time data processing and fine-tuning. |
Benchmark of GenAI on Financial Applications
| Data | Task | Test | Evaluation | License |
|---|---|---|---|---|
| NER (Alvarado et al., 2015) | named entity recognition | 980 | Entity F1 | CC BY-SA 3.0 |
| FiNER-ORD (Shah et al., 2023b) | named entity recognition | 1080 | Entity F1 | CC BY-NC 4.0 |
| FinRED (Sharma et al., 2022) | relation extraction | 1,068 | F1, Entity F1 | Public |
| SC (Mariko et al., 2020) | causal classification | 8,630 | F1, Entity F1 | CC BY 4.0 |
| CD (Mariko et al., 2020) | causal detection | 226 | F1, Entity F1 | CC BY 4.0 |
| FNXL (Sharma et al., 2023) | numeric labeling | 318 | F1, EM Accuracy | Public |
| FSRL (Lamm et al., 2018) | textual analogy parsing | 97 | F1, EM Accuracy | MIT License |
| FPB (Malo et al., 2014) | sentiment analysis | 970 | F1, Accuracy | CC BY-SA 3.0 |
| FiQA-SA (Maia et al., 2018) | sentiment analysis | 235 | F1 | Public |
| TSA (Cortis et al., 2017) | sentiment analysis | 561 | F1, Accuracy | CC BY-NC-SA 4.0 |
| Headlines (Sinha and Khandait, 2021) | news headline classification | 2,283 | Avg F1 | CC BY-SA 3.0 |
| FOMC (Shah et al., 2023a) | hawkish-dovish classification | 496 | F1, Accuracy | CC BY-NC 4.0 |
| FinArg-ACC (Sy et al., 2023) | argument unit classification | 969 | F1, Accuracy | CC BY-NC-SA 4.0 |
| FinArg-ARC (Sy et al., 2023) | argument relation classification | 496 | F1, Accuracy | CC BY-NC-SA 4.0 |
| MultiFin (Jørgensen et al., 2023) | multi-class classification | 690 | F1, Accuracy | Public |
| MA (Yang et al., 2020a) | deal completeness classification | 500 | F1, Accuracy | Public |
| MLESG (Chen et al., 2023a) | ESG Issue Identification | 300 | F1, Accuracy | CC BY-NC-ND |
| FinQA (Chen et al., 2021) | question answering | 1,147 | EM Accuracy | MIT License |
| TATQA (Zhu et al., 2021) | question answering | 1,668 | F1, EM Accuracy | MIT License |
| Regulations,long-form question answering | 254 | ROUGE, BERTScore | Public | |
| ConvFinQA (Chen et al., 2022b) | multi-turn question answering | 1,490 | EM Accuracy,MIT License | |
| ECTSum (Mukherjee et al., 2022) | text summarization | 495 | ROUGE, BERTScore, BARTScore | Public |
| EDTSum | text summarization | 2000 | ROUGE, BERTScore, BARTScore | Public |
| BigData22 (Soun et al., 2022) | stock movement prediction | 1,470 | Accuracy, MCC | Public |
| ACL18 (Xu and Cohen, 2018) | stock movement prediction | 3,720 | Accuracy, MCC | MIT License |
| CIKM18 (Wu et al., 2018) | stock movement prediction | 1,140 | Accuracy, MCC | Public |
| German (Hofmann, 1994) | credit scoring | 1000 | F1, MCC | CC BY 4.0 |
| Australian (Quinlan, [n. d.]) | credit scoring | 690 | F1, MCC | CC BY 4.0 |
| LendingClub (Feng et al., 2023) | credit scoring | 2,690 | F1, MCC | CC0 1.0 |
| ccf (Feng et al., 2023) | fraud detection | 2,278 | F1, MCC | (DbCL) v1.0 |
| ccfraud (Feng et al., 2023) | fraud detection | 2,097 | F1, MCC | Public |
| polish (Feng et al., 2023) | financial distress identification | 1,736 | F1, MCC | CC BY 4.0 |
| taiwan (Feng et al., 2023) | financial distress identification | 1,364 | F1, MCC | CC BY 4.0 |
| ProtoSeguro (Feng et al., 2023) | claim analysis | 2,381 | F1, MCC | Public |
| travelinsurance (Feng et al., 2023) | claim analysis | 3,800 | F1, MCC | (ODbL) v1.0 |
| FinTrade,stock trading | 3,384 | CR, SR, DV, AV, MD | MIT License |
Applications of GenAI in Finance
Customer Personalization
One of the primary applications of GenAI in finance is providing personalized product recommendations and offers. By analyzing user data, financial institutions can generate unique suggestions for investment portfolios, financial products, and services. This capability extends to both customers and institutions, with 72% of customers believing products are more worthwhile when tailored to their individual needs. For example, generative AI can help firms deliver flexible and relevant conversations that improve the overall customer experience by adapting the conversational style to match that of the customer
Fraud Detection and Risk Management
GenAI plays a critical role in enhancing fraud detection and risk management in financial institutions. By generating synthetic examples of fraudulent transactions, GenAI can help train machine learning algorithms to recognize and differentiate between legitimate and fraudulent patterns in financial data. This capability allows for faster detection and prevention of fraud, ultimately improving the overall security and integrity of financial operations. Additionally, by learning from historical financial data, GenAI models can make predictive analytics about future trends, asset prices, and economic indicators, providing valuable insights into potential risks and opportunities
Automation and Operational Efficiency
The adoption of GenAI in finance has led to significant improvements in automation and operational efficiency. For instance, financial services institutions leverage machine learning (ML) technologies like computer vision, optical character recognition (OCR), and natural language processing (NLP) to streamline customer onboarding and know-your-customer (KYC) processes. GenAI can automate the generation of financial reports, thereby freeing up expert time for strategic analysis and reducing errors for greater accuracy. This automation extends to various back-office functions such as accounting, legacy software maintenance, document analysis, and responding to regulator requests.
Regulatory Compliance
Financial institutions are required to explain their decisions and actions to internal and external stakeholders, including prudential supervisors. GenAI can aid in fulfilling regulatory requirements, such as those related to anti–money laundering and combating the financing of terrorism. While many regulators have yet to publish guidelines specifically for the use of AI in financial transactions, financial institutions should take internal measures to update their anti-money laundering and client relationship policies and procedures when introducing AI technology to automate their customer-facing interactions.