A Snapshot of AI/ML Usage in Financial Institutions in 2023
Discover how generative AI and machine learning are reshaping the financial sector while addressing governance, ethical, and regulatory challenges.
Introduction and Overview
The financial services industry has been at the forefront of incorporating artificial intelligence (AI) and machine learning (ML) technologies. The IIF-EY Annual Survey Report for 2023 provides an in-depth look at the current state of AI/ML adoption, governance, and the emerging role of generative AI within financial institutions.
Historical Context and Survey Methodology
While the IIF has published reports on AI/ML since 2018, this 2023 survey includes insights from 65 global financial institutions. The survey incorporates diverse perspectives and methodologies, highlighting a continued evolution in AI/ML practices across multiple geographic regions and institution types.
Current AI/ML Use Cases
A narrow majority of institutions have fewer than 50 AI/ML use cases, while 15% have over 350 AI/ML models. Financial institutions use AI/ML to enhance processes, discover new risk segments, and achieve cost savings. Most respondents foresee generative AI being deployed internally in the next 12 months, later expanding to customer-facing applications and ecosystem integrations.
Generative AI’s Impact
Generative AI is expected to significantly expand model inventories, with 86% of institutions predicting a notable increase within three years. The majority of survey respondents believe generative AI will have revolutionary impacts. However, 77% also impose restrictions on its use, primarily due to concerns over data privacy and model explainability.
Governance and Oversight
AI/ML governance remains critical, with 66% of respondents appointing or planning to designate a C-suite manager for AI/ML ethics and oversight. Tollgate processes are widely used to decide the appropriateness of deploying AI/ML models. Furthermore, 78% of institutions have established policies to manage risks associated with generative AI, emphasizing model risk management and compliance.
Ethical Considerations
Ethical issues such as data privacy, safety, and equitable treatment are paramount. Most institutions conduct awareness programs and training on AI/ML ethics for their employees. The report also indicates the diverse strategies employed to prevent biased outcomes, including implementing institutional codes of ethics and engaging external advisors.
Validation and Control Techniques
To ensure AI/ML robustness, institutions employ a comprehensive set of validation techniques, including ongoing monitoring, data quality checks, and in-sample/out-of-sample testing. These processes are critical in mitigating risks and ensuring AI/ML models’ reliability and fairness.
Regulatory and Supervisory Engagement
Many institutions have engaged or plan to engage with regulators on AI/ML issues, mainly focusing on explainability and bias. Regulatory developments influence AI/ML adoption, with voluntary standards and frameworks increasing in guiding practices.
Conclusion
As the potential and risks of generative AI continue to unfold, financial institutions are adapting governance frameworks and regulatory strategies to harness AI/ML’s benefits responsibly. Future IIF-EY surveys will keep exploring these themes, ensuring ongoing dialogue and improvement within the industry.
