文章摘要
Griffin, a UK bank, has commenced building an "agentic bank" focused on integrating AI agents and large language models (LLMs) with its core banking infrastructure.
- The project aims to empower autonomous AI agents to handle complex financial tasks, including fraud detection, customer service inquiries, and loan underwriting processes.
- A key development involves establishing a secure technical architecture and internal tools, enabling AI agents to safely interact with sensitive banking systems and proprietary data.
- Specific banking functions, API calls, and database operations are being abstracted into modular "tools" that AI agents can be trained to invoke responsibly.
- This framework emphasizes providing LLMs with the necessary context and awareness to leverage internal tooling and interact with various data sources, all while ensuring robust auditability and human supervision.