Overview of Agentic AI in Financial Services
On January 27, a significant publication highlighted the observations regarding the emerging use of agentic AI within the financial services industry. AI agents are defined as autonomous systems that possess the capability for planning, decision-making, and executing actions to achieve specific goals without relying on predefined rules or continuous human oversight. This sets them apart from traditional automation tools, which typically require more direct human control.
Characteristics of AI Agents
These agents can operate with varying degrees of autonomy and oversight. However, they inherently lack human judgment and tacit knowledge, leading to potential challenges surrounding transparency and supervision.
Identified Risks of AI Agents
Several risks associated with AI agents have been identified, including:
- Scope and Authority: The potential for AI agents to act beyond their intended parameters.
- Auditability and Transparency: Difficulties in tracking and understanding the actions of AI agents.
- Data Security: Risks related to the misuse or inadvertent disclosure of sensitive information.
- Domain Knowledge: Insufficient expertise for managing complex tasks effectively.
- Misaligned Reward Functions: Challenges in aligning the objectives of AI agents with organizational goals.
- Persistent Risks: Unique risks associated with Generative AI, such as bias, hallucinations, and privacy concerns.
Types of AI Agents
FINRA has classified several distinct types of AI agents that are currently being utilized in the financial sector:
- Conversational Agents: These agents interact with users through natural language interfaces.
- Software Development Agents: These automate coding and manage infrastructure tasks.
- Fraud Detection and Prevention Agents: Specialized in identifying and mitigating fraud.
- Trade and AML Surveillance Agents: Focused on monitoring trades and anti-money laundering practices.
- Process Automation and Optimization Agents: Designed to streamline various processes.
- Trade Execution Agents: Responsible for executing trades efficiently.
In conjunction with this research, an infographic was released detailing these different types of AI agents, providing a visual representation of their functionalities and applications.
