The investment management sector is undergoing a significant transformation with the integration of artificial intelligence (AI) into its operations. Many firms are utilizing AI agents to assist portfolio managers, analysts, and compliance officers. However, a common challenge persists: the difficulty in clearly defining the type of intelligence these tools embody.
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?
Establishing a framework for AI in investment management
Every instance of AI integration provides an opportunity to establish clear boundaries and define the scope of these tools. Without a proper classification of AI capabilities, governance becomes challenging, and scaling the technology effectively is hindered. To address this challenge, a collaborative effort between DePaul University and Panthera Solutions has resulted in a multi-dimensional classification system for AI agents in the investment sector. This article summarizes findings from the academic paper titled A Multi-Dimensional Classification System For AI Agents In The Investment Industry, which has been submitted for peer review.
This classification system aims to provide a common terminology for practitioners, boards, and regulatory bodies to assess AI agents based on their autonomy, functions, learning capabilities, and governance structures. Investment leaders will benefit from understanding how to create an AI taxonomy tailored to their firms and the necessary steps for mapping out their deployed AI agents.
The risks of a vague taxonomy
Without a universally accepted taxonomy, there is a risk of both over-relying on and under-utilizing a technology that is already transforming capital allocation strategies. This could lead to unforeseen complications in the future.
A well-structured AI taxonomy should not stifle innovation. Instead, if crafted thoughtfully, it should empower firms to define the problems that AI agents address, clarify accountability, and outline how model risk is minimized. Lacking this clarity can result in a tactical rather than a strategic approach to AI adoption.
Functional vs. systemic approaches to AI integration
Investment managers typically approach AI in one of two ways: as a set of functional tools or as an integrated component of the overall investment process. The functional perspective utilizes AI for tasks such as risk assessment, sentiment analysis using natural language processing, and summarizing portfolio exposures. While this enhances efficiency and consistency, it does not fundamentally alter the decision-making architecture, maintaining a predominantly human-centric organization with AI serving as a supplementary enhancement.
Conversely, many firms are adopting a systemic approach, integrating AI agents into the investment design process as active participants rather than mere tools. In this model, the roles of autonomy, learning capacity, and governance are explicitly defined, fostering a decision-making ecosystem where human insights and machine reasoning work in tandem.
Creating a smarter investment organization
The distinction between functional and systemic approaches is crucial. While functional adoption may produce faster tools, systemic adoption leads to more intelligent organizations. Both approaches can coexist, but it is the latter that offers a long-term competitive advantage.
As noted by neuroscientist Antonio Damasio, all forms of intelligence seek to maintain homeostasis—a balance with their environment. The financial markets are intricate adaptive systems that similarly require equilibrium between data and judgment, automation and accountability, as well as profitability and sustainability. An effective AI framework would reflect this complexity by categorizing AI agents along three critical dimensions.
Key dimensions for AI classification
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?0
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?1
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?2
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?3
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?4
Unlike traditional models such as ChatGPT, which primarily respond to queries, agentic AI offers more advanced capabilities. These agents can observe, analyze, and make decisions, sometimes executing actions on behalf of humans within specific parameters. This raises a critical question for investment firms: Should AI be regarded as a decision-support tool, an autonomous research analyst, or a delegated trader?5