The landscape of investment management is undergoing a significant transformation due to the rise of artificial intelligence (AI). As AI technologies increasingly become integral to the daily operations of portfolio managers and analysts, it is essential for firms to clearly define and understand the types of AI they are utilizing. This clarity not only enhances operational efficiency but also ensures proper governance and risk management.
In this context, agentic AI emerges as a pivotal component. Unlike conventional models such as ChatGPT, agentic AI possesses the capability to observe, analyze, and make decisions within established parameters, often acting on behalf of human operators. Investment firms face a critical choice: will they view AI as a supportive decision-making tool, an independent research analyst, or a fully autonomous trader?
Establishing a framework for AI classification
Every implementation of AI provides an opportunity to define its boundaries and applications. The ability to classify AI systems is paramount; without this classification, governance becomes challenging and scalability is hindered. To assist in this endeavor, our research team, in collaboration with DePaul University and Panthera Solutions, has developed a multi-dimensional classification system specifically tailored for AI agents in investment management.
This classification framework offers investment professionals, boards, and regulators a standardized language to assess agentic systems based on several criteria, including autonomy, function, learning capacity, and governance. By understanding these dimensions, leaders in the investment sector can better design AI taxonomies and create a framework for mapping the deployment of AI agents within their organizations.
The risks of a lack of taxonomy
The absence of a shared taxonomy can lead to mismanagement of AI technologies, resulting in both over-reliance and underutilization. Such missteps could complicate the already intricate processes of capital allocation. A well-structured AI taxonomy should not stifle innovation; rather, it should enable firms to articulate the specific problems that each AI agent addresses, identify accountability, and outline strategies for mitigating model risk.
Two approaches to AI integration in investment management
Investment managers typically adopt one of two approaches when integrating AI into their workflows: treating AI as a functional toolset or embracing it as a systemic component of the decision-making process. The functional approach primarily involves utilizing AI for tasks such as risk assessment, sentiment analysis via natural language processing, and portfolio analysis summaries. While this enhances efficiency, the fundamental decision-making framework remains predominantly human-centric, placing AI in a supportive role.
Conversely, a growing number of firms are opting for the systemic approach. In this model, AI agents are woven into the investment decision-making process as active participants rather than mere tools. Here, aspects such as autonomy, learning capabilities, and governance are explicitly defined, resulting in an ecosystem where human judgment and machine intelligence coexist and evolve together.
Understanding the implications of each approach
Recognizing the distinction between these two approaches is crucial. While functional-driven adoption provides quicker tools for analysis, a systemic approach fosters smarter organizational behaviors. Although both strategies can coexist, only the latter has the potential to deliver sustained competitive advantages in the investment landscape.
Mapping AI agents within the investment framework
To effectively integrate AI, it is vital to consider how these agents operate within the investment process. The investment value chain typically consists of five stages: idea generation, assessment, decision-making, execution, and monitoring, all of which are further entwined with compliance and stakeholder reporting. AI agents can enhance each of these stages, but it is crucial that decision-making rights remain aligned with their interpretability to avoid governance oversights.
Additionally, it is essential to evaluate the comparative advantage that AI agents bring to investment firms. This includes identifying whether they enhance informational, analytical, or behavioral advantages. While AI may not directly generate alpha, it can amplify existing strengths, allowing firms to align agent types with their unique skill sets.
Addressing uncertainty with AI
In this context, agentic AI emerges as a pivotal component. Unlike conventional models such as ChatGPT, agentic AI possesses the capability to observe, analyze, and make decisions within established parameters, often acting on behalf of human operators. Investment firms face a critical choice: will they view AI as a supportive decision-making tool, an independent research analyst, or a fully autonomous trader?0
In this context, agentic AI emerges as a pivotal component. Unlike conventional models such as ChatGPT, agentic AI possesses the capability to observe, analyze, and make decisions within established parameters, often acting on behalf of human operators. Investment firms face a critical choice: will they view AI as a supportive decision-making tool, an independent research analyst, or a fully autonomous trader?1
In this context, agentic AI emerges as a pivotal component. Unlike conventional models such as ChatGPT, agentic AI possesses the capability to observe, analyze, and make decisions within established parameters, often acting on behalf of human operators. Investment firms face a critical choice: will they view AI as a supportive decision-making tool, an independent research analyst, or a fully autonomous trader?2