The age of the AI agent
The investment management industry is at an evolutionary crossroads when it comes to the adoption of artificial intelligence (AI). AI agents are increasingly used in the daily workflows of portfolio managers, analysts and compliance officers, but most companies can’t precisely describe the kind of “intelligence” they have deployed.
Agentic AI (or AI agent) takes large language models (LLMs) many steps further than commonly used models such as ChatGPT. This is not just about asking a question and getting an answer. Agentic AI can observe, analyze, decide, and sometimes act on behalf of a human within defined boundaries. Investment firms must decide: is it a decision support tool, an autonomous research analyst or a delegated trader?
Every adoption and implementation of AI presents an opportunity to set boundaries and delineate the tools. If you can’t classify your AI, you can’t control it and you certainly can’t scale it. To this end, our research team, a collaboration between DePaul University and Panthera Solutions, has developed a multi-dimensional ranking system for AI agents in investment management. This article is an excerpt from an academic article: “A multidimensional classification system for AI agents in the investment industry”, which was recently submitted to a peer-reviewed journal.
This system provides practitioners, boards, and regulators with a common language for evaluating agentic systems based on autonomy, function, learning, and governance. Investment leaders will understand the steps needed to design an AI taxonomy and create a framework for mapping AI agents deployed at their companies.
Without a shared taxonomy, we risk both over-relying and under-using a technology that is already changing the way capital is allocated, which could lead to further complications down the road.
Why a taxonomy matters
AI taxonomy should not limit innovation. If carefully designed, it should allow companies to articulate the problem the agent solves, who is responsible, and how model risk is mitigated. Without such clarity, AI adoption remains tactical rather than strategic.
Investment managers today treat AI in two ways: solely as a functional set of tools or as a systemically integrated part of the investment decision process.
The functional approach includes using AI for risk scoring, natural language processors for sentiment extraction and co-pilots that summarize portfolio exposures. This improves efficiency and consistency, but leaves the core decision architecture unchanged. The organization remains people-oriented, with AI acting as a peripheral amplifier.
A smaller but growing number of companies are following the systemic route. They integrate AI agents into the investment design process as adaptive participants rather than as auxiliary tools. Autonomy, learning capacity and governance are explicitly defined. The company becomes one decision ecosystemwhere human judgment and machine reasoning coexist and evolve.
This distinction is crucial. Function-driven adoption results in faster tools, but systemic adoption creates smarter organizations. Both can coexist, but only the latter provides a sustainable comparative advantage.
Intelligent integration
Neuroscientist Antonio Damasio reminded us that all intelligence strives for homeostasis, equilibrium with its environment. Financial markets are complex adaptive systems (Lo, 2009) and therefore must also maintain a balance between data and judgment, automation and accountability, profit and planetary stability. A smart AI framework would reflect that ecology by mapping AI agents along three orthogonal dimensions:
First think about the investment process: where in the value chain is the agent active?
Typically, an investment process consists of five phases: idea generation, assessment, decision making, execution and monitoring, which are then embedded into compliance and stakeholder reporting workflows. AI agents can expand each stage, but decision-making power must remain proportional to interpretability (Figure 1).
Figure 1.
Mapping agents to the five stages below (Figure 1) clarifies accountability and avoids governance blind spots.
- Generate idea: Perception layer agents like RavenPack transform unstructured text into sentiment scores and event functions.
- Idea assessment: Co-pilots like BlackRock Aladdin Co-pilot surface portfolio holdings and scenario summaries, accelerating insight without removing the human touch.
- Decision point: Decision Intelligence systems (as illustrated by the Panthera Decision GPS scheme above) are designed to build asymmetries between risk and return, based on the most relevant and validated evidence, with the aim of optimizing the quality of decisions.
- Execution: Algorithmic trading agents act within explicit risk budgets, under conditional autonomy and constant supervision.
- Supervision: Agentic AI autonomously tracks portfolio exposures and identifies emerging risks.
In addition to these five phases, this scheme can improve compliance and reporting to stakeholders. AI agents can perform pattern recognition and flag breaches, as well as translate complex performance data into narrative output for customers and regulators.
Second, look at the comparative advantage: which competitive advantage does it increase: informational, analytical or behavioral?
AI does not create Alpha, but can strengthen an existing lead. One method of mapping taxonomy is to distinguish between three archetypes (Figure 2):
- Informational benefit: Superior access or speed of data. Short-lived and easy to trade.
- Analytical advantage: Superior synthesis and inference. Requires own expertise; defensible but time-consuming.
- Behavioral benefit: Superior discipline in exploiting the prejudices of others or avoiding your own prejudices.
Figure 2
Strategic alignment means matching an agent type with a specific skill set of an investor/firm. For example, a quant house can use reinforcement learning for greater analytical depth, while a discretionary company can use co-pilots to monitor the quality of reasoning and maintain behavioral discipline.
Third, evaluate the complexity range: under what degree of uncertainty does it function: from measurable risk to radical ambiguity?
Markets fluctuate between risk and uncertainty. Extending Knight and Taleb’s typologies, we distinguish four operative regimes.
Figure 3
Governance: from ethics to evidence
Upcoming regulations, such as the EU AI Act and the OECD Framework for the Classification of AI Systems, will codify explainability and accountability. A taxonomy that links these mandates to practical governance tools would be considered best practice. A classification matrix then becomes both a risk management system and a strategic compass.
Strategic implications for CIOs
The adaptive nature of the financial sector requires improved intelligence and systems designed to extend, not replace, human adaptability. People contribute to contextual judgment, ethical reasoning, and meaning making; agents contribute to scale, speed and consistency. Together they improve the quality of decisions, the ultimate KPI in investment management.
Companies that design based on decision architecture, rather than algorithms, will increase their advantage.
Therefore:
- Map your ecosystem: Catalog AI agents and plot them within the framework to reveal overlaps and blind spots.
- Prioritize comparative advantage: Invest where AI enhances existing benefits.
- Institutionalize learning loops: Treat every implementation as an adaptive experiment; measures the impact on decision quality, not efficiency.
In practice
Enhanced intelligence, properly classified and governed, makes capital allocation not only faster but wiser, learning as it is allocated. So classify before you scale. Align before you automate. And remember: when it comes to the quality of decision-making, design trumps luck.
#Design #Beats #Luck #Taxonomy #Investment #Firms #Evolve #CFA #Institute #Enterprising #Investor