Designing Agentic AI-Based Screening for Portfolio Investment

2026-03-24

AI Summary

This study addresses the limitations of traditional portfolio construction methods in effectively integrating fundamental and news sentiment information and adapting to dynamic asset universes. The authors propose a three-layer multi-agent AI framework in which two large language models—one grounded in fundamentals and the other in news sentiment—collaboratively screen high-quality assets and generate trading signals through a negotiation mechanism to dynamically narrow the candidate pool. Portfolio weights are then optimized using high-dimensional precision matrix estimation. Innovatively, the concept of “reasonable screening” is introduced, modeling the number of selected assets as a random variable and proving that the Sharpe ratio remains consistent under mild screening errors. Empirical results on S&P 500 data from 2020 to 2024 demonstrate that the proposed approach significantly outperforms both non-screening benchmarks and conventional screening strategies, achieving higher Sharpe ratios.

Technology Category

• Multiagent Systems: Mechanism Design
• Search and Optimization: Metareasoning and Metaheuristics
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Search and Retrieval-Augmented AI: Agentic search

Abstract

We introduce a new agentic artificial intelligence (AI) platform for portfolio management. Our architecture consists of three layers. First, two large language model (LLM) agents are assigned specialized tasks: one agent screens for firms with desirable fundamentals, while a sentiment analysis agent screens for firms with desirable news. Second, these agents deliberate to generate and agree upon buy and sell signals from a large portfolio, substantially narrowing the pool of candidate assets. Finally, we apply a high-dimensional precision matrix estimation procedure to determine optimal portfolio weights. A defining theoretical feature of our framework is that the number of assets in the portfolio is itself a random variable, realized through the screening process. We introduce the concept of sensible screening and establish that, under mild screening errors, the squared Sharpe ratio of the screened portfolio consistently estimates its target. Empirically, our method achieves superior Sharpe ratios relative to an unscreened baseline portfolio and to conventional screening approaches, evaluated on S&P 500 data over the period 2020--2024.

Problem

• portfolio screening
• agentic AI
• Sharpe ratio
• asset selection
• large language models

Innovation

• Agentic AI
• Large Language Models
• Portfolio Screening
• Precision Matrix Estimation
• Sensible Screening