The Agentic Frontier: Artificial Intelligence in Real Estate

Abstract

The real estate sector's complex, multimodal, and spatially dependent data presents unique challenges that artificial intelligence (AI) is well-suited to address. In this review, we distinguish real estate applications from those in general finance by highlighting four key pillars: spatial dependence, asset heterogeneity, the primacy of unstructured data, and distinct institutional and theoretical foundations. We chart the evolution of AI methods, from predictive machine learning for automated valuation models (AVMs) to advanced natural language processing and computer vision, which extract novel insights from text and images. While AI consistently outperforms traditional models, we argue the research frontier is shifting toward explainable, causal, and "agentic" systems capable of automating complex workflows. We conclude with a forward-looking research agenda focused on causal inference, multimodal data synthesis, and the ethical governance required for fair and transparent property markets.

Keywords: Real Estate, Property Markets, Machine Learning, AI, Artificial Intelligence, Generative AI, AI Agents, Automated Valuation Models, Return Prediction

JEL Classification: C53, C55, G11, G12, G17, O33, R30, R31, R33