AI Economist Agent: An Agentic Framework for Model-Grounded Economic Analysis with RAG, Knowledge Graphs, and Large Language Models

Researchers propose an AI economist agent that combines large language models with knowledge graphs and retrieval-augmented generation (RAG) to produce grounded economic analyses. Rather than relying solely on LLM-generated narratives, the framework grounds economic claims in explicit model-based computations and retrieved evidence, tested on inflation analysis and bank stress-testing scenarios.

The development of an AI economist agent represents a meaningful advance in applying machine learning to economic analysis, addressing a fundamental limitation of current LLMs: their tendency to generate plausible-sounding but unfounded claims. By anchoring language model outputs to formal economic models, knowledge graphs, and verifiable data sources, this framework bridges the gap between narrative generation and analytical rigor that economists demand.

The motivation stems from economics' inherent requirement for grounding claims in both theoretical frameworks and empirical evidence. While LLMs excel at synthesizing information and producing coherent narratives, they lack mechanisms to verify consistency with economic theory or ensure factual accuracy. This framework solves that through agentic planning—allowing the system to strategically retrieve relevant evidence, select appropriate quantitative models, and link conclusions back to sources.

The practical applications tested—U.S. inflation persistence analysis and commercial real estate stress-testing narratives—reveal real-world utility in high-stakes domains where analytical credibility matters. Financial institutions, policymakers, and researchers rely on transparent, traceable reasoning; this agent-based approach delivers both fluency and accountability.

For the broader AI and fintech industries, this work signals growing demand for responsible AI systems that operate within domain-specific constraints. Rather than treating LLMs as standalone analytical tools, the framework demonstrates how to orchestrate them as components within larger reasoning systems. This pattern has implications for risk management, regulatory compliance, and institutional adoption of AI in regulated sectors where explainability and accuracy are non-negotiable.

• →AI economist agent combines LLMs with knowledge graphs and RAG to ground economic claims in formal models and data rather than relying on unfounded narratives.
• →Framework employs agentic planning to strategically retrieve evidence, select appropriate models, and generate reports with explicit computational traceability.
• →Successfully tested on high-stakes applications including inflation analysis and bank stress-testing for commercial real estate exposure.
• →Addresses critical limitation of LLMs in regulated industries requiring transparent, verifiable reasoning and theoretical consistency.
• →Demonstrates broader pattern of constraining language models within domain-specific frameworks to improve credibility and institutional adoption.