How to build an AI agent that reads 10-K Risk Factors

This tutorial provides a practical guide to building an AI-powered agent capable of analyzing SEC 10-K risk factor disclosures. Using TypeScript, the Valyu API for data retrieval, and structured output from language models, the agent extracts specific business risks, links them to original source passages, and suggests follow-up investigative questions. The process involves defining a schema for risk assessment, verifying that model-generated quotes exist within the source text, and saving the evidence in a structured JSON format. The article also explores optional extensions, such as using Jev for thesis validation, performing quarterly updates via 10-Q filings, and conducting basic earnings-quality checks. By automating the retrieval and verification of financial disclosures, developers can create more reliable tools for investment research, ensuring that AI interpretations remain grounded in verifiable evidence from official company filings.