Building an AI Agent That Searches the Web and Makes Investment Decisions

2026-06-29 · Source: Andrej Baranovskij · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Intermediate, short

Summary

The "Sparrow Agent Implementation: Part Two" details an AI agent designed to assist with investment decisions by integrating web search and a local LLM. This agent processes investment positions, enriching initial data from a JSON file with "holding years" and "annual income" to provide a comprehensive risk analysis. It then utilizes a Tab B client to search the web for additional position information, saving these results to ensure reproducibility. Finally, all gathered data is fed into a local Gemini 4 LLM, which, guided by a prompt favoring selling after long holding periods (e.g., 8 years), recommends either "hold" or "sell" for each position, often providing reasoning. The system highlights Sparrow's agentic capabilities, allowing for chained instructions beyond simple data extraction, though emphasizing that human verification of decisions is crucial.

Key takeaway

For AI Engineers developing financial decision-making agents, integrating external web search with local LLMs like Gemini 4 offers enhanced contextual analysis. You should design workflows combining diverse data sources, like holding years and annual income, with real-time web information. This refines risk assessments. Implement result caching for web searches to ensure decision reproducibility and always include human oversight for critical investment recommendations.

Key insights

An AI agent integrates web search and LLM analysis for nuanced investment hold/sell decisions.

Principles

• Combine diverse data for better risk assessment.
• Reproduce web search results for consistency.
• LLMs can provide decision reasoning.

Method

The agent extracts data, calculates risk, searches the web via Tab B client, saves results, then feeds all to a local Gemini 4 LLM for hold/sell decisions.

In practice

• Use Tab B client for structured web searches.
• Save search results to a JSON file.
• Prompt LLMs to favor specific outcomes (e.g., selling).

Topics

• AI Agents
• Investment Decisions
• Web Search Integration
• LLM Applications
• Risk Analysis
• Gemini 4

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• IPO Finance Agent: Evaluation of LLM Financial Analysts beyond Finance Agent v2, with Automated Rubric Generation -- the Case of the SpaceX (SPCX) IPO — IPO Finance Agent improves LLM financial analysis for IPOs using contextual retrieval and automated rubric generation.
• Tool Calling, Explained: How AI Agents Decide What to Do Next — Tool calling enables LLMs to trigger external actions by deciding which function to call and with what arguments, forming agentic AI workflows.
• Building Search for AI Agents with Exa CEO Will Bryk — AI agents demand a new search paradigm focused on comprehensive, controllable, and deep information retrieval, distinct from human-optimized engines.
• Building an AI Agent That Turns Web Data Into Sales Intelligence — AI agents transform dynamic web data into timely, actionable sales intelligence, overcoming stale CRM data.
• Connecting the Dots with Context Graphs — Stephen Chin, Neo4j — Context graphs integrate LLMs with knowledge graphs to provide grounded, explainable, and auditable AI-driven decision-making by connecting disparate enterprise data.
• How to build Multi Agents for FINANCE: Outperforming Anthropic — Specialized, locally runnable LLMs can outperform large proprietary models in domain-specific tasks through targeted training and novel evaluation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Andrej Baranovskij.