Company Research Agent
Harness the power of Tavily’s APIs to perform in-depth company research.
Why Use Tavily for company research?
This is one of the most popular use cases for Tavily. Our powerful APIs can easily be integrated with agentic workflows to perform in-depth, accurate company research.
Tavily offers several advantages for conducting in-depth company research:
-
Comprehensive Data Gathering: Tavily’s advanced search algorithms pull relevant information from a wide range of online sources, providing a robust foundation for in-depth company research.
-
Flexible Agentic Search: When Tavily is integrated into agentic workflows, such as those powered by frameworks like LangGraph, it allows AI agents to dynamically tailor their search strategies. The agents can decide to perform either a news or general search depending on the context, retrieve raw content for more in-depth analysis, or simply pull summaries when high-level insights are sufficient. This adaptability ensures that the research process is optimized according to the specific requirements of the task and the nature of the data available, bringing a new level of autonomy and intelligence to the research process.
-
Real-time Data Retrieval: Tavily ensures that the data used for research is up-to-date by querying live sources. This is crucial for company research where timely information can impact the accuracy and relevance of the analysis.
-
Efficient and Scalable: Tavily handles multiple queries simultaneously, making it capable of processing large datasets quickly. This efficiency reduces the time needed for comprehensive research, allowing for faster decision-making.
How does it work?
We have prepared a Jupyter Notebook that demonstrates how to run a weekly research process on companies using Tavily, the LangGraph framework, and OpenAI for content generation.
This notebook outlines a comprehensive workflow that dynamically gathers relevant information on a company, processes the data, and generates a detailed PDF report.
The notebook utilizes several components to achieve its goal through a structured set of nodes representing different stages of the research process:
You can view examples of company reports generated by the code in the notebook here.
Possible Improvements
-
Advanced Content Generation Filtering: Enhance the workflow by incorporating more advanced filtering techniques, such as selecting the top K most relevant documents using relevance scoring or similarity measures. This pre-content generation filtering step, combined with keyword-based filtering, ensures that only the most pertinent information is used, allowing the content generation step to focus solely on producing high-quality, accurate reports without the burden of additional filtering.
-
“Human on the Loop” Feedback Integration: Allow a human to provide feedback on the sources found during research and guide the agent to retrieve the most relevant sources. Instead of predefining keywords before the execution of the process, enable the dynamic inclusion or exclusion of keywords and other adjustments that only a human can provide when needed, improving the user experience as well as the quality and accuracy of the sources used.
-
Advanced Curation Step: Instead of adding raw content to the RAG documents, include only the relevant information extracted from the raw text. This can be achieved through live chunking or using an LLM to extract or summarize the pertinent information, ensuring that the content fed into the generation step is highly focused and relevant.
-
Customized Workflow for Specific Needs: Tailor the workflow to meet specific requirements by defining a precise report format or prioritizing the use of particular sources. For instance, you can specify that only certain trusted domains should be used for data retrieval or create a structured template that the generated reports must follow. This customization enhances the relevance and precision of the research output, ensuring it aligns closely with your unique needs.