FinSight: Towards Real-World Financial Deep Research
Jiajie Jin, Yuyao Zhang, Yimeng Xu, Hongjin Qian, Yutao Zhu, Zhicheng Dou
2025-10-23
Summary
This paper introduces FinSight, a new AI system designed to automatically create professional-quality financial reports, something current AI struggles with.
What's the problem?
Creating good financial reports is really hard work, requiring both detailed data analysis and clear presentation. Existing AI systems aren't capable of handling all the steps – gathering information, analyzing it, and then putting it into a well-written and visually appealing report – to the level a human financial expert can.
What's the solution?
The researchers built FinSight, which uses a system of different AI 'agents' working together. One key part is an agent that can write and execute code to collect and analyze data. They also developed a way to automatically improve charts and graphs to make them look professional, and a writing process that builds detailed reports with supporting evidence. Essentially, they created a system that mimics how a human analyst would approach the task, but with AI.
Why it matters?
This work is important because it brings us closer to fully automating the creation of financial reports. This could save companies a lot of time and money, and potentially make financial information more accessible. The fact that FinSight performs better than other AI systems suggests it’s a significant step towards AI-generated reports that are as good as those created by human experts.
Abstract
Generating professional financial reports is a labor-intensive and intellectually demanding process that current AI systems struggle to fully automate. To address this challenge, we introduce FinSight (Financial InSight), a novel multi agent framework for producing high-quality, multimodal financial reports. The foundation of FinSight is the Code Agent with Variable Memory (CAVM) architecture, which unifies external data, designed tools, and agents into a programmable variable space, enabling flexible data collection, analysis and report generation through executable code. To ensure professional-grade visualization, we propose an Iterative Vision-Enhanced Mechanism that progressively refines raw visual outputs into polished financial charts. Furthermore, a two stage Writing Framework expands concise Chain-of-Analysis segments into coherent, citation-aware, and multimodal reports, ensuring both analytical depth and structural consistency. Experiments on various company and industry-level tasks demonstrate that FinSight significantly outperforms all baselines, including leading deep research systems in terms of factual accuracy, analytical depth, and presentation quality, demonstrating a clear path toward generating reports that approach human-expert quality.