📄 中文摘要
数据产品通过提供支持资产（如示例问题-SQL对）使最终用户能够更深入地理解其数据。然而，制作有用的数据产品具有挑战性，通常需要领域专家手工制作支持资产。该研究提出了一种系统，通过在持续优化循环中操作的专门AI代理，自动化数据产品的改进。该系统通过提出问题、监控多维质量指标以及支持人机协作控制，将数据转化为可观察和可精炼的资产，从而在自动化与信任和监督之间取得平衡。
数据产品优化的自主控制中心
出处: Agentic Control Center for Data Product Optimization
发布: 2026年3月12日
📄 English Summary
Agentic Control Center for Data Product Optimization
The proposed system automates the improvement of data products through specialized AI agents operating in a continuous optimization loop. By surfacing questions, monitoring multi-dimensional quality metrics, and supporting human-in-the-loop controls, it transforms data into observable and refinable assets. This approach addresses the challenges of producing useful data products, which typically require domain experts to handcraft supporting assets. The system enables end users to gain greater insights by providing supporting assets, such as example question-SQL pairs that can be answered using the data or views over the database tables, thereby balancing automation with trust and oversight.