ChatBI: NetEase AI‑Powered Business Intelligence Platform – Architecture, Technology, and Real‑World Applications

In the era of digital transformation, large language models are reshaping data analysis, and NetEase’s ChatBI integrates AI and BI to deliver a more intelligent, efficient, and low‑threshold analytics experience for enterprises.

The NetEase Data Analytics (BI) platform provides end‑to‑end capabilities including data ingestion from over 40 sources, data preparation via custom SQL or drag‑and‑drop ETL, data modeling with materialized wide tables and relational models, and data application modules such as data tables and the newly introduced ChatBI conversational analytics.

AI brings three main opportunities to data analysis: natural‑language query (NL2SQL/NL2API), multi‑turn interactive Q&A, and advanced insights like chart interpretation and causal inference, while also introducing challenges of hallucination, accuracy, performance, and scenario integration.

ChatBI’s product architecture consists of a front‑end layer, intent‑recognition module, and a core query pipeline (pre‑processing, a self‑developed NL2SQL large model, post‑processing, and the data‑query infrastructure). The NL2SQL model is fine‑tuned on diverse domain data, surpasses GPT‑4 in accuracy, and supports custom functions, UDFs, and complex aggregations.

The post‑processing stage converts generated SQL into a domain‑specific DSL, enabling process verification, user intervention, and rule‑based adjustments, which are then transformed into visual query configurations for rendering.

To meet performance demands of conversational analytics, ChatBI employs materialized view technology that materializes join relationships into single wide tables, supports selective column/row materialization, and integrates with ClickHouse MPP for sub‑second query responses.

ChatBI also offers extensive open capabilities: single sign‑on and token authentication, data‑model synchronization APIs, plug‑in support for third‑party LLMs and RAG systems, and embeddable query APIs, allowing seamless integration into various business systems.

Real‑world deployments include HR chatbot assistants, self‑service data retrieval for NetEase Music, intelligent Q&A for Chongqing Tobacco, and campus analytics for a university, all demonstrating significant improvements in query speed, accuracy, and user adoption.

The presentation concludes with a summary of ChatBI’s technical innovations and its role in advancing AI‑augmented business intelligence.