Integrating AI into Business Intelligence: Practices, Challenges, and Product Architecture

The rapid development of big data and AI is driving a transformation in the Business Intelligence (BI) field, enabling more powerful tools for analysts and decision makers.

Market Trend: BI is evolving from traditional to agile and now to intelligent BI, leveraging AI capabilities such as natural language processing, recommendation algorithms, and AIGC to automate analysis.

Intelligent BI Vision and Implementation Path: Users can ask business questions in natural language, which the system translates into data problems, locates relevant data assets, generates SQL or Python code, and finally produces analysis conclusions.

Business question translation

Data asset location

Query condition generation (SQL/Python)

Analysis conclusion summarisation

AI Applications in Intelligent BI:

Natural language understanding to lower the analysis barrier for non‑technical users.

Retrieval‑augmented generation (RAG) for more accurate knowledge retrieval.

Code generation (SQL, Python) from natural language instructions.

Intelligent reasoning and prediction for anomaly detection, trend analysis, and opportunity identification.

Challenges and Solutions:

Accuracy issues due to insufficient domain‑specific training data and hallucinations.

Maintaining context in multi‑turn conversations.

Complex system interaction and integration costs for seamless user experience.

Industry Cases: Tableau and Power BI adopt Copilot‑style AI assistants; Tencent’s OlaChat demonstrates Text2SQL, natural‑language‑driven drag‑and‑drop analysis, AI‑generated visualisations, and automated data interpretation.

Product Architecture (ABI): An open, modular design allows AI capabilities to be integrated as services or agents across multiple data platforms, supporting flexible deployment of Text2SQL, code correction, and other agents.

Q&A Highlights: Metrics for AI‑driven data products include usage, retention, growth, and accuracy; Text2SQL accuracy is around 75% (correct) and 85% (correct + basic correct); the architecture enables integration with Power BI and custom BI solutions via APIs.

The presentation concludes with a discussion on future product forms that guide users through the entire data workflow, offering intelligent recommendations, multi‑step analysis canvases, and end‑to‑end AI assistance.