Tencent has made its artificial intelligence-based data analysis platform, Tencent Beacon, available to external companies and private users. Users can register on the website beacon.qq.com. Previously, Beacon was used exclusively by internal Tencent teams for self-service data analysis and visualization for about ten years.
The platform has now been completely redesigned into a comprehensive set of AI-powered data analysis tools. This allows business employees to ask questions in plain language, create dashboards, track the reasons for metric changes, configure monitoring, and collaborate without needing to write SQL code.
Tencent reports that the updated Beacon AI is used by over one hundred internal teams, including QQ Studios, Tencent Video, Tencent News, WeChat Pay, Yuanbao, and TiMi game studios. The company provided examples where an end-to-end operational dashboard, which traditionally required four to six working days in standard Business Intelligence (BI) tools, was created in less than two hours. Furthermore, a monitoring panel was completed in just eight minutes after the user opened the chat.
Beacon AI integrates with three categories and ten types of data sources. These include MySQL, StarRocks, PostgreSQL, and ClickHouse databases, as well as local Excel and CSV files, and online documents from Tencent Docs, WeCom smart sheets, and Feishu. Users can tag fields as dimensions, metrics, or temporal dimensions so that the AI interprets them correctly.
When a user mentions a dataset in the chat and describes the desired outcome, the AI begins delivering results as it understands the request, matching fields, selecting charts, and displaying the dashboard. Many chart types are supported. Dashboards can be edited by pointing to an element and describing the required change, correcting the AI-generated SQL, or directly modifying the HTML dashboard, while maintaining version history and rollback capability.
For attribution purposes, users can click on an unusual chart and ask for the reason behind a numerical value change. The alerting system can trigger when thresholds are reached, data is missing, or queries fail, with notifications sent via WeCom or email. The platform comes with twelve data analysis skills and three visualization skills, supports more functions through a skill marketplace and team-created skill sets, and connects to internal knowledge bases and other systems via the Model Context Protocol (MCP).
A feature called Self-Harness transforms feedback, corrections, and successful methods from past tasks into reusable experience. Regarding data security, Tencent asserts that the model only sees metadata, such as field names and data types, during reasoning. Actual values are filled via placeholders and are not read by the model, and every step is recorded in an audit log. Team spaces offer three levels of permissions, and administrators can restrict the use of AI models for team members. The enterprise version supports private deployment.