Tencent Beacon AI Officially Opens to the Public: Complete Data Analysis with a Single Sentence

Tencent recently announced that its AI data analytics platform "Tencent Beacon" is now officially open to the public, and both enterprise and individual users can register and use it via the official website beacon.qq.com. Over the past decade, Tencent Beacon focused on self-service analytical insights and data visualization for Tencent's internal businesses. This time, the brand has been upgraded to an AI data analytics and insights suite, offering business teams one-stop capabilities such as natural language data querying, dashboard generation, attribution analysis, metric monitoring, and team collaboration.

After more than ten years of technological accumulation and product iteration, the newly upgraded Beacon AI is now widely used across more than 100 internal Tencent business teams, including QQ, Tencent Video, Tencent News, Yingyongbao, WeChat Pay, Yuanbao, Tencent Customer Service, and TiMi Games, accumulating rich experience in data analysis and day-to-day business operations analysis.

Compressing a business dashboard from "one week" to "hours"

In the traditional model, when business staff want a dashboard, they need to go through steps such as submitting requirements, waiting for scheduling, extracting data, creating charts, and adjusting formats, with a cycle that usually takes about a week; if the metric definition changes, they have to be scheduled again. Beacon AI's approach is to let business staff complete this chain themselves: describe requirements in natural language, and AI directly generates the dashboard; subsequent modifications, follow-up questions, attribution, and push notifications are all completed within the same workbench, with no need to write SQL.

In past usage cases, we found that building a full-link operations dashboard required 4 to 6 working days on traditional BI tools. After using Beacon AI, the time from requirement to delivery (including iterative tuning) was no more than 2 hours. For building a business monitoring dashboard, it took only 8 minutes from opening the dialog box to finalizing the dashboard, whereas previously scheduling and waiting usually took 1 to 2 weeks. Beacon AI helps businesses greatly save working time and improve work efficiency.

Data access: Beacon AI supports access to 3 major categories and 10 types of data sources. Among them, databases and data warehouses support MySQL, StarRocks, PostgreSQL, and ClickHouse; files and online documents support local Excel/CSV, Tencent Docs, WeCom Smart Sheets, Feishu Docs, and more. In addition, custom SQL can be used to define calculated views that are not stored in the database. When connecting to a database, you can configure display names for fields, automatically extract data types, and set business semantics such as "dimension/metric/time dimension" to help AI accurately understand metric definitions.

Dashboard generation: One sentence produces a chart, with the process visible in real time. In the dialog box, users "@dataset" and describe requirements, for example, "conversion funnel by channel over the past 30 days," and AI generates in a streaming manner in the order of "understand requirement → match fields → select chart → render dashboard." The platform supports dozens of chart types, including KPI metric cards, line charts, bar charts, Sankey diagrams, sunburst charts, box plots, heatmaps, funnel charts, candlestick charts, and data tables, and automatically recommends charts based on field types and data volume.

Editing and adjustment: Three convenient ways to modify, so dashboards are no longer "one-time." Conversational modification—after enabling "AI Picker," click any chart element and explain the change in one sentence; element selection editing—directly adjust styles and associated fields, and view and modify the SQL generated by AI behind the chart; code editing—directly modify the dashboard HTML source code and embed custom components. Dashboards retain complete version history and support version comparison and one-click rollback.

Attribution analysis: Automatically break down the causes of metric anomalies. Users click an anomalous chart card and directly ask follow-up questions, for example, "Why did sales increase on November 24?" Beacon AI can then generate a visualized attribution dashboard with one click. A product operations staff member for livestream e-commerce governance said that in the past, investigating a metric anomaly required manually writing 5 to 8 SQL queries for cross-validation, taking 3 to 4 hours; now it can be completed in about 10 minutes.

Monitoring and push notifications: Let data find people proactively. Beacon AI's intelligent alerts support three types of detection: "metric threshold, empty data, and query failure." Detection frequency can be set from every minute to every month. You can select only a local area of the dashboard for a screenshot, which is delivered on a schedule via WeCom or email. Each push retains the execution time, channel, status, and content snapshot for traceability.

Analysis results are reusable, moving from one-off querying to daily operations

Dashboards generated by Beacon AI are dynamic artifacts, and data is automatically refreshed according to the configured strategy, so daily viewing does not require calling the large model again each time. A dashboard can be converted with one click into a knowledge document with explanatory text for reporting, or saved as a template, truly retaining analytical capabilities instead of starting from scratch every time.

In terms of capability expansion, the platform comes with 12 data analysis skills and 3 visualization skills, supports installing more skills from the skill marketplace, uploading team-built skill packages, and connecting to enterprise internal systems such as knowledge bases and R&D collaboration platforms through the MCP protocol.

Enterprise-level governance: business autonomy, asset accumulation, and management with boundaries

Team asset accumulation, autonomous learning, and evolution: Self-Harness turns feedback, error correction, and effective methods during task execution into reusable experience, continuously improving analysis paths and execution strategies. From understanding business context to adapting to individual preferences, every collaboration becomes the starting point for better performance next time, making AI understand the business better and become more adept with use.

In terms of data security, Beacon AI processes metadata and real data separately. During large model inference, it only touches metadata such as field names and data types. Real values are dynamically replaced and backfilled through placeholders and are not read by the model, ensuring privacy and security. At the same time, every analysis step in Beacon AI can retain a full-link audit log, and key figures are traceable and reviewable.

For work permissions, Beacon AI's team space provides three permission modes: management, editing, and viewing, which can be refined to folders and individual files. Team administrators can manage members, limit the available AI models, and hide files that members are not authorized to view. The enterprise edition also supports private deployment and sensitive data probes.

Tencent Beacon AI is committed to helping business staff independently complete daily data querying, data analysis, attribution insights, dashboard creation, and team collaboration using natural language. A data analyst summarized after using Beacon AI for 70 days: in the past, 80% of the time was spent writing SQL, tuning SQL, taking screenshots for PPTs, and aligning metric definitions; now 80% of the time is used to look at insights, discuss causes with the business, and drive decisions. Starting today, users can log in to beacon.qq.com to register and use it. New users receive 500 points upon first registration, and 500 points are granted each month thereafter. Let more people be freed from "moving data" and give time back to thinking—start with Beacon AI.