Data Analyst
Analyzes data the operator provides — CSV, Excel, a connected Google Sheet, or pasted tables — to find real patterns, trends, and outliers, answer business questions with verified statistics, and build reports, charts, and dashboards. Every number is computed with code (never eyeballed), nothing is fabricated, correlation is never sold as causation, and data-quality caveats (missing values, small samples, outliers) are surfaced honestly. It has no BI tool, data warehouse, or product-analytics connector — the operator exports from their tools and it does the analysis.
看看它如何工作
看看这位 AI 助手可以怎样主动帮你处理工作，而不是把麻烦交回给你。
公司
A subscription box startup
场景
The founder pastes last quarter's monthly sales and wants to know why one month dipped
它可以帮你做什么
- 1Loads the figures with python and computes the month-over-month change for real, leading with the headline drop
- 2Profiles the data first, reporting row count, date range, and any blanks before drawing a conclusion
- 3States the dip as a fact but flags that three months is too short a window to call a cause
- 4Lists the confounders rather than claiming the price change caused the fall, since it is only correlated
- 5Asks for the underlying daily rows by product or channel so the next pass can pinpoint where the fall happened
公司
An online clothing shop
场景
The owner wants to know which customer segment is most valuable from an exported orders CSV
它可以帮你做什么
- 1Cleans and dedupes the order rows, then groups and ranks segments by total spend with pandas
- 2Shows the breakdown as a paste-ready table directly in the chat with the computed numbers
- 3Flags which segments have too few rows to trust, warning a 3x gap on a tiny sample is noise
- 4Renders a bar chart as an image and offers to drop it in Google Drive or send it over Telegram, since chat cannot show it inline
- 5Masks customer names and emails in any shared version, reporting patterns by segment instead of individual rows
公司
A small SaaS team
场景
The operator asks the agent to pull their conversion funnel straight from GA4
它可以帮你做什么
- 1Explains honestly that it has no GA4 connector and cannot see the events directly
- 2Refuses to invent a conversion number to fill the gap rather than fabricating a plausible figure
- 3Tells the operator exactly which GA4 report to export as a CSV so it can compute the funnel for real
- 4Once the export arrives, computes the drop-off at each step and flags where the biggest leak is
- 5Offers to re-run the same analysis on a weekly schedule once Google is connected, noting it is a scheduled check not real-time