p_01 · Sales · Microsoft Fabric
AI Sales Analytics
Fabric lakehouse to an interactive D3 dashboard. Analysis, design and code produced by Claude Code agents that I directed and reviewed.
- Microsoft Fabric
- OneLake
- T-SQL
- Python
- D3.js
- Claude Code
01 / Problem
Show that an end-to-end analytics product, from governed cloud data to a dashboard people can use, can be built quickly with AI agents doing the drafting and a BI engineer owning the decisions.
02 / Data
- Source table dbo.sample_data in a Microsoft Fabric Lakehouse on OneLake: Microsoft's public financial sample, 700 rows.
- Read through the SQL analytics endpoint with T-SQL, signing in with Microsoft Entra ID.
03 / Approach
- Python pipeline (mssql-python, azure-identity) exports and shapes the table.
- A build script combines the page source and the CSV into one self-contained HTML file.
- Claude Code agents drafted the analysis, dashboard design and code. I set the questions and reviewed the output.
- Dashboard: headline KPIs vs the same months of 2013, auto-picked insights, monthly trend, breakdowns, a country by product heatmap, year over year and a searchable data table.
stage_1
OneLake Lakehouse
dbo.sample_data
stage_2
SQL analytics endpoint
T-SQL · Entra ID
stage_3
Python pipeline
mssql-python · azure-identity
stage_4
D3 dashboard
GitHub Pages
04 / Live dashboard
preview · d3 dashboard
05 / Business impact
- A working Fabric to dashboard pipeline, refreshable with three commands.
- Works on phones, with light and dark modes.
- Shows the way I work: AI speeds up delivery, and I stay accountable for every number.
06 / Tools
- Microsoft Fabric
- OneLake
- SQL analytics endpoint
- T-SQL
- Python
- D3.js
- Claude Code