“Weekly revenue reporting takes 8+ hours of manual SQL pulls, Excel reconciliation, and narrative writing. This agent does it in under 2 minutes.”
Problem → Approach → Results
Manual weekly revenue reporting consumed 8+ hours per cycle. Finance teams were spending more time assembling data than interpreting it. Leadership visibility was delayed by days.
Built a simulated multi-product revenue dataset. Wrote SQL pipelines using CTEs and window functions to compute actuals vs. plan, week-over-week delta, and 4-week rolling trends.
Engineered a structured prompt that passes variance data to the Claude API and returns a 3-sentence exec summary, top 3 drivers, key risk flag, and recommended action.
Orchestrated the end-to-end pipeline in n8n: weekly trigger → data pull → Python analysis → Claude API call → formatted report output.
“The goal wasn't to replace the analyst. It was to give the analyst their week back.”