Luxury fashion retail · Asia · 2 brands, 8 regions
AI-augmented analytics & BI for a luxury fashion retailer
Weekly reports turned daily, and advanced analytics opened up to non-experts — without sensitive data ever leaving the company.
Proves · Applied AI + Data + Automation
- Daily
- reports (was weekly)
- 8
- regions
- 2
- brands
- In-house
- AI keeps data private
Challenge
Reporting was manual and slow: teams across retail, finance, merchandising and CRM could only get key reports about once a week, and deeper analysis — segmentation, product affinity — needed a scarce data specialist. Bringing in AI was tempting, but sending sensitive customer and sales data to external AI tools was a non-starter.
Approach
We started where AI projects usually fail — the data. After heavy cleaning and modelling into a reliable, well-structured foundation, we taught the AI our data schema, never the real data. From the schema it writes Python to answer a business question and return exactly what's needed: an Excel extract, a customer list, a dashboard or a report. A local AI model (Qwen) runs the first pass — summaries, insights, suggestions — on real data in-house, so nothing sensitive leaves the company. The AI-written code runs on AWS Lambda, outputs land in S3, and Power Automate delivers reports on schedule to inboxes and SharePoint.
Outcome
Reporting went from weekly to daily across 2 brands and 8 regions, with fewer errors thanks to less manual handling. Just as important, it freed analysts from rebuilding reports to do genuinely deeper work — and put analyses that once required a data scientist (RFM segmentation, clustering, product association) into business teams' hands. CRM can now reach the right customers more precisely; merchandising uses product affinity to shape promotions and ranging.
Privacy-safe by design
Cloud AI only ever sees the data schema; a local model (Qwen) handles first-pass analysis on real data in-house. Sensitive data never leaves the company.
Clean data first
The longest, most important step. AI is only as good as the data under it — so the foundation came before any model.
Enterprise-grade automation
AI-written Python on AWS Lambda → S3 → scheduled delivery via Power Automate to email and SharePoint.
Adoption & enablement
Tiered training for nearly the whole company, a Super User per team, SOPs for first-level support, and regular Super User exchange sessions — co-ordinated with the French HQ and Asia teams.