An omnichannel shoe retailer's owners spent every Monday assembling numbers instead of acting on them. We built a pipeline that unifies POS and web data and has AI write a plain-language trading brief, delivered on WhatsApp at 7:00 — eight hours of weekly reporting, gone.
Typical results
0 hrs
Manual reporting weekly
7:00
Monday delivery time
4
Channels unified
11
Slow movers flagged monthly
Figures reflect typical results for this kind of automation.
The challenge
Sales lived in four places: three store POS systems and a web shop. Every Monday one of the owners exported reports, stitched spreadsheets and hunted for what changed — roughly eight hours before any decision got made. Insights arrived Wednesday at best, slow-moving stock was noticed a season late, and reorders ran on gut feel.
Before and after
Weekly reporting effort
−100%Lag from week close to insight
−90%Slow movers caught monthly
5.5×The automation
We connected all four channels to a single data flow that runs every Sunday night: sales, margin and stock roll up per SKU, sell-through is computed, and slow movers and reorder candidates are ranked. OpenAI then writes the week as a story — what sold, what stalled, what to reorder and why — in the owners' language. The brief hits their WhatsApp at 7:00 Monday, with a linked sheet carrying every figure behind it.
How it works
- 01
Sunday-night sync
Sales, margins and stock pull from three store POS systems and the web shop.
- 02
Crunch
Sell-through, margin and velocity are computed per SKU across all channels.
- 03
AI narrative
OpenAI writes the week in plain language: what sold, what stalled, what to reorder.
- 04
Monday 7:00
The brief lands on the owners' WhatsApp before the stores open.
- 05
Drill down
A linked sheet carries every figure behind the story for deeper digging.
Starts from
- 3-store POS
- Web shop sales
Delivers
- WhatsApp owner brief
- Reorder sheet
- KPI dashboard
The impact
Reporting effort dropped from eight hours a week to zero, and the lag from week close to insight fell from 72 hours to seven. The AI now surfaces around 11 slow movers a month, where manual review caught maybe two, and reorder decisions are made Monday morning with data instead of instinct.
Questions about this automation
Where does the data for the weekly brief come from?
Every Sunday night the flow pulls sales, margins and stock from each store's POS system and from the web shop. Products are matched by SKU across channels, so each product is counted once, and sell-through, margin and sales velocity are calculated per SKU before any writing begins.
How does the AI decide what to highlight?
The numbers are calculated first, so the AI does not invent figures. Slow movers and reorder candidates are ranked by fixed rules, then OpenAI receives those results and writes them up in plain language: what sold, what stalled, what to reorder and why. Every figure in the brief also appears in the linked spreadsheet.
Can the owners check the numbers behind the story?
Yes. Each brief links to a spreadsheet with the full SKU-level figures for that week, by store and by channel, so any statement can be traced in a few clicks. The spreadsheet keeps past weeks for comparison. Simple automations like this are usually live within days.