In partnership with Open AI
Hospitality · Inventory Intelligence
Bar inventory counted by camera, reordered by message
n AI inventory monitor for bars and restaurants that scans a shelf, estimates what’s left, flags low stock, and opens the reorder conversation automatically.
Bars and restaurants lose margin in the gap between what’s on the shelf and what anyone has time to count. Applore built a proof of concept that replaces the manual stocktake with a camera: staff photograph the shelf, the system recognises each bottle and estimates how much liquid remains, and low-stock items are routed straight into a reorder conversation with the right supplier over WhatsApp. Built fast and deployed to the cloud, the POC covers the full loop — scan, review, detect, reorder — and was put in front of a real operator. The engineering was done in Claude Code, with the vision workload handled by the Gemini API and the app deployed serverless on Google Cloud Run.
The structural limitations
Tracking partial stock during live service is the real difficulty — presence isn’t enough, quantity is the hard part.
- 01
Manual counts are labour-intensive and routinely skipped during service
- 02
Partial bottles make quantity, not just presence, the hard thing to track
- 03
Low-stock items are noticed too late, causing 86s and rushed orders Reordering lives in scattered calls and messages with no system of recor
- 04
No continuous view of stock between formal stocktakes
- 05
Reordering lives in scattered calls and messages with no system of record
How we solved it.
Replace the clipboard with a camera — staff photograph the shelf, the system recognises each bottle and estimates remaining liquid, and low-stock items route straight into a reorder conversation with the right supplier.
Bars lose margin in the gap between what’s on the shelf and what anyone has time to count. Manual stocktakes are slow, skipped on busy nights, and out of date the moment they’re done.
The layers that make the change stick
Intelligence and engineering applied where they drive measurable change.
- 01
Shelf scanning & bottle recognition
A photo identifies each bottle by type and brand, removing manual logging.
- 02
Quantity estimation
Estimates how full each bottle is, so partial stock is counted rather than guessed.
- 03
Inventory review
Recognised stock is presented for quick human confirmation before numbers are committed.
- 04
Low-stock detection & supplier mapping
Items below threshold are flagged and matched to their supplier.
- 05
WhatsApp reorder communication
Reorders are drafted and sent over WhatsApp, turning detection into action.
- Manufacturing · Auto
Maintenance ops, rebuilt for a 105-country manufacturer without disrupting production.
How Applore rebuilt maintenance operations for JK Tyre across 9 plants and 105+ countries — 65% faster task assignment, 70% less manual reporting.
Read case study - Industrial Safety · PPE
From minutes to under three seconds real-time site verification at industrial scale.
Applore built a real-time QR verification platform for KARAM Safety, cutting site credential checks from minutes to under 3 seconds with a digital audit trail.
Read case study - Enterprise SaaS · Field Sales
Kohler's field sales platform CRM and ERP on one stack, shipped in twenty weeks.
Applore shipped Kohler's field sales CRM and ERP on one stack in 20 weeks — full offline support, unified dashboard, 45% faster order processing.
Read case study
Next chapter
Building something this ambitious?
Tell us where the system is breaking. We respond within one business day.
Talk to our team