Case Study — La Boîte à Grains × Quartier AI
Case Study — La Boîte à Grains × Quartier AI
The Challenge
Collaboration
What Was Deployed
Results
Conclusion
« Before, invoice reconciliation was a task my team did manually, item by item. We knew we were losing money on price discrepancies, but we didn't have the capacity to catch everything. Since we implemented artificial intelligence with Quartier AI, it's automatic. The system detects discrepancies, generates credit requests, and my manager just has to validate them. We've eliminated 80% of the manual task, achieved 100% accuracy, and easily recovered 40 hours per month on just one process, which we're reinvesting elsewhere in the store. »
Alexandre Savard
Owner - La Boîte à Grains
La Boîte à Grains has been an Outaouais institution since 1980. As a natural and organic grocery store, the company operates several retail locations in Gatineau and delivers across Quebec and Ontario. With dozens of active suppliers and promotional flyers planned up to a year in advance, purchasing management is complex — and negotiated prices change with each flyer.
The problem: suppliers don't always invoice at the agreed-upon promotional prices. And without a system capable of verifying everything, these discrepancies went unnoticed.
The process relied entirely on manual verification. Upon receiving each invoice, the team compared unit prices between the Excel purchase order and the invoice PDF, item by item. In case of a discrepancy, a paper credit request form was filled out and sent to the supplier. The purchasing and finance teams spent approximately 40 hours per month solely on these reconciliation and follow-up tasks — time that the increasing volume of flyers made it harder and harder to absorb.
The financial stakes were real, but invisible: no one knew precisely how much money was leaving the organization each month in unclaimed discrepancies.
Quartier AI began with a one-month strategic audit — including comprehensive process mapping, identification of friction points, and definition of solution criteria. The diagnosis was clear: the problem wasn't human, it was structural. The verification burden needed to be removed, not the people.
The solution came in the form of BAG OS, a custom-built extension of EntrepriseOS tailored for La Boîte à Grains' purchasing operations. Development and deployment were completed in two months, making it a total three-month project from audit to production.
The chosen approach incorporates a human-in-the-loop logic at the final stage: the agentic system performs the analysis, detects discrepancies, and generates credit requests — but it's the manager who validates and sends them. No blind automation. A final human check ensures the quality of exchanges with suppliers, while eliminating the tedious verification task.
BAG OS centralizes purchase orders imported as CSV or XLSX, automatically monitors the dedicated invoice email inbox, extracts key data from each PDF received via agentic OCR, cross-references information in real-time, and generates credit documents in a standardized format — invoice number, unique identification per store, detail by brand, UPC, invoiced amount, and credit amount to be claimed.
Each incoming invoice immediately triggers a complete processing workflow, with no human action required.
All key data from each PDF invoice is read, structured, and associated with the correct supplier and store in seconds.
The cross-referencing occurs in real-time, in sync with supplier deliveries, without accumulation or delay.
Standardized format, unique identifier per store, brand-level detail with invoiced amounts and credits to claim. Ready to send with one click.
Each incoming invoice immediately triggers a complete processing workflow, with no human action required.
The manager receives an alert, reviews the generated document, validates it, and forwards it to the supplier. The resolution cycle is reduced from several days to just a few minutes.
Within the first week of going live, BAG OS detected discrepancies that the manual process had never uncovered.
Within four weeks, $3,000 to $4,000 in supplier credits were recovered — money that had been silently leaving the organization for months due to the lack of a system capable of verifying everything. Based on these initial results, the annual projection for credit recovery is between $30,000 and $50,000 (estimated).
The 40 monthly hours freed up — recovered from this single reconciliation and credit request management process, identified during the strategic audit — are being reallocated to higher-value tasks. The reconciliation error rate is zero: detection is automatic, systematic, with no possibility of omission. And the time-to-value was immediate: the first measurable results appeared on the very day of deployment into production.
Beyond the numbers, BAG OS made the invisible visible. La Boîte à Grains now knows exactly what its suppliers are billing them for — and what they owe them.
Food retail businesses that manage complex promotional purchases almost always underestimate the true cost of manual verification, both in lost time and unclaimed money. It's often only when a system takes over that the true extent of the problem becomes measurable.
Does your organization manage promotional purchases with multiple suppliers? The discrepancies you don't see are likely costing you more than you think.