A leading QSR chain was losing revenue to underbilling across hundreds of outlets. BootLabs deployed a real-time computer vision system on existing CCTV infrastructure that cross-references items served against live POS transactions — and alerts managers within 30 seconds.
A leading quick service restaurant chain was experiencing consistent revenue leakage across its outlet network — food items served at the counter that were never entered into the POS system, either through staff error or deliberate underbilling. Across hundreds of outlets, even small per-transaction losses compounded to a significant revenue gap. Manual auditing was practically impossible: reviewing CCTV footage retrospectively was labour-intensive and always after the loss had occurred. The chain needed a system that could catch discrepancies in real-time, before the transaction window closed. BootLabs deployed a computer vision AI system on the chain's existing CCTV infrastructure that monitors food service in real-time, identifies items being handed to customers, and cross-references against live POS transactions — flagging discrepancies to outlet managers within 30 seconds.
Underbilling — whether through error or intention — was happening across hundreds of outlets simultaneously. Even a small per-outlet daily loss multiplied across 500+ locations represented significant annual revenue erosion, yet no single incident was large enough to trigger traditional audit mechanisms.
Manual CCTV review after shift end could identify what happened but couldn't prevent it or enable real-time intervention. By the time an incident was flagged, the transaction was complete and the revenue was gone. Post-hoc evidence was valuable for HR action but did nothing to recover the immediate loss.
POS data and CCTV footage lived in entirely separate systems with no link between them. Outlet managers had no mechanism to correlate what was served with what was billed without manual, time-intensive review — making systematic loss prevention practically impossible across a network of this size.
Object detection models — fine-tuned on the chain's specific menu items, packaging, and tray configurations — run on edge compute devices connected to existing CCTV cameras at the service counter. No footage leaves the outlet. Inference is local, low-latency, and privacy-preserving by design.
The CV system reads live transaction data from the outlet's POS via API. As items are detected being handed to a customer, the system checks whether equivalent items appear in the current or immediately preceding transaction — within a configurable time window that accounts for natural service flow.
When a mismatch is detected — item served but not billed, or quantity served exceeding quantity billed — a discrepancy event is logged with a confidence score, timestamp, camera frame snapshot, and POS transaction ID. High-confidence events trigger an immediate manager alert; lower-confidence events queue for review.
Outlet managers receive real-time push notifications on their device with a frame capture of the detected discrepancy and the POS comparison. A central dashboard aggregates all outlet-level events for operations and loss prevention teams, with trend analysis by outlet, time of day, and item type.
The system went live on existing CCTV infrastructure — no new cameras, no hardware procurement delays. Edge inference kept latency low enough for real-time intervention while keeping footage entirely on-premise. From day one of deployment, outlet managers were receiving actionable alerts within the transaction window, enabling interventions that simply weren't possible before. The 30-second alert window was the critical design constraint that unlocked real loss prevention rather than just post-hoc reporting.
Real-time alerts allow managers to intervene before transactions close, converting what was previously unrecoverable leakage into recoverable billing corrections at the moment of service.
The presence of AI-driven monitoring across all service counters improves billing compliance behaviour among staff — reducing the frequency of incidents over time as the deterrent effect takes hold at scale.
The central dashboard enables the loss prevention team to prioritise outlets showing the highest discrepancy rates for targeted intervention — focusing resource where revenue risk is greatest rather than applying uniform audit pressure across all 500+ locations.
The system runs entirely on existing CCTV camera infrastructure via lightweight edge devices — no camera replacement, no network overhaul, and no capital expenditure on new surveillance hardware across the outlet network.
Tell us about your challenge and we'll set up a focused 30-minute session.