Retail shrink — the gap between what should be in a store and what actually is — costs the Indian retail industry an estimated ₹45,000–60,000 crore annually. Employee theft, external shoplifting and administrative errors all contribute. AI is addressing all three simultaneously in ways that were impossible with traditional security infrastructure.

The Anatomy of Retail Theft

Organised retail crime is methodical. Reconnaissance visits map camera positions and staff patterns. Distraction techniques draw security attention. Concealment happens in dressing rooms, bags and clothing. Exit strategies exploit busy periods or emergency exits. AI counters each of these steps specifically.

Reconnaissance Detection

AI-based customer behaviour analytics identifies individuals who enter a store multiple times without purchasing — a strong precursor to organised theft. A person who visits three times in a week, spends 80% of their time near high-value merchandise and is consistently observed checking for staff positions matches a documented reconnaissance pattern.

Combined with facial recognition, a watchlist of known offenders can be maintained and cross-matched against in-store visitors in real time — with a push notification to the store security manager when a match is detected.

Real-Time Shelf Monitoring

AI object detection can monitor specific shelf zones and detect when products are moved from a shelf to a bag or pocket without being taken to a checkout. This is the most technically demanding application — requiring high-resolution cameras, precise zone calibration and reliable product classification — but increasingly viable as camera resolution and AI accuracy improve.

Self-Checkout Fraud Detection

Self-checkout fraud — deliberately scanning a lower-priced item in place of an expensive one, or not scanning items at all — is estimated to account for 20–40% of self-checkout shrink. AI cameras monitor the checkout belt in real time, comparing what is placed on the belt versus what is scanned, and flagging discrepancies for review without requiring every transaction to be manually checked.

EAS + AI Integration

Traditional EAS (Electronic Article Surveillance) systems alarm when a tagged item passes through the pedestal. AI integration adds context: when an EAS alarm triggers, a PTZ camera automatically zooms in on the exit point and video is auto-saved and tagged with the alarm event. Floor security is notified with a video clip rather than just an alarm tone — enabling faster, better-informed response and reducing the confrontations that arise from reacting to alarm noise without visual confirmation.

Results from Indian Deployments

Across Indian retail deployments combining AI video analytics with EAS and facial recognition, documented outcomes include: shrink reduction of 35–52% in year one, 70% reduction in organised retail crime incidents, and a measurable reduction in staff theft following awareness of AI monitoring (the deterrent effect).