FACILITIES AND LOGISTICS

Vision agents for workplace safety and security

Security teams cannot watch every camera. Wexa built vision agents that detect intrusion, loitering, falls and known persons on existing CCTV, and send each event with evidence to a person for review.

Download this case as PDF Book a demo

detection on a CPU, no GPU needed
p95 per frame for the full 1080p pipeline
incident types plus cross-camera re-identification
of events reviewed by a person

CLIENT: Warehouse and facilities operator
REGION: India
WHEN: 2026
ENGAGEMENT: Proof of concept
WEXA COMPONENTS: Vision models at the edge · incident engine · Wexa workbench · human review

01Case 29

The challenge

Too many cameras for operators to watch
Incidents noticed after the fact
Privacy law on face data

What Wexa built

Detection, tracking, pose, zones, faces and re-identification on CCTV, RTSP or video files
Incident engine writing events with evidence images and model versions, pending review
Edge-deployable: CPU container today; GPU, Jetson and mobile targets documented
Privacy by design: consented gallery only, unknown faces never stored, retention rules

02Case 29

How it runs on Wexa

FigureHow this use case runs on Wexa: client systems connected, ontology discovered, context graph in the Wexa memory layer, agents and process flows, and governance. Highlighted components are central to this use case.

03Case 29

Vision pipeline

04Case 29

Results

Measure	Before / without Wexa	With Wexa
Camera-monitoring operators across three shifts	24 operators	8 operators
Annual cost of the process	$216K	$72K
Losses from theft and safety incidents avoided per year	–	$150K
Time to implement the AI solution	6 months	4 weeks
AI build cost (one-time)	$480K	$30K
AI operating cost per year (models, infrastructure, upkeep, platform)	$312K	$166K
Total AI cost in year 1	$792K	$195K