PLATFORM
Canopy Platform Edge Devices AI Agents Privacy by design
INDUSTRIES
Industries overview Hotels & Hospitality Retail Smart Cities Construction Out-of-Home
PROOF
Case Studies Insights
TECHNOLOGY
Technology overview Edge Compute Data Sovereignty OAIC Compliance
COMPANY
About Us Contact
Sign in Get in touch
CASE STUDY

Property group migrates 20 sites to edge AI

A phased migration that cut cloud costs while improving response time.

Case Study · Property — 20-site edge migration

An Australian property group was running cloud-based computer vision across 20 sites. The cloud bill was growing, and the response times were not getting any faster.

Commercial building equipment room housing edge compute hardware

The problem

The property group had signed up with a US cloud CV vendor in 2022. By 2025, the per-site egress cost had doubled, the model latency on regional sites was 400–500ms, and the security team was increasingly uncomfortable with raw video crossing the Pacific. They wanted to move, but they did not want a forklift upgrade.

Our deployment

We ran a phased migration. Two sites first, on a parallel Canopy edge deployment, running alongside the cloud pipeline for four weeks. We compared counts, latency and cost. Then we migrated the other 18 sites, four at a time, on weekends. The cloud vendor was turned off site by site.

The result

60%Reduction in per-site compute cost
< 20msMedian inference latency
0Raw video offshore

The migration took three months, with zero downtime. The security team was able to retire the cross-border data flow. The property group now runs all 20 sites on Canopy edge units, with a single Sydney-hosted control plane.

Canopy Console — Fleet 20 sites · aggregated · Sydney control plane Sites on Canopy edge 0 5 10 15 20 W0 W3 W6 W9 W12 2 sites · 4-week parallel run 20 Median inference latency Cloud, regional sites 400–500 ms Canopy edge < 20 ms 20 of 20 sites · 0 raw video offshore
Aggregated fleet view of all 20 sites after the phased migration.

Implementation details

The migration ran in parallel. We installed Canopy CU-2 units at two pilot sites alongside the existing cloud pipeline, and ran both for four weeks. We compared counts, latency and cost. The edge pipeline matched the cloud pipeline on counts within 3%, cut inference latency from 420ms to under 20ms, and reduced per-site egress cost by 60%. The security team retired the cross-border data flow for those two sites, and signed off on the pattern.

The remaining 18 sites were migrated four at a time, on weekends. We worked with the property group's IT team to schedule the cutover. The cloud vendor was turned off site by site. The property group's BMS and CAFM systems continued to receive events from the Canopy control plane, on the same JSON schema the cloud vendor had provided — no integration work was needed on their side.

Ongoing support

The migration is complete. All 20 sites run on Canopy edge units, with a single Sydney-hosted control plane. We visit each site once a year for a cleaning and inspection. Firmware updates are pushed automatically. The property group has added two new sites in the last year, on the same pattern.