Video used to be something you reviewed after the fact. Today, Artificial Intelligence (AI)-powered video analytics is turning cameras into real-time operational sensors, catching incidents while they’re still small instead of hours later. Following our recent webinar with Beecham Research and Digital Barriers, we sat down with Rupa Datta, vice president of Connected Services at Semtech, and Clive Sawkins, CEO of Digital Barriers, to hear about what it takes to deploy that reliably at scale.
Q: What Is Driving the Shift to AI-Powered Video Analytics?
Rupa Datta: Customers deploying cameras and routers were recording huge volumes of footage but not seeing a return on it — video was reviewed after an incident, not during it. AI now lets us put real-time response in place: detect an incident, act on it while it’s still small and compress the data so customers aren’t paying to transmit full uncompressed video around the clock.
Q: What Technology Made Real-Time Video Analytics Possible?
Clive Sawkins: A few things converged. Cameras became more intelligent and more powerful at the edge. Edge compute became dramatically more affordable and reliable. And video compression software — ours, specifically — matured to the point where we can move AI-generated video and metadata across constrained uplink networks affordably, without sacrificing the quality the back-end AI needs to process it.
Q: What Are Real-World Examples of AI Video Analytics in Action?
California’s Alert California program, run with CAL FIRE, monitors over 1,200 AI-enabled cameras statewide and flagged roughly 3,600 wildfire incidents in a recent 12-month period — more than half detected before anyone called 911. In France, the FireGuard platform in Finistère is credited with helping prevent more than €10 million in damages in its first four months. The same pattern shows up in traffic management (London’s AI-driven signal rerouting) and emergency response (Georgia DOT cut stranded motorist location time from 23 minutes to three minutes).
Q: What are the Biggest Barriers to Scaling AI Video Analytics Deployments?
Clive Sawkins: The first question is always: can you deliver high-quality, reliable video streaming over a variable network (cellular, wireless, radio, and satellite)? The second question is the speed of deployment. With a SIM card and the right hardware, there’s no fiber to dig in.
Rupa Datta: On the customer side, it’s quality and cost predictability. A pilot with one or two cameras always looks fine. At scale, over time, customers hit network congestion and video degradation — and unpredictable overage costs on standard Internet of Things (IoT) data plans.
Q: How Does the Semtech and Digital Barriers Solution Work Together?
Together, the two companies deliver Semtech Video Compression — a native integration, not two products bolted together. Semtech contributes Smart Connectivity (multi-network SIM routing, a fully owned and geographically redundant core network and the AirVantage® platform for fleet management) and its AirLink® XR60 router. Digital Barriers contributes EdgeVis, the adaptive compression codec purpose-built for mission-critical video.
Rupa Datta: “The XR60 is not the only piece of hardware that Smart Connectivity is actually compatible with. So, if you have a router that’s in-field with the same sort of compute power that the XR60 has, then talk to us about how to get that software. Smart Connectivity can run on any type of hardware as long as it is compatible with the software, meaning the Digital Barriers software. So, you can get the same value proposition not only with the XR60, but other comparable routers.”
Clive Sawkins: And on performance, we’re seeing 50–90% bandwidth savings compared to traditional video transmission. The codec adapts in real time as network conditions change — it doesn’t buffer or drop; it adjusts.
Q: How Does Semtech Video Compression Perform Against Standard Video Streaming?
The webinar included a side-by-side test: an AXIS camera running standard H.264/Zipstream compression versus Semtech Video Compression running on an AirLink XR60, both feeding a Milestone Video Management System (VMS), with live Wireshark traffic monitoring. As the test vehicle accelerated and the scene became more complex, the standard stream spiked toward 4Mbps trying to keep up. The Semtech Video Compression stream stayed flat and kept pace with scene changes — the practical result of the codec adapting to available bandwidth in real time.
Q: Why Does Low-Latency Video Analytics Matter Operationally?
Rupa Datta: Clarity and low latency directly affect the decision you can make. A blurry face doesn’t help you identify anyone. A foreign object on a runway needs to be seen clearly enough, in real time, to act before a plane takes off.
Q: What Other Industries Use AI-Powered Video Analytics?
Beyond traffic and wildfire detection: oil and gas remote monitoring in low-connectivity areas, worker safety on construction and mining sites, manufacturing floor anomaly detection, national infrastructure protection, and even real-time monitoring aboard container ships at sea.
Q: How Is Video Analytics Connectivity Secured?
Rupa Datta: Smart Connectivity uses a proprietary applet on the SIM that automatically routes to the least congested available network — this runs globally today. EdgeVis is hosted across two geographically redundant Semtech-owned data centers, and Semtech doesn’t throttle video after compression. Smart Connectivity has maintained greater than four-nines uptime over the past year.
Q: What Does a Typical AI Video Compression Deployment Timeline Look Like?
Rupa Datta: A greenfield deployment can be live in six to eight weeks. If AirLink XR60 routers are already in the field, adding the Digital Barriers EdgeVis software and activating SIMs can take three to six weeks. Comparable third-party routers with sufficient compute power fall in a similar window.
Q: What Is the Migration Path for Customers Who Already Have AirLink XR60 Routers Deployed?
Rupa Datta: The EdgeVis container can be pushed to existing XR60s remotely, over the air — no on-site visit or hardware swap required.
Q: What Advice Do Experts Give for Deploying AI Video Analytics?
Clive Sawkins: Treat it like any other project — define your ROI target, understand your network and cost model and be clear on the driving objective (public safety, cost optimization or resilience).
Rupa Datta: Look beyond short-term total cost of ownership. The larger return comes from catching and resolving incidents faster, which compounds into operational efficiency over time.
The Bottom Line on AI-Powered Video Analytics
AI is turning video from a passive record into a real-time operational sensor. The barrier to scaling it has been the trade-off between video quality and cost predictability. Semtech Video Compression, with Digital Barriers EdgeVis natively integrated on the AirLink XR60 (and available on comparable third-party hardware), is designed to remove that trade-off, with up to 90% bandwidth savings and predictable connectivity.
This Q&A only covers the highlights. Watch the full webinar on demand to see the live side-by-side technology demonstration in action, plus the full audience Q&A with Rupa Datta and Clive Sawkins.
Watch the on-demand webinar → When Video Becomes a Sensor: Turning AI-Powered Video Into Real-Time Operational Intelligence
Semtech®, the Semtech logo, AirLink®, and AirVantage® are registered trademarks or service marks of Semtech Corporation or its subsidiaries. Other product or service names mentioned herein may be the trademarks of their respective owners.
