From AI Technology to AI Solution

You can’t swing a manufacturer-branded lanyard around a security trade show floor these days without hitting 12 booths touting their latest artificial intelligence (AI) product offerings. The vast majority of these products are cameras, servers or edge appliances talking about object detection, or false alarm reduction for remote monitoring applications.

There is no argument that we have come a very long way in a very brief time with the accuracy of video analytics based on AI technology, but is the AI technology itself really the top consideration here?

Flashing lights and playing prerecorded messages in response to analytic detections is a common assumption around utilizing AI products for perimeter protection. But once you’ve deployed a few systems, you quickly realize that using a remote monitoring partner (or your own security operations center) to view the incidents and respond in real time is the linchpin of an effective solution. And this is where you start to look not just at the capabilities of the AI itself, but also the capabilities and integrations of the entire product.

Ten, or even five, years ago, getting reliable and accurate detections often involved cumbersome calibration and setup procedures that needed to be performed for each video input. Now, any product worth consideration is mostly point-and-shoot in this regard, and the only manual configuration requirement is defining your perimeter or area of protection and then configuring the appropriate settings to route events to your operators so that they can view live video and do audio talk-downs in response.

When evaluating AI-based products, people tend to focus on the AI aspect and fail to also consider the crucial parts that relate to enabling remote monitoring or, if using the AI functionality for things like people counting and business intelligence, integrations into your preferred video management system platform. This, however, may actually be the most critical part of the system. It does no good to detect an intruder if we cannot reliably and easily put that event in front of a central station operator and give them the ability to do their job.

Too often, trade show demos and product shootouts will focus heavily on all of the neat things the AI software can do, but then fail to point out that the user will need to use a cumbersome mechanism to actually view the output of the AI process. AI can be a great enabler for security operations at all levels, but only if we can put its output into a format and interface that users can easily access and utilize. Next time you are evaluating an AI product, don’t forget the P – the process or API that takes it from a technology to a solution for your application.

The views and opinions expressed in guest posts and/or profiles are those of the authors or sources and do not necessarily reflect the official policy or position of the Security Industry Association.

This article originally appeared in All Things AI, a newsletter presented by the SIA AI Advisory Board.