Video Analytics: Innovation and Market Challenges

For years, video surveillance has been an exercise in observation. Someone watches, someone reviews footage, someone reacts. Artificial intelligence promises to change that, and in part it does. But the biggest obstacle we have faced in bringing these solutions to market is not technical. It is an expectation.

The market has come to believe that AI can do everything, and that video analytics is an off-the-shelf solution, already designed to fit any vertical. Many companies and brands have promoted these technologies that way, creating confusion and expectations that no single product can meet. Our experience points the other way: off-the-shelf solutions do not necessarily address clients’ real needs. That is why the most important, and most difficult, task has been understanding what each organization truly wants. A technical team that immerses itself in the client’s environment and defines the problem before choosing the technology is better positioned to address the client’s needs.

In practice, that means starting with questions rather than features: what decision is this analytic meant to support, who will act on it and what does that person need to see first? A solution built around those answers will look different from one organization to the next, even when the underlying technology is the same. It is the opposite of the off-the-shelf approach, and it is where, as we often say, what truly differentiates us is not just the code, but our empathy and commitment.

Once that first obstacle is overcome, a deeper one appears. The opportunity we see is not limited to Latin America; it is a global trend. There is a barrier in how people interact with video analytics: the market is not yet prepared to digest the large volumes of information these tools generate. The reason is habit. Organizations are used to treating cameras passively, as observers. When analytics arrive, watching is no longer enough: there is data to interpret, prioritize and connect to daily operations. More detections do not mean better decisions if no one can process them. Breaking that paradigm is, for us, the next step in video surveillance.

That principle shapes how we design the AZOR ecosystem. Instead of showing everything, the analyst panel receives only AI-filtered events for validation, reducing continuous screen monitoring and analyst fatigue. Unstructured video becomes metadata that enables forensic searches in seconds rather than hours of manual review. The web interface we recently launched for video surveillance as a service (VSaaS) and on-premises deployments features a multitenant architecture and global configuration. It follows the same idea: letting each organization interact with its analytics simply, from a single environment. This launch also includes code and performance improvements aimed at optimizing the use of processing resources.

  ​Video analytics does not replace the judgment of the people supervising; it extends it.   

The architecture follows the same logic. Processing video close to the source and sending only metadata, alerts and keyframes can cut bandwidth consumption by 80 to 90 percent. In this way, analytics reaches operations without forcing clients to replace what already works.

All of this also demands a learning curve from the market: understanding what each tool offers, recognizing its limits and linking its results to everyday work. Video analytics does not replace the judgment of the people supervising; it extends it. Organizations that grasp this will stop being spectators of their own cameras. Each camera will stop being a window and become a sensor.