For decades, banks have relied on traditional CCTV systems that primarily act as passive recorders. While the system was useful for post-incident investigations, it was limited in its ability to provide real-time visibility into what was happening inside the branch and around the ATM.
AI is now turning these static video systems into proactive intelligence tools that enhance security, improve operations, and provide real-time insights. This change is redefining how financial institutions detect risk, streamline branch workflows, and protect customer trust.
Global Banking & Finance Review recently spoke with Ofir Mulla, co-founder and CTO of AI video security provider Lumana, to learn more about what banks and financial institutions need to know about how AI is transforming video surveillance.
How have banks and financial institutions traditionally used video surveillance in their branches and physical operations?
Banks have historically relied on CCTV as a deterrent and as a record for post-incident investigations. Cameras were installed at counters, safes, ATMs, entrances, and parking lots, but the system was largely passive and footage was only reviewed after something went wrong. Security teams often had to manually sift through hours of video, and integration with alarms and access controls was limited or inconsistent. As a result, surveillance was useful for investigations but had little real-time awareness or operational value.
What new capabilities does AI bring to traditional CCTV and security camera systems in branch environments?
AI has transformed traditional cameras from mere recorders to real-time intelligence tools. Branch teams can detect behaviors that may indicate risk, such as loitering near ATMs, unusual movement patterns, or extended periods of inactivity at workstations. Beyond security, AI can provide operational insights by tracking queue lengths and other operational bottlenecks within branches. Managers can use this data to optimize layouts, staffing levels, and other efficiencies across the site.
AI solutions like Lumana can also provide early signs of risk that human observers often miss. For example, Lumana can detect ATM tampering, identify potentially escalating confrontations, and flag after-hours unauthorized access or policy violations in sensitive areas. It can also detect safety hazards such as smoke, slip-and-fall accidents, and sudden congestion. Operationally, it also highlights long waiting times, periods of understaffing, and inefficient branch layouts, allowing banks to act before problems impact customers or escalate into losses.
What privacy, compliance, and data governance obligations should banks consider when AI models are trained on sensitive footage involving customers and employees?
Video footage in a banking environment is regulated personal data, so governance must be built in from the beginning. Banks must limit use to defined purposes, implement strict access controls, and ensure a clear audit trail of all workflows. Technically, banks should limit exposure by processing video at the edge if possible and maintaining clear documentation of model training sources, updates, and performance checks. Compliance teams must ensure the legal basis for data collection and provide transparent privacy notices to customers and employees to comply with regulatory obligations and maintain trust.
How should banks assess whether their existing camera infrastructure is capable of supporting AI analytics or whether it needs to be upgraded?
Most modern camera networks can support AI, especially in edge computing. Banks must evaluate image quality, camera angle, and frame rate to ensure footage is suitable for accurate analysis. It is also important to evaluate network capacity and bandwidth constraints.
The best approach is to start with a small pilot at a branch or ATM cluster to validate performance, identify gaps, and make adjustments before expanding. This allows financial institutions to achieve immediate success without large upfront investments.
Where does Lumana fit into this ecosystem of AI-driven video surveillance?Lumana is a hybrid cloud AI video surveillance system designed to work with the cameras banks already own. Instead of requiring a complete replacement, existing cameras can be transformed into intelligent devices that offer centralized management, real-time alerts for specific activities and behaviors, rapid investigations, and AI dashboard capabilities that turn video data into actionable insights.
AI is processed at the edge to provide the low-latency recognition banks need, while enhancing privacy and minimizing data movement. For financial institutions balancing security, compliance, and operational efficiency, Lumana provides a scalable path to advanced AI capabilities without disrupting existing systems, leveraging the infrastructure they already have in place while increasing visibility, reducing response times, and meeting regulatory expectations.
Beyond security, what measurable business or operational benefits can banks achieve by modernizing video surveillance with AI?
AI-powered video security systems dramatically reduce investigation time from hours to minutes by allowing staff to search footage using natural language or specific attributes. Banks can detect fraudulent ATM and teller operations early, reducing losses and business losses due to fraud. Additional examples include AI-derived insights about wait times and customer flow to improve service quality and staffing allocation. Enhanced incident capture also reduces liability risk and supports more efficient insurance claims. Agencies that implement AI consistently report faster response times and measurable improvements in fraud detection and branch efficiency.
What emerging trends in AI video surveillance should banks prepare for in the next 2-3 years?
AI models increasingly adapt to each branch’s unique environment, reducing false positives and improving accuracy over time. Video analytics is integrated with transaction data, access control logs, and fraud monitoring systems to uncover more complex patterns of criminal activity. Privacy regulations will continue to evolve, with anonymization, on-device processing, and federated learning becoming standard expectations. Operational decisions such as staffing, layout adjustments, and customer flow design will become increasingly informed by continuous insights gained from video.
Any final thoughts or recommendations for banks looking to upgrade their video surveillance systems?
Start with a clear goal, not a list of technologies. Identify one or two high-value use cases, such as tamper detection or queue monitoring, and start a small pilot using your existing cameras and edge hardware. Build privacy and governance into your systems from the beginning so you can scale responsibly. We also ensure that video platforms are integrated into fraud, operational, and security workflows rather than operating in isolation.
Agencies that measure results (faster investigations, fewer incidents, shorter queues) will build the strongest business cases and scale most effectively.
Ofir Mulla, Co-founder and CTO, Lumana
