6 Questions to Ask Before Deploying AI-Powered Video Analytics 

During an AI-powered analytics webinar, I conducted a survey, posing a straightforward question about attendees’ future technology adoption plans.   

The results indicated that 50 percent of nearly 100 participants are either currently in a pilot phase or have intentions to initiate a pilot for AI deployment in their video infrastructure within the next 12 months. This figure harmonizes closely with my market observations, highlighting a substantial interest and eagerness to embrace AI technology at this juncture.  

I’ve witnessed firsthand the interest in AI adoption, and clearly that interest is only growing.  Over the course of my career, I’ve orchestrated effective deployments of AI-powered video analytics for numerous F10 corporations, major critical infrastructure providers, and federal government agencies. While each business has its own distinct technical requirements and needs, I have noticed a consistent pattern in the decision-making criteria they follow.   

Below, I’ll explore the six most common and pressing questions that businesses ask before incorporating AI into their video security.  

1. What goals do we want to achieve with AI-powered video analytics?  

It’s crucial to have a clear vision and a well-defined business driver when starting on any initiative. Otherwise, you might find yourself with so many pilots that it feels like you’re running an airline!   

In my previous work with clients, I’ve employed various goals and associated KPIs to measure success. Below, I’ve covered a few of those goals that I’ve set with clients in the past and how we measured them.  

Goal #1: Better Security 

According to the 35th annual Hayes Annual Retail Theft Survey, the average loss per retail theft incident was over $800 in 2022. One of the best ways to measure this goal is by utilizing AI and seeing whether a reduction was made against the average loss during incidents of theft. 

Investigation time is another metric that can be used to determine the success of AI. We have seen customers reduce their investigative time by 50% with AI, meaning they were able to handle twice the number of investigations with the same staff. 

Goal #2: Enhanced Safety 

According to the 2022 National Council on Compensation Insurance’s (NCCI) Workers Compensation Statistical Plan database, injuries resulting in fracture, crush, or dislocation cost employers $60,934 per incident. If you’ve got spaces where heavy equipment and people cross paths, there’s clear value in using analytics to detect when a person is in a location they shouldn’t be. 

Goal #3: Improved Operational Efficiency 

Using analytics to measure wait times for customers in a drive through lane or waiting to check out at a register provides immediate value to businesses in sectors where employee-client interaction is so integral. 

Additionally, understanding not only what your customers are buying but how they’re interacting with your store as they move through it is incredibly valuable data, all of which AI can provide. For this goal, it’s important to focus on real time and historic trends for occupancy, people counting, and heat mapping. 

Goal #4: Superior Customer Experience 

All stages of the customer journey can be analyzed with AI-powered video, including dwell times. This could be journey times, customer wait time, queue lengths, food delivery time, employee present, or walk-offs for in person ordering. 

2. What legal and ethical considerations should we consider before utilizing AI technology? 

There are four distinct areas to focus on when evaluating the legal and ethical considerations of AI deployments. 

Data Privacy and Local Regulations 

Organizations need to ensure that the collection, storage, and analysis of video data comply with data privacy laws, such as the General Data Protection Regulation (GDPR) in the European Union or the California Consumer Privacy Act (CCPA) in the United States.  

This may include obtaining consent from individuals before collecting their data and ensuring that the data is stored and processed appropriately. 

Transparency and Accountability 

Organizations need to be transparent about their use of video data and ensure that individuals are aware of how their data is being collected and used. They also need to be accountable for the use of the data and ensure that it is being used in an ethical and responsible manner. I recommend a regular process to review use cases with stakeholders to ensure everyone is aligned on the actual deployment.  

Potential Discrimination and Bias 

Businesses must be vigilant about potential biases in data or algorithms used in computer vision technology. These biases can lead to discrimination against certain individuals or groups. Addressing these biases is paramount, and ensuring fair and non-discriminatory use of technology is a shared responsibility. To gauge this, it’s prudent for users to question vendors about their technology’s training, data sources, and performance on industry-standard bias tests.   

Boundaries Around Ethical Use of the Technology 

Businesses need to ensure that the computer vision technology is being used in an ethical manner and is not being used to infringe on the rights of individuals or groups. This may include avoiding the use of the technology for overly invasive purposes for which there is no clear business justification.  

3. What else will we need to support a video analytics deployment? 

I like to call this the four P’s for supporting analytics. All four need to come together to deliver a successful AI-powered video analytics deployment:  

Picture (for Increased Clarity) 

Analytics don’t need 4K everywhere, but if you are detecting faces or license plates, the resolution of objects being captured is important. In general, the more detailed your detection is, the more picture quality, resolution, and lighting matter.  

Placement (for Ideal Visibility) 

Cameras must be in the right place, meaning they shouldn’t be so far away or obscured that they can’t do their job. This might mean moving or adding a camera in order to properly deploy the analytic you need.  

Performance (for Real-Time Computing) 

AI-powered video analytics requires computing power to analyze video data in real-time. It’s best if this computing is embedded in the camera or your VMS appliance, but the cloud can be used as well. Either way, budgeting for needed hardware upgrades should be part of your deployment plan.  

Process (for Streamlined Operations) 

Detecting and collecting analytic data is only helpful if you do something with it. Make sure the VMS you choose is easy to use and delivers intelligence for you to act on. Also, that you have operational processes in place to ensure users are leveraging the data from your AI investment.  

4. How should we train our company to properly use AI? 

My experience has been that there are five key areas to focus on when thinking about how to train your staff to properly use AI-powered technologies.   

Technical Training in AI-Powered Video Analytics Systems 

A couple key members of your staff should be trained in the technical aspects of the AI-powered video analytics system, including how to set up, configure, and use the system. This may include training on the hardware and software components of the system, as well as on data management and analysis tools.  

Data Analysis and Interpretation 

Staff who will be using the software should be trained in how to interpret the results of the AI-powered video analytics system. This may include training on how to identify patterns and anomalies in the data, how to analyze data for specific business purposes, and how to interpret the results of the AI-powered video analytics system to make informed decisions.  

Ethical Considerations for Using AI Technology 

We discussed this earlier, but it’s a necessary aspect of AI deployment. Staff should be trained in the ethical considerations surrounding the use of AI-powered video analytics, including the importance of protecting individual privacy and ensuring that the system is used responsibly.   

Ongoing Technical Support and Training 

Staff should have access to technical support when they need it to ensure the system is used effectively while troubleshooting any issues that may arise. This support framework may encompass a responsive help desk or dedicated technical support team readily available to address inquiries and offer guidance.  

Regular Training Updates for Effective Use 

As the system evolves and new features are added, staff should receive regular updates and training to ensure that they are up to date on the latest capabilities of the technology and how to use it effectively.   

5. How will this help our company in the long term? 

With AI, you’re not buying a static product. AI evolves and changes over time, and deployments get better in a couple important and consistent ways.  

More Data to Improve Accuracy 

More data is, well, more better when it comes to improving the accuracy of AI-based systems. What this tangibly means is that as more customers choose to opt in to programs that share selected video data with their AI providers, the tech gets better in terms of accuracy.   

Improved Algorithms in Computer Vision Technology 

Developers are continually refining and improving the algorithms used in computer vision technology, which is leading to better accuracy and more advanced capabilities.  

Simultaneous Advancements in Hardware 

Not to be outdone by the software, the hardware technology is advancing apace as well. Growing interest has only fueled rapid advances in technology, and most of the things we could only imagine today will soon be done in a combination of camera and cloud infrastructure without the need for dedicated appliances.  

Growing Expertise and Knowledge 

As security teams become more familiar with computer vision technology and how it can be utilized in physical security applications, they can better optimize and tailor the system to their specific needs. This can include improving the accuracy of the system, reducing false positives, and developing more effective response protocols.  

Integration with Other Systems 

As computer vision technology is integrated with other physical security systems, such as access control or alarm systems, it can become more effective and efficient in identifying and responding to potential security threats.  

6. How does AI complement our business outside of video security? 

Taking advantage of security footage begins with understanding its most powerful asset: data.   

Rich and contextual data on what is happening in the physical spaces of your organization is, figuratively and literally, gold. But to supercharge this visual data, the metadata needs two things: 

1. Organizations need an underlying philosophy about how to treat and leverage their data and the insights it produces. Data-derived insights need to be like vapor: existing everywhere around every user and business process.   

2. The metadata needs a robust way to connect with other business systems, like the ERP, real estate/space planning tools, and building management systems. This is done via APIs. APIs and their usefulness have grown exponentially to easily connect different business systems. 

Tangibly, if the real estate team can benefit from insights about space utilization coming from existing video infrastructure, that metadata should make its way, via an API, into whatever real estate planning tool the real estate team is using. The insights should meet them at the point of need, not make them come to a separate application and workflow. This two-fold approach provides the foundation for an organization to take the raw video coming off cameras and turn it into more profitable spaces and more efficient processes.   

In the process of helping dozens of organizations start and expand their AI deployments to work on their video infrastructure, the six questions covered here have proven, time and again, to be a helpful starting point on the journey.  

Interested in seeing if AI-based video analytics is the right fit for your company? Book a demo with OpenEye to see how we can equip your business with powerful, advanced technology for a complete security solution.  

Watch the webinar as well to dive deeper into the world of AI-powered video analytics. 

About the Author

Brent Boekestein leads OpenEye’s Enterprise team. Prior to that, he co-founded Vintra, an AI-powered video analytics company that was used by multiple Fortune 100 companies and the US Government prior to its acquisition by OpenEye and Alarm.com in April of 2023. 

Brent has delivered over two dozen talks globally on the Internet of Things and artificial intelligence. He holds an MSc from the University of Manchester and is a co-author of two patents in the field of computer vision.