Everything that can be automated needs to be automated. This approach significantly reduces costs, speeds up processes, and lightens the support team and proctor workload.
Frequently, large customers conduct daily testing simultaneously with tens of thousands of users. At this rate, they do not have enough time and capacity even for a random video records review. Ideally, they want all monitoring processes to be automated. Specifically for such clients, we have enhanced our algorithms. Now, AI not only detects infringements but also tracks their duration. This increased the accuracy of catching cheaters, on average, by 30-35%, and in some cases even by 55%.
As a result, we were able to develop a fully automated online proctoring mode. AI independently monitors infringements and, based on this information, determines whether the exam was passed honestly or not. We still believe that post-review by human proctors is highly desirable. However, if customers' resources and time are limited, they can use the automatic mode.