Evaluation of the current business processes with human assessors would require a lot of staff with the same qualification using same methodology. Additionally, objectivity of the collected data and conclusions would be questionable, since the staff behavior and performance would change under observation (people normally work harder when they feel their performance is being evaluated). It would also require each assessor to observe and register actions of multiple staff simultaneously for many hours without missing details. This would involve "human factor" and subjectivity at each stage, reducing the resulting data reliability.

VisionSystems.tech designed and developed a complex software solution that analyzes video data captured by existing customer's video surveillance system to identify and evaluate regular restaurant operations. The software automatically identifies the type of operation each employee is performing at every moment of time, the number of guests in a line for making order and amount of time taken to clean each table after guests leave. Collected stats was displayed in a clear and consistent way at the central self-service portal where authorized users can generate pre-built or self-created statistical reports and visual layouts.

Customer's management obtained a powerful instrument for reliable evaluation of production process in multiple restaurants. After collecting operation performance data for the first two weeks new instrument allowed business to identify specific deviations in particular processes that influenced overall facility performance. In some cases, management could take immediate business-decisions. Other, required additional audit but it took much less time and resource as particular targeted process was identified in advance. In all cases performance indicators demonstrated immediate boost.

The core of AI system is a software that uses machine-learning algorithms. It can run both on remote server and on specialized computers (Video Processing Units) located in each facility allowing to create either centralized, decentralized, combined or cloud architecture.

Software uses data captured by existing video surveillance system to analyze video streams. At the initial set-up stage system administrator uses admin panel to define zones for each operation (chopping, cooking, preparation, washing, guest table cleaning, cashier desk, etc) on camera-captured images. Staff uniform is marked with unique graphical markers that help computer vision clearly identify each employee. The Machine Learning model created by VisionSystems' Data Scientists identifies operation performed by each employee at each moment. This information is stored in database to be used manyfold for analytical reports providing insights. Authorized personnel can view pre-designed graphic performance visualizations as well as create their own reports with built-in report designer.

This solution can evolve in multiple directions. New features can be added to increase restaurant staff engagement. E.g. you can gamify the staff operations by providing them access to team or personal performance data and establish internal performance excellence award system. Analytics can provide way more insights if it is combined with restaurant business management software. It for example could help to see non-obvious dependencies between sales rate of a particular meal in particular restaurant with it's cooking process deviation in it.

Adding new Computer Vision features will let QSR analyze drive-thru performance, sanitary standards (wearing gloves, and facemasks) and many other aspects increasing customer satisfaction, mitigating risks and preventing loss.