If you went to a QSR in the twentieth century, your order was taken on a notepad, and the page was torn out and posted for the kitchen crew to prepare. Tracking and moving the “ticket” through the kitchen was the innovation that allowed workers to build your meal like an assembly line.

As we entered the 21st century, information technology like KDS and POS systems made it easier and faster to coordinate between the front and back of house. Your order moved instantly from the POS to each relevant station, so the crew got the information they needed nearly instantly.

Today, the AI revolution is the biggest ongoing change in the management of the QSR. If concrete information already moves instantaneously, the next innovations are the ones that predict the future and provide AI-optimized guidance to the crew at each station.

If you go to a QSR today, there’s a strong chance an AI agent is involved in the preparation of your meal. There are a number of points in the production process where an AI agent might be aiding the crew or making decisions to ensure your food gets made as freshly, quickly, or accurately as possible.

Where are AI Agents in the Kitchen?

The goal of QSRs is to serve delicious food as quickly and accurately as possible. Of course, restaurants can serve delicious food by switching to premium ingredients, but there’s a cost tradeoff associated with it. GM’s can have their crew serve quickly, but they’ll sacrifice freshness or accuracy. If crew triple check the accuracy of each order, service times rise. Tradeoffs in the kitchen demand a solution like an AI agent for the restaurant to get as much as possible of the two seemingly opposed goals.

Speed of service is directly correlated with customer satisfaction in a QSR, but service times have risen by 20 seconds in 2019 alone. Customers want speed, but are not willing to sacrifice quality to get their food sooner. How can a GM deliver both? An AI agent, the QSR Brain, monitors both the upcoming demand and the current inventory levels to determine exactly how much to cook right now and guides the crew members to keep the optimal amount of inventory available.

AI Agents manage inventory and demand, requesting fresh food production when it’s going to be sold.

The AI Agent has advantages over people trying to make the same decisions. While a person can only focus their attention on one task at a time, the AI Agent can monitor a plurality of camera streams, while constantly calculating demand and production schedules for several stations. With computer vision detection on each of the streams, the computer can make detailed and informed decisions considering vastly more information than any manager or crew member could. The result of these decisions is efficiency, speed, and ultimately improved profits.

Order accuracy presents another tradeoff affecting customer satisfaction. The typical approach to improve order accuracy is to take extra time to check each order, or to simplify the menu so inaccuracies become less of an issue. Since customers like to customize and stores gain throughput when they sell food faster, neither of these solutions offers a beneficial effect for customers.

An Order Accuracy AI agent adds extra eyes on the production process, improving accuracy without the tradeoffs. Vision sensors check for each item as it’s added to the bag, comparing what they see against the data from POS to track down inaccuracies. These AI agents inform crew members when an item has been forgotten or mistakenly added to an order. The instant feedback improves accuracy without slowing down service.

Computer Vision AI Agents monitor the accuracy of orders as they’re constructed.

Next time you’re in a QSR or ordering in the drive thru, take a look inside the kitchen to see if there might be an AI agent helping the crew serve your food faster or better. You’ll be surprised at how many are involved in the process.

Al Indig is an engineer and project manager at PreciTaste with personal experience in AI project implementation for F500 clients. Read his other articles here: