Chick-fil-A is now using AI to monitor social media feedback from customers in order to detect and prevent foodborne illness, QSR reports.
At last week’s ReWork Deep Learning Summit in Boston, Chick-fil-A’s senior principle IT leader of food safety and product quality, Davis Addy, explained how this tech works.
For most QSRs, gathering feedback from social media is key for getting insights into what’s working and what isn’t with the business. That includes spotting any mentions of food safety issues or potential foodborne illnesses like norovirus, which on average causes 19–21 million cases of acute gastroenteritis annually in the U.S. Restaurants and catered events are one of the most common settings for norovirus, according to the CDC.
But to find these mentions, Chick-fil-A (or any QSR, for that matter) has to sift through hundreds if not thousands of customer reviews every day, many of them written with poor (or no) grammar and what Addy called “mixed sentiments and off-topic musings.” In other words, references to legitimate foodborne illness issues are often hidden amid hyperbole, misspellings, and other facets of modern communication.
Hence, more AI. Chick-fil-A has developed a custom AI platform that’s hosted on Amazon’s AWS Comprehend, a natural language processing service that will check data from Chick-fil-A’s social media channels every 10 minutes and filters for over 500 related keywords like “nausea” and “food poisoning.” The system can then uses machine learning to check the sentiment of the feedback and determine its legitimacy. In the case of legitimate issues, the system notifies store managers.
At the end of 2018, researchers at Google and Harvard unveiled a somewhat similar model called FINDER that uses machine learning to scan anonymous and aggregated data on searches that indicate food poisoning (e.g., “stomach bug”) from users who save location data onto their phones. It uses that information to pinpoint restaurants which are potential sources of foodborne illness, then notifies the city’s health department. FINDER is still in pilot mode; tests have so far taken place in Chicago and Las Vegas.
Chick-fil-A says its system currently operates on 78 percent accuracy and that the company is working with Amazon to improve those numbers. Since machine learning is a type of AI that learns over time rather than needing to be programmed, the hope is that it can gradually improve its ability to both detect keywords and also determine if they come from legitimate complaints about food illness.
Once it’s done that, AI’s job ends. Restaurant operators and managers have to take that information and decide how to act on it, at least for now. With humans and robots/AI starting to work side by side in an increasing number of restaurant settings, AI could not only track down foodborne illnesses, but also help to prevent them in the first place.