top of page

KFC Dumi-

An AI robot which understands different accents to take orders.

To accommodate young consumers’ behavior so that they can fully manage and enjoy their dining experience

Challenge

While designing Dumi, we realized a great challenge facing us, which was to develop Dumi to understand different accents.

 

In China, there are 56 ethnic groups speaking more than 80 dialects, which may sound completely different from Mandarin.

 

Therefore, we needed to design a robot that can understand the major dialects and accents. We also wanted Dumi to be able to make some small talk with consumers to make the experiences even more fun.

The Solution

KFC Original+ Store, Shanghai

We used the AI technologies such as natural language processing, machine translation, multi-mode computing, Internet data mining and plenty of conversational tests, so that Dumi can accurately recognize the accents from the three largest dialect-systems: the Sichuan dialect, the Northern dialects and Cantonese.

 

Meanwhile, using facial recognition, Dumi can tell the age, gender or even mood a person. With this information and recognizing the region of the customers’ accent, which indicates the preferable tastes, Dumi is able to recommend a suitable meal.

KFC launched this Original+ Store when they celebrated its 5000+ store opening. This brand new digital dining experience became the highlight of the grand opening of KFC’s 5,000th restaurant. It gained over 600 new articles from various media outlets and created tremendous discussions on social media. It also attracted many consumers to the restaurant to explore fresher and more interesting digital dining experiences.

 

We believe in the long run, Dumi will contribute greatly to KFC’s operational efficiency. It will shorten consumers’ waiting times to enhance the consumers’ satisfaction as well as reduce the number of service staff, which in turn increases customer loyalty and profitability.

Results

Facial Recognition

KFC Dumi- AI Service Robot - Case Video

bottom of page