Yum China
genAI conversational ordering agent in a brand app
In January 2026, Yum China opened SmartK, an AI ordering agent, to all users of the KFC super app; already used by 2 million members, mostly for breakfast and coffee, in a context where 94% of sales go through digital.
Key points
- Yum China rolled out SmartK, an AI ordering agent, to all users of the KFC super app in January 2026.
- The assistant has already been used by 2 million members, mostly for breakfast and coffee.
- Context at scale: 94% of sales digital, 10.4 billion dollars in digital sales, 590 million loyalty members.
- Case confirmed by Yum China's Q4 2025 earnings call (primary source), a customer-facing assistant in production.
Objective
Streamline order-taking in the KFC super app and push personalized suggestions at the scale of a loyalty base of several hundred million members, in a context where nearly all sales go through digital.
The deployment
SmartK is an AI-based ordering assistant integrated into Yum China's KFC super app. It helps the customer place an order in the app and offers personalized suggestions during the journey. Yum China opened it to all users of the KFC super app in January 2026. At the time of the fourth quarter 2025 earnings call, management indicates the feature has already been used by 2 million members, with stronger adoption on breakfast and coffee orders. The setup sits on an already very deep digital base in China: 94% of sales go through digital, digital sales reach 10.4 billion dollars in 2025, and the loyalty program counts 590 million members for 265 million active members. SmartK is distinct from Q-Smart, the AI assistant Yum China is testing on the operations side for restaurant managers (scheduling, inventory, quality). It is also distinct from Yum! Brands' AI marketing setup in the United States: Yum China is a legally separate, separately listed entity that operates KFC and Pizza Hut in China. The AI model underlying SmartK is not detailed publicly in the available sources.
Results Proof A
The rollout of SmartK to all users of the KFC super app and the figure of 2 million member users are stated by management on Yum China's Q4 2025 earnings call (verbatim transcript, primary source). The context figures (94% digital sales, 10.4 billion dollars, 590M members) are confirmed by the same quarter's results slides. The headline result is an adoption metric, not a quantified conversion lift.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
How it runs, concretely
For ops teams-
1Opening the super app customer
The member opens the KFC app and reaches the SmartK assistant in the ordering journey.
-
2Understanding the request AI
SmartK interprets the customer's request and suggests suitable items or add-ons, with stronger relevance on breakfast and coffee.
-
3Personalization from the profile AI
The assistant draws on the member's history and context to tune the suggestions to the basket in progress.
-
4Confirmation and order customer
The customer confirms the suggested basket, and the order goes into the app's usual digital flow.
The order history and the member's context (time of day, breakfast and coffee habits). Without loyalty data tied to the member, the assistant loses personalization and falls back on generic order-taking.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- A base of identified members with order history tied to each profile
- A structured product catalog, with offers and availability by time slot (breakfast, coffee, lunch)
- A session context usable in real time (time of day, pickup or delivery point)
Org prerequisites
- A brand app where a significant share of orders already go through digital
- A product and digital team that owns the in-app ordering journey
- A legal basis and GDPR- and AI Act-compliant information for personalization and conversational interaction in the EU
Possible stack
- A genAI conversational assistant integrated into the app's order flow
- A recommendation engine connected to the member's history and the catalog
- A contextual personalization layer by time slot and purchase habit
First step: Target a high-repetition, low-basket-complexity ordering moment, for example breakfast and coffee, to launch the assistant where personalized suggestion has the most traction.
The same pattern, proven in other industries
Amazon
genAI shopping assistant
Amorepacific
genAI beauty advice agent built on a skin diagnostic
Booking.com
genAI trip planning assistant
Sources
- S1 Yum China (YUMC) Q4 2025 Earnings Call Transcript Primary archive pending
- S2 Yum China Q4 2025 slides: sales acceleration and record expansion drive earnings beat Established press archive pending
- S3 KFC, Pizza Hut operator Yum China embraces AI to boost efficiency, profitability Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.