JD.com
AI-generated live commerce host
In 2024, JD.com's AI avatar of its founder generated about 50 million yuan of sales and more than 20 million views in one hour, while the JoyStreamer digital human cuts a live broadcast's cost by 90% compared with a human host.
Objective
Run live shopping rooms 24/7 without deploying human hosts, and open live commerce to merchants who cannot afford a studio and a team.
The deployment
JoyStreamer is JD.com's digital human model that hosts live shopping rooms in place of a human host. The avatar presents products, answers live chat questions, and handles the sale end to end. The most visible demonstration dates to April 2024: an avatar of founder Richard Liu (Liu Qiangdong), named the Procurement and Sales Digital Human, held a livestream that exceeded 20 million views in the first hour and generated around 50 million yuan in sales. JD then made the tool free for merchants, which pushed adoption up. At the end of 2025, the group highlights more than 50 AI tools for merchants and a shift of JoyStreamer from a broadcast assistant toward a live-growth engine that plans, presents, hosts, and analyzes the session.
Results Proof C
Established press naming JD.com (China Daily HK, Global Times) and the brand's official blog, consistent sources on the workings and the figures. The conversion and cost values are figures communicated by JD through the press, not audited results, which keeps it at C. China Daily and Global Times are Chinese state media, to be read with that bias.
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.
The stack in detail
- plateforme JoyStreamer JD.com's in-house digital human model: avatar, voice and video synthesis, hosting of the live commerce session
- llm JoyAI LLM JD's in-house LLM powering the sales script and the live chat answers
- infra Moteur de live commerce JD livestream rooms of the JD app, connected to merchants' catalog, prices, and promotions
- integrateur JD Explore Academy JD.com's internal R&D team that developed and maintains the digital human model
How it runs, concretely
For ops teams-
1Session preparation merchant
The merchant loads the products, prices, and sales script; the avatar and voice are configured.
-
2Broadcast AI
JoyStreamer presents the products live, runs the script, and highlights the promotions.
-
3Interaction AI
The avatar reads the chat and answers buyers' frequently asked questions during the live.
-
4Post-session analysis data team
The system reports views, conversion, and sales; the team adjusts the catalog, timing, and pitch.
Live chat engagement and the conversion rate during the session. If the catalog and prices are not up to date, the avatar presents a wrong offer and the room loses its credibility.
How your customers perceive this type of use
Sourced studiesUn ecart net separe les annonceurs des consommateurs : 77% des annonceurs voient l'IA positivement contre 38% des consommateurs (Yahoo/Publicis, 2024). Les mesures implicites confirment le rejet declare : en EEG, les pubs generees par IA produisent une activation memorielle plus faible que les pubs traditionnelles et sont decrites comme agacantes, ennuyeuses et confuses (NIQ, 2024). La disclosure a un effet ambivalent : elle augmente fortement la confiance quand elle est remarquee (Yahoo/Publicis), mais 27% des jeunes consommateurs disent faire moins confiance a une entreprise dont la pub est creee par IA (IAB, 2024).
Acceptance conditions
- Une disclosure visible : quand la mention IA est remarquee, la confiance globale envers l'entreprise augmente de 96% (Yahoo/Publicis 2024)
- Une qualite visuelle suffisante : les visuels IA de basse qualite augmentent l'effort cognitif et distraient du message (NIQ 2024)
Red lines
- Le contenu IA non declare puis identifie : 72% des consommateurs disent que l'IA rend l'authenticite difficile a etablir (Yahoo/Publicis 2024) et les marques utilisant des pubs IA sont plus souvent jugees inauthentiques ou non ethiques par les consommateurs que par les dirigeants (IAB 2024)
- Les mannequins et personnes generes par IA : 46% des consommateurs n'en veulent pas dans la publicite, l'inquietude premiere etant les standards de beaute irrealistes (Attest 2025)
Sources: Yahoo / Publicis Media (terrain Ebco) 2024 · IAB (avec Attest) 2024 · NIQ (NielsenIQ) 2024 · Attest 2025
How to replicate
Inference, not sourcedData prerequisites
- Structured product catalog with up-to-date prices and stock
- Sales scripts by category
- Voice and video corpus for the avatar (or off-the-shelf avatar)
Org prerequisites
- Compliance rule on disclosing the synthetic nature of the avatar
- Consent if using a real person's image
- Handoff model to a human on sensitive cases
Possible stack
- Digital human generator (avatar + voice synthesis)
- LLM for the script and the live chat
- Integration with the catalog and the live commerce engine
The plan, step by step
- Step 1Choose a high-volume category, clean up the catalog and prices, script the sessionsDeliverable: Reliable catalog + sales scripts per product
- Step 2Configure the digital human solution (avatar, voice) and lock the legal framework (image consent, disclosure of the synthetic nature)Deliverable: Avatar validated technically and legally
- Step 3Connect the LLM to the script and the product FAQ, rehearse internallyDeliverable: Pilot session recorded and validated off traffic
- Step 4Launch the pilot during off-peak hours on real traffic, with human handoff on sensitive casesDeliverable: Conversion and chat engagement measured vs baseline
- Step 5Extend the hours then the categories, adjust scripts and promotions on session dataDeliverable: Rooms running near-continuously + cost/conversion dashboard
First step: Choose a high-volume category with a clean catalog, and script a pilot session before opening up traffic.
The same pattern, proven in other industries
Intuit
done-for-you genAI assistant integrated into the product to reduce friction and convert
L'Oreal (marques ModiFace)
AR virtual try-on (synthetic makeup rendering on the face)
Gatorade
self-service genAI product personalization on a D2C site
Sources
- S1 Virtual humans helping spur e-commerce Established press archive pending
- S2 Chinese e-commerce giants tap AI solutions to unlock business opportunities Established press archive pending
- S3 The Future of Industry Marketing with JD.com's AI-Powered Solutions Interested party 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.