Lazada

Objective

Increase conversion and basket size on the marketplace by answering buyers' product questions at the moment of consideration, personalizing recommendations and review summaries, and lowering the cost of producing quality product listings for the million active sellers, across six Southeast Asian markets where appetite for shopping AI is already strong.

The deployment

In October 2024, Lazada announced a suite of genAI features for its six Southeast Asian markets (Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam), organized around two sides. On the buyer side, the AI Lazzie assistant answers product questions, compares features, summarizes customer reviews, and personalizes recommendations based on browsing behavior and purchase history; to this are added smart reviews, AI-curated offers and vouchers, and a SmartStack mechanism that automatically combines rewards and seller vouchers. On the seller side, a Product Listing Agent generates titles and descriptions, translates listings into the region's languages through the Marco MT tool, and suggests a competitive price based on real-time demand; a Refund Agent handles refund requests autonomously. The whole thing is built on Qwen, the large language model of the Alibaba Group. Lazada states that it connects about 160 million active users to more than one million active sellers each month. In the pilot phase, Lazzie reportedly increased orders by 42 percent and interactions by 50 percent; these figures are a pilot, not a generalized result. The Product Listing Agent optimized more than 13 million listings, with page views up as much as 180 percent in one week, and 35 percent of refund requests were handled autonomously with 99 percent accuracy (press, November 2025).

Results Proof B

Official Lazada press release (T1_primary, PR Newswire, Oct. 2024) announcing the genAI suite across six markets and quantifying the marketplace scale (160 million users, one million sellers), consistent with established press (The Manila Times, Nov. 2025) reporting the impact figures attributed to Lazada (13 million listings optimized, autonomous refunds, and the pilot at +42 percent). Public and concordant figures but none at the financial-results level; the +42 percent is explicitly a pilot.

The internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.

1. 1
   Buyer question client

   The customer asks Lazzie about a product, a comparison, or a review, in the app or on the site.

2. 2
   Conversational response and recommendation AI

   Lazzie interprets the request through Qwen, retrieves listings, reviews, and history, summarizes the reviews, and offers personalized recommendations and vouchers.

3. 3
   Seller-side listing generation AI

   The seller submits a product; the Product Listing Agent generates the title and description, translates it into the market's language through Marco MT, and suggests a competitive price.

4. 4
   Autonomous refund handling AI

   The Refund Agent processes part of the refund requests and issues return codes, with no human intervention on standard cases.

5. 5
   Supervision and steering data team

   The teams track conversion, interactions, seller page views, and the autonomous refund rate to extend the scope and adjust the models.