Huayuesc launches AI Sourcing — LLM matching beta

Huayuesc has officially released its AI Sourcing feature — using an LLM (large language model) to understand complex Vietnamese RFQs and automatically match them with the most suitable suppliers in a database of 4,200+ partner factories. A free beta test is open to the first 100 buyers who sign up from April 5, 2026.

The old problem: keyword search does not understand context

Previously, a Vietnamese buyer would send an RFQ with a complex description such as 'Need to source 500 ergonomic mesh chairs with aluminum bases and headrests, target retail 4-5 million/pc, OEM logo printed on the side' — the old keyword search only matched 'mesh chair' and 'ergonomic', ignoring the price, OEM and target-retail requirements. The result: the buyer received 30+ irrelevant quotes and wasted time filtering them manually.

The solution: AI Sourcing with LLM + RAG

CSR's AI Sourcing uses a RAG (Retrieval-Augmented Generation) architecture: (1) A Vietnamese-tuned LLM (custom fine-tuned on 50,000+ historical CSR RFQs) understands context, target retail, OEM requirements and implicit priorities. (2) A vector database of 4,200 suppliers with embeddings of their capabilities, main products, price range and MOQ flexibility. (3) A re-ranking layer that uses business signals (rating, audit score, response time, dispute history) to select the top 5-7 best-matching suppliers.

Beta results (40 buyers tested in March 2026)

Especially impressive: AI Sourcing matched many complex RFQs that keyword search missed. For example, a buyer needed 'a 4-7 seat Nordic modular sofa in gray/beige for an 80m² apartment, target retail 25-40 million' — the AI understood the Nordic style (straight legs, bouclé fabric), the dimensions suited to a small apartment, the mid-to-high price, and correctly matched KUKA Hangzhou + ZuoYou Shenzhen.

How to use it

• ▸Go to /buying-request and click 'Use AI Sourcing (BETA)' in the top right corner.
• ▸Describe your needs in natural Vietnamese — the more detail the better (target retail, customer demographic, OEM requirements).
• ▸The AI returns the top 5-7 suppliers in 60 seconds + match reasons + a price/MOQ/delivery-time comparison.
• ▸Click 'Send RFQ to all matched' — the RFQ is automatically forwarded to the suppliers, with responses within 24h.

“AI Sourcing does not replace people — it lets the CSR Sourcing team focus on hard cases (custom OEM, large orders) instead of manually screening every RFQ. Efficiency is up 4x.”

Roadmap

Q3/2026: AI Pricing — predicting DDP prices with accuracy above 92% based on historical prices + real-time ocean freight + exchange rates. Q4/2026: AI Negotiation Assistant — suggesting negotiation replies to Chinese suppliers in Chinese. 2027: Multi-modal AI — buyers upload a photo of the product they want and the AI finds suppliers + compares visual similarity against the 4,200-supplier catalog.