AI customer service: 4 minutes to 12 seconds
100K+ SKUs and member rules behind a RAG brain — stores and members answered in seconds, 24/7.
Smart customer service
Chain retail · 联华零售 · 2025
Lianhua Retail operates thousands of stores across hundreds of cities. Store staff and members flood the hotline with questions about products, promotions and return policies, served by a large human agent team with an average multi-minute handle time. Nights and holidays queue up painfully.
Internal knowledge is scattered across ERP, CRM, WeCom, Lark and an internal wiki, with rules changing every month. FAQ hit rate is low and transfer-to-human rate is high, and member satisfaction keeps falling. We needed to ship ask-correctly-and-escalate in one pass — without ripping out the existing systems.
How Ouryun ships Smart customer service for 联华零售
We put AI Gateway at the front door, a model platform underneath, and unify all of Lianhua's business systems, member rules and SKU knowledge. RAG powers retrieval-augmented answers, governance covers auditing and cost, and custom apps align the agent workspace, WeCom, Lark and IVR experiences.
AI Gateway as the single front door
ERP, CRM, member platform and internal wiki plug in via API / MCP, routed by tenant, scenario and model version. Zero disruption to legacy systems.
Dual-track model platform
A primary model plus a fallback, with a domestic LLM as primary and a fine-tuned support model as safety net. CPU / GPU scale elastically.
RAG knowledge base
120K SKU specs, 20K member rules, return SOPs, promo rules and store playbooks are chunked, embedded and indexed, with monthly incremental sync.
Full-chain governance
Conversation auditing, sensitive-word blocking, SLA dashboards, token usage and cost accounting — all in one place.
Aligned experience across four surfaces
Web workspace, WeCom, Lark and IVR all share the same UX, with one-click transfer to a human agent and channel-switch continuity.
From connect to ship, in four steps
Connect
In 2 weeks we adapt ERP, CRM, member platform and the wiki through API / MCP, and inventory every knowledge asset ready for vectorization.
Govern
In 3 weeks we deploy the model platform, configure primary / fallback models, build RAG indexes, and ship the audit dashboard with sensitive-word policies and tenant permissions.
Deploy
In 4 weeks we deliver the agent workspace, WeCom / Lark plugins and IVR — gray-released to 50 stores first and iterated twice.
Scale
In 8 weeks we roll out nationwide, retrain RAG monthly on real conversations, and significantly reduce the transfer-to-human rate.
120K SKUs and 20K member rules on a RAG knowledge base — 20x faster for stores and members.Full case study delivered and reviewed by the Ouryun team.
Numbers that prove value
Bring this capability into your business
More Ouryun case studies
Manufacturing AI quality inspection
An AOI + AI inspection appliance for PCB production that performs AI-based second-pass review on AOI NG images and reduces manual review load.
Domestic AI inference resource pool
A unified domestic AI inference resource pool for a large manufacturing group, supporting multiple large models and group-level AI platform operations.