01Overview
AI for commerce work
Commerce work keeps generating product, order, review and inquiry data, and the speed and accuracy of search, recommendation and support tie directly to revenue. Plateer AI Labs applies the commerce-focused sLLM Polar, the agent platform XGEN and the commerce platform X2BEE together.
Products and capabilities used
02Challenges
What commerce teams run into
- 01
Keyword search misses intent
Phrases like "a light jacket for the commute" rarely match product-name keywords, so the right products often don't show up.
- 02
Repeat inquiries tie up support staff
Inquiries with set answers — delivery, exchange, sizing — take a large share of support time, leaving less for complex cases.
- 03
Reporting starts with data gathering
Collecting sales and market data from several systems takes time away from analysis and decisions.
03Approach
How it is applied, step by step
- 1
A model trained on commerce data
Polar is an sLLM trained on product, review and inquiry data, and can run on in-house servers.
- 2
Semantic search and recommendation
Korean embeddings and RAG interpret query intent to find products, and recommendations reflect order and stock data.
- 3
Support and reporting agents
XGEN is used to build FAQ, review-summary and sales-report agents and connect them to existing systems.
- 4
Integration with the commerce platform
AI features are attached to the product and order flow of commerce platforms such as X2BEE, used directly in operating screens.
04Typical work
- 01Natural-language product search and top-N recommendations
- 02Review summaries and automated answers to repeat inquiries
- 03Periodic reports built from market and sales data
05Before you adopt
- Search and recommendation quality depends on the consistency and volume of product data.
- For inquiries that need judgment, such as refunds and claims, handoff rules to agents are defined together.
- On-premises operation requires upfront infrastructure cost and custom implementation work.
