← All work
ai

LOVE&SAKE AI Assistant

Existing products and editorial knowledge turned into grounded sake conversations and recommendations.

LOVE&SAKE AI assistant welcome and question entry
Delivered conversation entry: people can begin with taste, budget or occasion rather than a context-free chat box.

Project facts

LOVE&SAKE already had rich editorial and product knowledge.

We made that content retrievable by AI and integrated conversation, recommendations, settings and feedback into WordPress and WooCommerce.

01WordPress/WooCommerce indexing
02RAG retrieval and sources
03Recommendations and feedback
04Workers, Vectorize, KV and D1

Working flow

Every question returns to real content and product data.

  1. 01
    Understand

    Interpret taste, budget, occasion and knowledge intent from the question.

  2. 02
    Retrieve

    Use article and available product records as the evidence for a reply.

  3. 03
    Recommend

    Explain the recommendation and return the visitor to a real product page.

01 01

Starting with knowledge the brand already owns

The system indexes articles and products, retrieves relevant passages for each question, then produces sourced answers and in-store recommendations. New site content enters the same indexing flow.

02 02

A system the team can inspect and improve

The customer widget shows answers, sources and products, while the admin side exposes settings, queries and feedback. Data sources and human controls remain inspectable.

03 03

Answers use the site’s editorial knowledge and live product data together.

The assistant retrieves articles and the latest eligible products, then uses semantic analysis and a higher-capability model to formulate a recommendation with a route back to the shop.

04 04

The assistant lives inside the established WordPress experience.

A small widget opens from the existing site instead of replacing its content structure. An admin workflow supports the manually managed product index and the next iteration of the service.

Next case

Envy Apple Campaign →
Working on something similar? Let’s talk →