Results

AI transformation for traditional industry, starting in southern Taiwan

We don't build showpieces. Every item here is an AI system currently running for a client, or one that has passed acceptance testing.

Ice-cream manufacturerAI customer service deployment

Acceptance passed
PROBLEM
High volume of B2B quote requests and after-sales messages, with human support going offline after hours.
SOLUTION
A semantic judgement layer (not a keyword table), automatic hand-off of urgent cases to a human, group notifications, and a dedicated support console — with two LINE accounts kept strictly separate.
RESULT
20 out of 20 on the client's own judgement test, and 7/7 on the second protection layer.
The knowledge base was distilled from four official documents and 90 real conversation logs.

Restaurant chainInstagram campaign automation

Live
PROBLEM
Running Instagram giveaways by hand was slow and error-prone.
SOLUTION
End-to-end automation of the giveaway process, delivered as a monthly subscription.
RESULT
Live in production, on a paid monthly subscription.

Pharmaceutical maker, two brandsAI agent team

In service
PROBLEM
A traditional manufacturer running both a B2B ingredients line and a B2C consumer brand — two audiences with entirely different messaging.
SOLUTION
Three isolated AI agents, each holding its own brand knowledge base and channels, with no cross-contamination.
RESULT
All three agents in service.

Before this, we built entire websites for traditional-industry clients through full automation — now we point that same automation at your operations.