Our work / Case study
Text and Voice AI Assistant for Customer Support, Built on Existing Inquiry Knowledge
Canadian Solar Japan
Context
What they needed
A high volume of repetitive inquiries, answered from knowledge scattered across call records, FAQs and inquiry logs.
What we built
Ekohe connected existing inquiry data, FAQs and product documentation to a conversational assistant that answers in text and voice, in several languages. Routine inquiries handled consistently, with less manual work for the support team.
Services and expertises
Details
Canadian Solar Japan's support team received a high volume of inquiries, many of them the same questions asked again. The answers existed, but they were spread across call records, FAQ pages, inquiry logs and product documentation. Each response meant searching, and consistency depended on who picked up the request.
The goal was not to replace the team's knowledge with a generic chatbot. It was to make what the team already knew available in every conversation, in the way people naturally ask: by typing or by speaking.
The assistant answers in text or voice, drawing on Canadian Solar Japan's own support history.
Ekohe built an AI-assisted support workflow connected to the existing inquiry systems. The conversational assistant draws on historical inquiries, FAQs and product documentation to answer questions in text or voice, in several languages. Because it reuses the knowledge already in place, nothing had to be rebuilt from scratch.
Routine and repetitive inquiries are now handled with consistent answers across customer touchpoints. The support team spends less time on manual handling, and more on the requests that need their expertise.
Text and voice, one source of knowledge
Text and voice interactions draw on the same knowledge base, so an answer does not depend on the channel someone chooses. Voice input lets people describe their question naturally, in their own words and language. The technology was chosen for this problem: Ruby on Rails and PostgreSQL for the support workflow, Next.js for the interface, and OpenAI models for language understanding. The inquiry data and the system remain Canadian Solar Japan's own.
Let's START