Marketplace seller support, instant answers across 5 languages

The challenge

Spark hosts 4,000 sellers across Ukraine, Poland, Czechia, Romania and Hungary, and a support team of just three people fielding hundreds of seller questions a day in five languages. The math never worked: response times stretched, and complaints about slow support started threatening seller retention.

Sellers are the supply side of a marketplace; lose their trust and the whole platform suffers. Yet most questions were repetitive policy and how-to queries the tiny team had to answer over and over, in five languages.

Scaling the team five-fold across five markets wasn't realistic.

The solution

A multilingual AI support agent now answers sellers instantly in their own language, across all five markets:

• Auto-detects and replies in five languages, the Seller Support Agent detects the seller's language (UA/PL/CZ/RO/HU) and answers natively, removing the routing delay across markets.
• Trained on the seller knowledge base, policies and how-tos live in the Knowledge Base, with the Knowledge Search Agent connected to internal Notion docs so answers are accurate and current.
• Instant FAQ resolution, repetitive policy and how-to questions are resolved instantly on WhatsApp, Email and the website widget, freeing the three-person team.
• Complex cases escalated with context, harder cases pass to a human via Hand-off Control and into Zendesk with the full conversation, so nothing is re-explained.
• Consistent across markets, one instruction in the Configuration Wizard keeps policy answers identical in every language.

4,000 sellers across five countries got instant answers in their own language, slow-support complaints subsided, and the three-person team scaled without growing.

What was deployed