One bot per brand and per country, trained on each store's site and its Zendesk tickets

Nordbutiker AB is a Swedish e-commerce company focused on light electric vehicles: e-bikes, e-scooters, e-mopeds and electric mobility scooters for seniors. It runs its own web stores in Sweden, Norway, Denmark and Finland under its own brands, Evobike for e-bikes and Blimo for mobility scooters among them, and its own staff handle everything from product development and purchasing to customer service and the warehouse. The company started in 2006, has about 35 people at its office and warehouse in Norrtälje, 45 minutes north of Stockholm, and has been part of the Seafire group since 2019.

Visit Nordbutiker at nordbutiker.se →

Several brands, four countries, four languages, one support team. A question about a Rull scooter in Denmark and a question about an Evobike in Sweden need different answers from different product pages, and the team wanted each brand to sound like itself rather than one bot answering for all of them.

The stores share one platform, which made the first setup harder than expected: crawling one brand's site picked up every brand's sitemap. Pointing each bot at its own sitemap fixed that.

Nordbutiker set eesel up in autumn 2025 with a bot for each brand and country combination. The team did the site crawls and the ticket training themselves, working through the setup with us over a couple of weeks.

• Zendesk. Each bot learns from the solved tickets for its own Zendesk brand, so the answers reflect what that store's agents actually said.
• Website chat. The chat bubble on each store answers from that store's pages, crawled from its own sitemap.
• The browser extension. Agents use eesel inside Zendesk to draft replies and to summarise the ticket in front of them. An internal Summarize Bot exists just for that.
• Other knowledge. The team has also fed in YouTube product videos and looked at using its product feed, so the bots know the catalogue without crawling every product page.

The team has asked for one thing in particular: a summary of a customer's whole ticket history, not just the open ticket, for the cases where a customer starts a new ticket instead of replying to the last one. That request is with our product team.