At a glance
tink is Europe's leading online smart home retailer, offering 200+ brands across multiple smart home protocols: from HomeKit and Matter to Zigbee. Operating across multiple European markets, tink helps customers build smarter homes with expert advice, a curated product catalogue, and seamless shopping experience.
HQ
Berlin, Germany
Industry
Ecommerce
Team size
Dixa features
Dixa Platform + Mim AI Agent
Email containment rate
Avg. email response time
Trustpilot rating
The Challenge
When Roger Sandmeier joined tink, Europe's leading online smart home retailer, the customer service operation was under pressure from every direction. The support team had grown to 16–20 full-time agents, yet customers were waiting up to 5 days for an email response. Customer satisfaction (CSAT) hovered around 30%. The board was unhappy with rising costs. And the agents themselves were burning out.
They told me: what am I doing here with my life? It's just answering the same answer the whole day. It's so annoying, boring."
Roger Sandmeier, Head of Customer Service
The root cause was familiar to any e-commerce support leader: an overwhelming volume of repetitive questions. Where is my order? Where is my refund? How does the warranty work? The same queries, over and over, burying the team and leaving no bandwidth for the complex cases that actually needed a human touch.
As a company with multiple suppliers, tink faced an added complication: when a customer asked about their order, the team often had to reach out to logistics partners first — adding days of delay before they could even reply.
The mandate Roger received was refreshingly simple: make it better.
"I didn't have a title when I started. I didn't have a job description. It was just: build your own job. You want to have better customer support — do it."
Finding the Right Tools
Roger's first move was migrating tink from Zendesk to Dixa. Having worked with Zendesk at a previous company, he knew it wasn't the right fit for what he wanted to build.
The AI journey started with Solvemate, a traditional rule-based chatbot–functional, but limited. When Dixa acquired Solvemate and announced it would build its own AI-native solution, Mim, Roger made the switch without hesitation. Two things mattered to him: having everything in one platform, and trusting the team behind the product.
"I knew you guys before and I worked with you before. I was really sure it will have a really good outcome when you do something."
The fact that both Dixa and the original Solvemate team were European companies also mattered. As Roger put it: "We have an industry here and you don't have to go all the time to the USA. We have good people here that can do good product."
Rolling Out Mim: Bold Moves, Fast Results
Roger's rollout approach was methodical at first and then fearless. He spent about a month testing Mim on a low-volume market, just himself, poking at edge cases and refining prompts. Then he did something most support leaders wouldn't dare: he rolled it out to tink's largest market.
By Black Friday 2025, instead of 80–90 seasonal hires, tink brought on 20 people and even that turned out to be too many. They scaled back to 10 within weeks. The longest email wait time during peak season? About one day, on Black Friday itself.
The Results
Roger's permanent support team went from 16–20 people to just two full-time agents and four working students. No layoffs, just natural attrition. People finished their studies, moved on, and tink simply stopped replacing them. All of this while the company was growing.
"The company was not shrinking in sales numbers, it was growing. More customers, more questions. But smaller team.
We just don't hire new people. And it's working. The people are not stressed. They can handle it with ease."
On customer sentiment: Roger is refreshingly honest that reactions are polarised. Some customers don't even realise they're talking to AI. Others immediately demand a human. But the overall picture tells its own story, tink's Trustpilot rating climbed from 3.5 to 4.6 stars.
"When you see the bigger picture with the reviews from the customers, they are happy."
Roger's Advice: Just Do It
When asked what he'd tell a peer at another company considering an AI agent, Roger doesn't mince words:
"Don't be afraid. Don't get stuck in the past. Don't fixate on old numbers. The world is changing, the customers are changing, and the numbers are changing."
His practical advice:
- Start by understanding your tickets. Sit down, spend a week doing tickets yourself. See what keeps coming back.
- Don't wait for perfection. Roll it out, learn, iterate. The AI gets better as you feed it more information and refine your prompts.
- Ignore early CSAT dips. They'll recover as the bot improves, or you'll find better metrics that actually reflect reality.
- Push for it, even if no one's asking. Don't wait for the board to tell you. Start small, prove it works, then show the numbers.
"When I was going back, I would even push it more. Just do it. Not waiting until something really big happens."
What's Next
Roger's vision for Mim at tink goes well beyond support. The next frontier is product consulting, using Mim to help customers navigate Tink's complex catalog of 200+ brands and multiple smart home protocols (HomeKit, Matter, Zigbee, and more).
"It's more like a colleague at the moment than just a machine that helps you or a tool. We tested a lot with product consulting. One of the board members already tested it and was happy."
tink is also preparing to expand internationally with a new Shopify-based shop, with Mim integrated from day one via Dixa. What started as a cost-reduction initiative is becoming a competitive advantage.
"The more you integrate and the more you get better at prompting, the more Mim evolves from just an AI agent to really a helping colleague."
Your team could be next.
From 5-day response times to 50 minutes. From 80 seasonal hires to 10. tink's results didn't happen overnight, but they started with one decision. See what Dixa and Mim can do for your team.