From 100% Tickets, to 50/50 Tickets and Strategy

CUSTOMER SNAPSHOT

Goodiebox is a subscription e-commerce brand that sends thousands of curated self-care boxes to members across its markets each month. The team describes its audience as members rather than customers, and the support function as Team Happiness. Maria Lauridsen, Head of Members and Retention, runs a KPI-driven operation with short time-to-market cycles and a growing member base. The ambition she set for the team was not only to work more efficiently, but to make the work itself better.
‍

THE CHALLENGE

Goodiebox's support process was heavily manual. Agents were spending real time on questions like where-is-my-order, over and over, in volumes that made up more than half of all combined inquiries. Repetitive by definition, and repetitive in a way that did not develop the people answering them.

Maria framed the problem as two problems stacked on top of each other. The operation needed to run better. The people running it needed a more effective setup.

"We needed to work smarter and more effectively, there were so many tasks that were repetitive. My mission was to lift performance but also create more interesting tasks for our Team Happiness Heroes."

She was also thinking ahead, past the current shape of the function.

"What will be the new structure for Team Happiness? Will full-time support heroes in the end be doing data analysis, member insights, optimizing processes and much more, while part-timers will do tickets or what is the future here?"

An AI Support Agent was the lever. The question was which one, and what kind.

THE SOLUTION

Goodiebox evaluated several vendors and narrowed the shortlist to two. The selection criteria Maria cared about were concrete: local presence, proven e-commerce cases, a user-friendly interface and a partner who understood what Goodiebox was trying to do.

The deciding moment was the architectural choice. One vendor pushed Goodiebox toward a purely generative setup. Kindly recommended a conversational setup, with the option to mix in generative where it made sense.

"Kindly recommended the conversational one and also said that we easily could have a mix between them both, where Boost.ai pushed for the purely generative one. Kindly always claimed that conversational was the best match for us and it became clear that they were thinking of our business needs."

Talking to an existing Kindly customer in a similar position helped Maria see what the post-decision reality would look like.

"We got in touch with Kindly's customer Hobbii. That really helped a lot since they were already in the process of making exactly the same changes as us."

Kindly was selected, and the AI Support Agent went live across the Goodiebox website and app. It handles the repetitive member questions, asks whether the answer helped, and routes to a human for the cases that actually need one, like payment issues.
‍

THE RESULTS

Goodiebox's initial target was 30 to 50% automation. The team hit 50% quickly, raised the bar to 60%, and passed that too. Today the AI Support Agent handles 72% of inquiries.

Fallback has moved in the right direction alongside it.

"We have a really low fallback rate. We started at 3% and now it consistently stays around 1.3%. We are very satisfied with these numbers."

The operational number that matters most to Maria is not automation rate. It is what it freed up.

"I think the most important result is that we have freed up resources and are now in a place where we can grow and develop our employees much more. And I think we were surprised that we reached the automation level of 50% so quickly."

Full-time agents who used to spend 100% of their time on reactive tickets now split their work roughly in half between tickets and analysis. The team uses that analysis time to predict payment and shipping issues before they hit the inbox, which changes the work from reactive to proactive.

Member engagement with the AI Support Agent is also strong on both website and in app. The greeting rate sits at 83% in the app and 61% on the website, which means members are actively engaging with the AI Support Agent.

This shows how customer service can scale in a way that benefits both members and the people delivering it.