Available for new assignments now

IKEA chatbot

Conversational customer support and self-service.

The project in short

The design and implementation of the new chatbot in IKEA* started in fall 2019 and I was part of the project for 1,5 years. In the design, my role as the only UX designer for the last year in the team included tasks such as:

  • UX design
  • UI design
  • Prototyping
  • User testing

* In reality Ingka, IKEA’s largest retailer that accounts for 85% of total sales volume.

The challenge for IKEA

Customers reach out to IKEA for questions that they could potentially find the answer to themselves, or even better solve themselves by using existing self-service capabilities. One way to help users would be a conversational self-service, i.e. a chatbot.

The process of the design

Concept: A single focused companion

Offering self-service on an existing order

After parts of the concept were already developed, I joined the project as a UI designer but continued as the UX/UI designer after about six months. The concept so far has been the same – to understand the customer’s intent and to be able to answer the bulk of the questions

Research: Reasons for contact

Based on transcripts from existing chats in a few selected markets the most common intents were covered such as order status and product availability.

Solution: Meet Billie

The solution, Billie a gender-neutral chatbot that helped the user by offering conversational self-service, links to supporting web pages, fetching possible answers from IKEA’s knowledge management system, or if needed offering a handover to a co-worker during office hours.

UI framework with complete specs

I created the UI design framework for how a conversation would be presented to the user, including when Billie presented content that was not a conversation, e.g. external content.

For the intent analysis, IKEA is using Google Dialogflow which then passed on the intent to another system that supplied the answer flows to the users.

Learnings

Part of a order status check flow done in Miro

This was my first experience with a conversational design which for me was another process using a lot of flows in Miro. The flows I made were bounced with the backend developers to understand what was doable, and how the flows needed to be tweaked and/or improved to fit how the backend was set up.