Kaufland AI Assistant

AI Assistant Chatbot

Transforming Search: The Ultimate AI-Powered Shopping Experience

Overview

In November of 2024, the higher management at Kaufland reached out to us to perform product discovery and and development of an AI assistant chatbox utilizing GPT-based models to improve the product search experience on our e-commerce platform.

I was assigned the ownership of this initiative meaning I was responsible to drive the discovery forward and develop a MVP. This is an ongoing project and once completed, the solution will seamlessly integrate with our existing platform, ensuring compatibility across devices while maintaining robust data security and privacy. Ultimately, this innovative tool is expected to improve search accuracy, increase conversion rates, and drive overall business growth.

Role

Product designer

My Contributions

Product strategy
Research and interviews
Ideation
Low to high-fi prototyping
User testing

Team

1 Product manager
1 Product Designer
2 front-end developers
3 back-end developers

Tools

Figma | Miro
Optimizely
Contentsquare
Tableau

Project Timeline

November 2024- ongoing

The Problem

In the competitive e-commerce landscape, traditional product search functionalities often fail to understand nuanced customer queries and provide personalized recommendations, leading to user frustration and lower conversion rates.

The Goal

The goal of this initiative is to develop an AI assistant chatbot powered by GPT-based models that enhances product search capabilities on our platform. This chatbot will interpret natural language queries accurately, offer personalized product recommendations, and provide an engaging and intuitive user experience, ultimately improving customer satisfaction, increasing conversion rates, and driving business growth.

User Research

As part of my user research, I conducted interviews with 5 users to gather detailed insights into their needs and preferences.
I also distributed surveys to a wider audience to collect quantitative data.
Additionally, I read reviews and conducted social media research on AI assistants to understand broader user opinions and trends.

User Interviews

Surveys

Reviews & Social media

Findings from research

Difficulty with Current Search: Users often struggle with finding specific products using the current search functionality, citing issues with relevance and accuracy.
Preference for Natural Language: Users expressed a strong preference for a search tool that understands natural language queries, making it easier to find products quickly.
Desire for Personalization: Many users want personalized recommendations based on their past purchases and browsing history to enhance their shopping experience.
Importance of Quick Responses: Users emphasized the need for quick and accurate responses from the AI assistant to avoid frustration and improve efficiency during their shopping journey.

Competitive Analysis/ Benchmarking

To gain a comprehensive understanding of the market landscape, I conducted a competitive analysis by examining leading e-commerce platforms that utilize AI assistant chatboxes. This involved evaluating their features, user experiences, and customer feedback. I assessed key functionalities such as natural language processing, personalization capabilities, and response times. By analyzing strengths and weaknesses of competitor offerings, I identified best practices and potential gaps in the market.

Ideation

I conducted an in-depth analysis of findings from user research and benchmarking to identify and understand different user types and was able to categorize users into distinct segments.

In addition to researching user types, I also explored potential user journeys to map out how users interact with our e-commerce platform while searching for a product. By analyzing these journeys, I identified key touchpoints, pain points, and opportunities for enhancement.

I also made sure to involve the stakeholders, product manager and developers during every step of the discovery to get their feedback and to keep them constantly updated with the updates.

User Flow

During the ideation phase, I generated several logic flows to conceptualize the functionality and interaction of the AI assistant chatbox. These logic flows outlined the various pathways and decision points users might encounter, helping to visualize how the chatbot would respond to different queries and actions. This process was crucial in ensuring a cohesive and intuitive user experience, laying a solid foundation for the subsequent design and development stages.

Designs

Following the research and ideation phase, I created the initial designs and screens for the AI assistant chatbox. These designs were based on the insights gathered from user research and potential user journeys, ensuring that the chatbox would be intuitive and user-friendly. The preliminary screens served as a foundation for developing the prototype and visualizing how the chatbox would function in real-world scenarios.

Prototype

After completing the initial designs and screens, I developed a clickable prototype of the AI assistant chatbox. This interactive prototype allowed for hands-on testing and provided a realistic representation of the chatbox's functionality and user experience. By simulating real-world interactions, the prototype was instrumental in identifying any design flaws and gathering valuable user feedback before moving on to full-scale development.

Testing and Iterating

After creating the initial designs and prototype of the AI assistant chatbox, I conducted a Hallway test with three of our colleagues and users. This informal testing method allowed me to gather immediate feedback on the usability and functionality of the chatbot. The insights gained from these sessions were invaluable, helping me identify areas for improvement and ensuring that the final product meets user expectations and provides a seamless shopping experience.

Test went live on 04.12.2024 and was paused on 19.12.2024

Findings

Positives:

Nice recommendations of products
Interactivity
The tone of the responses from the Asistant

Concerns:

Switching between product ideas
Some products were outside price range
Displayed used products
Suggestions are not clickable
Limited length of processable chats
ChatGPT unaware of recent events

Next steps

Create more use cases
Create business case
Consult legal team
More tests