I interviewed Marco Knecht from Jumbo for this article.

In this article, you will learn how a Swiss hobby and DIY store chain uses a Generative AI chatbot to advise customers in its online store, thereby significantly increasing the user experience and the overall success of the online store.

How is Jumbo already using Generative AI today?

Marco: At JUMBO, we use Generative AI as a central tool for customer interaction on our website. Our bot (JUMBot) is available around the clock to answer customers’ questions about DIY projects and offer them comprehensive support. Whether it’s woodworking tips, sewing instructions or ideas for creative projects, our AI is programmed to provide helpful and relevant answers. In addition, the bot can also recommend suitable products that meet the customer’s individual needs and preferences. In this way, we enable our customers to improve their DIY experience and realize successful projects by harnessing the power of Generative AI.

MTB: Good advice when buying a bike – the JUMBot provides a size recommendation and suitable bikes to go with it

What added value can be observed through the use of Generative AI?

Marco: By using Generative AI, we can identify and observe a variety of added values. This is particularly evident in our customer feedback, which is consistently positive. Our customers appreciate the opportunity to obtain comprehensive information and to be reassured in their decisions. It doesn’t matter whether they already have knowledge of a topic or are completely new to it. Generative AI provides exactly the information they need, acting in a similar way to the sales staff in our stores. This seamless integration of online and offline experiences closes one of the last major gaps in omnichannel. Customers can turn to our bot with confidence to ask questions, get advice and ultimately enjoy the best possible shopping experience. The use of Generative AI enables us to offer our customers a first-class service and further strengthen their trust in our brand.

Language: The JUMBot provides an answer in almost all languages

What has Jumbo learned so far with the new AI chatbot?

Marco: So far, we have gained valuable insights from the use of AI. One key advantage is the scalability of projects, which is made possible by the use of AI. One example of this is our bot, which provides different types of answers depending on the question, be it a step-by-step guide, a table or a simple text. In the past, we would have had to make individual specifications for each of these forms of response, whereas AI has this flexibility by nature.

Another learning point concerns linguistic diversity. Thanks to the use of AI, in particular Language Learning Models (LLM), our store can now serve almost any language without us ever having to carry out manual translations.

The targeted use of AI makes the implementation of projects significantly faster, more efficient and less error-prone. However, there is a risk of limited controllability. Nowadays it is no longer possible to check every single answer, as they are generated individually. Instead, the focus is on designing the framework conditions in such a way that the bot does not provide any incorrect or unwanted answers. These findings show that the correct use of AI brings significant benefits, but also comes with new challenges that need to be overcome.

What is planned for the future with regard to chatbots and generative AI?

The future of Generative AI at our company lies in deep integration into our website and our online store. Our bot currently functions as a standalone unit that customers must actively use in order to benefit from the advantages of AI. But we are striving to eliminate this separation and seamlessly integrate AI into the familiar store experience. In this way, customers can be unobtrusively supported by AI to improve the user experience and offer each customer added value.

Another focus is on optimizing our product data with the help of AI. We have already achieved some successes and will continue to intensify these efforts. This is because we recognize the importance of high-quality product data: it forms the basis for us to be able to advise customers better, which both helps customers and increases our sales.

What strategies are being pursued to improve the quality and diversity of the answers generated by Generative AI?

To improve the quality and diversity of the answers generated by Generative AI, we at JUMBO pursue a number of strategies. A central approach is to use AI to optimize and standardize all product data, including product titles, product benefits, description texts and technical attributes. This measure not only improves the classic on-page search and SEO ranking, but also the overall user experience. Our JUMBot, which accesses this data, also benefits from the increased data quality.

We are aware that product data is one of the main pillars of e-commerce. If something is not right here, no solid foundation can be laid for further developments. We therefore attach great importance to ensuring that our product data is of the highest quality.

By optimizing and standardizing this data, we can not only improve the performance of our Generative AI, but also continuously train the model. This ultimately leads to even better and more precise answers from our bot, which in turn improves the interaction with our customers and increases their satisfaction. Overall, the quality of our product data is a decisive factor for the success of our AI-supported solutions and for providing a first-class shopping experience for our customers.

Rako: The JUMBot can also provide complex answers, such as calculating the price per liter

How does JUMBO respond to potential challenges and risks associated with the use of Generative AI?

JUMBO responds to potential challenges and risks associated with the use of Generative AI with a comprehensive approach to controlling and monitoring the system. In contrast to traditional question-and-answer systems, where the answers can be checked and defined in advance, Language Learning Models (LLMs) generate individual answers to each question. This requires careful work in narrowing down the topics and ensuring that the responses generated meet expectations.

A key focus is to ensure that our AI only provides answers within the DIY topic area. We want to prevent the AI from addressing undesirable topics or performing tasks that are not in the interests of our company. We have therefore implemented sophisticated control options that allow us to precisely define the AI’s radius of action and prevent undesirable behavior.

Through these proactive measures, we ensure that the use of Generative AI at JUMBO is safe and responsible and that our customers can enjoy a positive and protected shopping experience.

No answer: For questions outside of JUMBot’s area of expertise, friendly and possibly helpful answers are provided.

Did you like the article?

This use case is part of my Generative AI Trend Report 2024. In addition to many facts and trends about Generative AI in the DACH region, the report also contains many interesting applications of Generative Artificial Intelligence in companies.