bonprix

Sascha Netuschil

Shaping Innovation through Data

Coming from an AI and Data Science background, one of the main mindset advantages is that you do not think about customer-centric solutions or solutions in general within the limits of what is possible to handle for a human in regards to granularity, complexity and frequency. Instead you think within the limits of the data available, the algorithms and models and the technological infrastructure. In this way, you can expand the solution space for a given problem by a big margin and drive innovation.

Let us take personalization of the customer interaction as an example, where you usually collect data, analyze it and make decisions about customer approaches based on these analysis. For a human to be able to do this, you need a tool like customer segments, which reduces the granularity of the data from millions of individual cases to a handful of segments and thereby to a level a human can handle. The same goes for complexity (focusing on a small number of parameters) and frequency (the process is repeated with low frequency). With this you end up with a labor-intensive and slow process that still treats millions of customers in the same way.

If you approach this problem not with a human-limit, but a technology-limit-mindset, you can treat every customer individually, include a wide range of data and signals generated by the customer and build automated processes that adapt in real time.

Scaling AI for Impact: Navigating Challenges to Deliver Business Value

One of the primary hurdles we have identified is the composition of our teams. While a few talented Data Scientists might be enough to prove a concept, scaling AI and truly exiting the PoC phase requires a much broader set of capabilities. This means bringing together not just Data Scientists, but also skilled Machine Learning Engineers and Product Managers.

Another critical aspect is the natural fear of “loss of control” when automation takes over tasks traditionally performed by humans. To overcome this, we place great importance on establishing monitoring frameworks. Interestingly, we have found that leveraging “human-friendly” approaches, such as monitoring algorithms in the context of familiar concepts like customer segments, helps immensely in building trust.

“Do not get lost in the hype of the technology itself. Focus on where AI and data science can genuinely create value”

Finally, the emergence of Generative AI (GenAI) created the need for new skillsets. Traditional Data Scientists often lack the specialized knowledge required for working in Generative AI, for example prompt engineering. Our strategy is the upskilling our existing talent through learning-by-doing and internal knowledge sharing.

AI-Powered Retail Smarter Experiences Streamlined Operations

One of our objectives to enhance customer experience is the personalization of the customer journey. Through the use of machine learning models, we utilize large amounts of data from browsing behavior and purchase history to real-time interactions. This allows us to dynamically tailor product and promotion recommendations, article rankings, and content to each individual and their real-time context.

When it comes to Generative AI, our strategy is currently focused on optimizing our internal processes and reducing their costs rather than customer-facing solutions. By doing so, we do not only achieve immediate cost savings and efficiency gains but also prepare ourselves for future customer-facing use cases by gaining technological experience and building up a solid and scalable AI infrastructure.

However, we also have such internal optimization projects with an impact on the customer experience. One example is our use of a multimodal AI model to generate textual descriptions of our visual content. These descriptions are then used in our online shop for accessibility purposes for vision-impaired people.

Aligning AI Strategy with Market Trends and Customers

Regarding market trends in Generative AI, you might feel a sense of missing out on the “latest and greatest” technology, if you do not adopt it right away. However, in our experience you can build impactful solutions on proven models and use cases.

Regarding broader customer expectations and market trends, our strategy is that of fast followers not early adopters. This means we aim to leverage established technologies and use cases that have already proven their value in the market. To ensure we are still on the right track, we reassess our technological approach within the context of new projects, monitor the market and, crucially, also conduct our own UX research, getting direct feedback from our customers to understand their evolving needs and expectations.

Guiding AI Integration for Retail Technology Success

First and foremost, my advice is approach the topic from a clear value perspective. Do not get lost in the hype of the technology itself. Focus on where AI and data science can genuinely create value. Prioritize use cases that promise high value relative to their complexity. Also, prioritize internal stakeholders who are truly passionate about these features. Their commitment will be invaluable in project success and in driving adoption.

Secondly, it is vital to think beyond the Proof of Concept (PoC) from day one. A successful PoC is a great start. However, the real challenge is in productizing and maintaining these solutions at scale. Make sure, you have all the resources available to achieve this goal.

Finally, regarding Generative AI, my advice is to start smart and build experience incrementally. I recommend you first look at internal processes that involve a lot of manual, repetitive work or significant external costs. Internal stakeholders will often love these solutions because they can see the effect directly in their departments. Based on these experiences and first successes, you can build confidence to approach more complex, customer-facing solutions.