AI-driven UX – how AI improves the shopping experience

Modern e-commerce is entering a stage in which the shopping experience is no longer evaluated solely through the prism of store aesthetics, speed of operation, or the number of features. What increasingly matters is whether the user can quickly find the right product, whether they receive relevant suggestions, whether the offer matches their purchasing context, and whether the entire process gives them a sense of simplicity and confidence in their decision. It is precisely in this area that artificial intelligence is beginning to play an increasingly important role. McKinsey points out that consumers increasingly expect tailored online interactions, and that AI and generative AI allow companies to scale the personalization of experiences far more effectively than traditional approaches. At the same time, commerce technology providers are developing not only recommendation mechanisms, but also semantic search, AI-based merchandising, and conversational commerce, all of which are intended to shorten the path from intent to purchase.

From our perspective, however, the most important thing is that AI does not improve the shopping experience on its own. It is not enough to implement a recommendation model, a chatbot, or a personalization engine for a store to suddenly become more intuitive. AI begins to bring real value only when it is embedded in a well-designed e-commerce architecture, has access to structured data, and supports specific moments in the shopping journey rather than merely generating an impressive add-on to the interface. In practice, this means shifting the thinking from the level of “how to add AI to the store” to the level of “how to use AI so that the user can buy faster, more easily, and with greater confidence.” This direction is consistent both with McKinsey’s observations regarding the scaling of personalization and with the development of AI tools for search, discovery, and recommendations on the side of commerce platforms and cloud providers.

AI-driven UX does not begin with graphics, but with removing friction

Many conversations about UX in e-commerce still focus on the visual layer, while the user’s real problem very often concerns something much simpler: they cannot find the right product, do not understand the differences between variants, get lost in filters, or lose confidence at the checkout stage. Research shows that the average cart abandonment rate remains at a very high level – currently 70.19% – and that problems related to checkout and the usability of the purchasing process still rank among the most common reasons for lost orders. This is important context because it shows that the shopping experience usually does not break down in a spectacular way. Most often, it falls apart in the places of everyday friction that the user encounters while searching, comparing, and completing a purchase.

This is precisely where AI-driven UX begins to make sense. Not because it replaces classic experience design, but because it helps to understand user intent faster and better adapt the interface to it. And this is where the need for personalization appears. Personalized commerce is an approach in which AI-powered search and discovery reduce the time needed to find the right product, support semantic understanding of queries, limit zero-result search cases, and allow for better arrangement of product order on category listings. Google Cloud, in turn, is developing Vertex AI Search for Commerce and conversational commerce precisely so that users can search for products more naturally and so that the store can respond better to imprecise, descriptive, or contextual questions. From the UX perspective, this means one thing: less effort on the buyer’s side and a greater chance that the path to the product will be shorter, more intuitive, and less frustrating.

This also changes the way we think about optimization. Not long ago, UX was improved mainly through manual layout testing, heuristic audits, and further interface iterations. Today, AI can support this process far more broadly – by analyzing behavioral patterns, identifying moments of declining attention, detecting recurring problems in shopping journeys, and helping to better adapt the logic of content presentation to actual user behavior. From our perspective, however, the most important thing is not to treat AI as a shortcut that bypasses the fundamentals. If the checkout is unclear, product data is inconsistent, and the catalog structure is chaotic, even the best model will not fix architectural or design errors. Fundamental usability problems still have a huge impact on cart abandonment, and AI has the greatest value when it strengthens a well-designed process instead of masking its weaknesses.

Shopping journey personalization is no longer an add-on, but is becoming the operating logic of the store

For a long time, personalization in e-commerce was understood rather superficially. Most often, it came down to displaying a few recommended products, reminding the user about an abandoned cart, or basic campaign segmentation. Today, this model is no longer sufficient. McKinsey emphasizes that as customer expectations rise, companies are beginning to use AI and generative AI to scale the personalization of experiences much more deeply, rather than merely for simpler targeting of messages. This means moving away from thinking about personalization as a separate marketing module and toward thinking about it as the operating logic of the entire shopping journey.

In practice, shopping journey personalization today means something much more than matching a banner to a user segment. The point is for the store to respond to the current context – purchase history, current behavior, source of entry, stage of the purchasing process, customer type, preferred delivery model, or assigned pricing logic. Shopware describes real-time personalization precisely as the ability to tailor content and experiences to the individual user based on their current interactions and behavior. In the B2B context, this scope is even broader because it may include dynamic pricing at the customer or project level, personalized dashboards, individual search results, or tailored payment and delivery methods. From our perspective, this is a very important shift because it shows that personalization is no longer an ornament of the shopping experience, but an element of its functionality.

This approach also has a very concrete business impact. If the user sees the right offer faster, does not have to work their way through irrelevant categories, and immediately reaches the right availability, price, and purchase variant, then the shopping experience becomes not only more pleasant, but simply more effective. The customer spends less time filtering and more time making decisions. From our perspective, this is exactly where the greatest value of AI in personalization lies: not in creating the impression of an “intelligent store,” but in shortening the distance between need and purchase. McKinsey and Shopware describe this direction in a similar way – as a shift toward more tailored, contextual, and real-time experiences.

Dynamic recommendations work best when they are part of the entire journey, not a single box on the product page

Product recommendations are one of the oldest areas of AI application in e-commerce, but their role is clearly changing. Until recently, many stores treated them as a simple cross-sell or upsell module, often operating according to rigid, manually defined rules. Today, dynamic recommendations are increasingly built as a layer that works across many points in the shopping journey simultaneously: in search results, on category listings, on the product page, in the cart, after purchase, and sometimes also in email communication and remarketing.

This is very important because the user does not experience recommendations in isolation from the rest of the store. For them, what matters is whether the suggestions are relevant, whether they appear at the right moment, and whether they help them make a decision rather than distract their attention. From our perspective, good recommendations should not only increase the basket value. They should also build the sense that the store understands the buyer’s needs, helps them navigate the offer, and reduces the number of unnecessary decisions. This is particularly important in stores with broad or complex catalogs, where the sheer number of products can work against conversion. In such environments, AI can support not only classic “customers also bought” scenarios, but also more advanced ones based on intent, product similarity, session context, or customer specificity.

It is also worth noting that modern recommendations are increasingly going beyond the moment of product selection itself. Shopware is currently developing features such as AI-generated checkout message, AI-generated summary of ratings, customer classification, search by context, or image search. These are important signals of the direction in which commerce platforms are evolving, because they show that AI is no longer confined to a single recommendation module. It is beginning to influence various layers of the shopping experience: from product discovery, through building trust on the product page, all the way to post-purchase communication. From our perspective, this is exactly the model that will gain importance – recommendations as part of a broader decision-support system, not a single widget added to the layout.

AI improves the shopping experience only when it has something to work with

The biggest mistake in thinking about AI in e-commerce is the assumption that the model itself will solve the problem of experience quality. In practice, the quality of recommendations, the relevance of personalization, and the effectiveness of AI-driven UX are only as good as the data, business rules, and architecture to which the AI has been connected. If the product catalog is inconsistent, attributes are incomplete, prices differ across channels, stock levels are delayed, and the promotion logic is unclear, then even the most advanced system will produce an experience that, from the user’s perspective, turns out to be inconsistent or misleading. In e-commerce today, there is a strong focus on semantic search, merchandising, recommendations, and discovery precisely because these areas require a solid data foundation in order to produce relevant results.

The same applies to governance. Organizations that achieve greater value from AI more often have clearly defined processes specifying when and how model outputs require human validation, and they also connect AI with workflow redesign instead of treating it as a loose experiment. In e-commerce, this is hugely important because AI already affects not only marketing content, but also the way products are searched, reviews are presented, customers are segmented, or recommendations are directed. From our perspective, the shopping experience truly improves only when AI operates within an organized decision-making model and does not undermine customer trust with incorrect, random, or inadequate suggestions.

In modern e-commerce, AI should strengthen what is already well designed

At CREHLER, we do not believe in the narrative in which AI will automatically fix every online store. From our perspective, a much more accurate approach is one in which AI strengthens well-designed foundations: a logical information architecture, a clear checkout, consistent product data, well-planned integrations, and a sensibly structured personalization layer. Only on such a foundation can an experience be built that is genuinely more intuitive, more contextual, and better tailored to the user.

That is exactly why the future of the shopping experience in e-commerce will not consist in adding more and more impressive features, but in connecting AI ever better with the real purchasing process. The user should find the product faster, understand the offer better, receive more relevant recommendations, and move more smoothly from interest to purchase. And at the same time, the brand should make better use of data, increase the accuracy of decisions, and build a more scalable growth model. In e-commerce projects – especially those developed on modern platforms such as Shopware – this increasingly means the need to combine AI-driven UX, personalization, and recommendations with architecture, integrations, and a good delivery model. From our perspective, this is where the real advantage begins, not at the level of the slogan “AI in the store” itself.