Advice in Online Shops: What an AI Chat Can Do

A physical store has salespeople who answer a question right away, pick the fitting product off the shelf and show an alternative when there is doubt. An online shop puts its goods in the window but usually leaves the visitor alone with their questions. This is exactly the advice gap a custom AI chat assistant closes when it knows the content of the website and the shop and helps in the conversation: it answers product questions, checks availability, explains shipping and returns and guides all the way to completion. The need is large, as German e-commerce turned over 83.1 billion euro (bevh) in goods in 2025, up 3.2 percent (bevh) on the previous year, while on average around 70 percent (Baymard Institute) of online shopping carts worldwide are not carried through to purchase. Between existing interest and an actual purchase there is a wide gap that advice at the right moment can narrow. This article shows the concrete use cases of a shop assistant, how it affects conversion and service load, what connecting it to a Shopware shop looks like and where the honest limit of what an assistant can do runs.

Enlarge graphic

Labels in the illustration

• Is the sneaker in size 43 in stock?
• Yes, size 43 is in stock —
• delivery in 2 days.
• Sneaker Aero
• Size 43 · white
• 89.00 euro
• Add to cart
• Free shipping from 50 euro
• Write a message...
• Four use cases
• ?
• Product questions
• Size, material, fit
• Availability
• checked when answering
• Shipping
• Cost and delivery
• Returns
• Clear return policy
• From doubt to purchase
• Open question
• Answer in chat
• Purchase
• 70%
• of carts are otherwise
• abandoned (Baymard Institute)
• The assistant clears questions before the cart is lost.
• Product questions, shipping and returns right in the chat — through to purchase

The Shop Assistant at Work

Where the Shop Lacks Advice

Online, almost everything that makes buying easy in a store is missing: the salesperson's glance, the quick answer, the reach for the right size. The visitor stands in front of a product page and has to work out for themselves whether the item suits them, when it will be delivered and what applies to a return. If they do not find the answer quickly, they leave the page. Yet for many, online shopping has long been routine: 34.2 percent (bevh) of online customers in Germany ordered goods online twice a week or more on average in 2025. Those who order this often know smooth processes from many shops and measure every other shop against them; an unanswered question then quickly becomes a reason to leave.

The competition for exactly this visitor has become tougher. A growing share of retail runs through online marketplaces, which now account for around 56 percent (bevh) of German online retail, and the market keeps growing: for 2026, bevh and the EHI Retail Institute expect nominal growth of 3.8 percent (bevh) in online retail of goods. For the individual shop this means that every interested visitor who leaves is likely to end up with a larger provider. All the more reason to lift the potential in the traffic that is already there instead of buying new traffic at a high price. An assistant that provides the advice no one else gives online starts exactly here. How such an assistant for online shops works in detail is shown on the corresponding industry page.

Four Use Cases in the Online Shop

The value of a shop assistant becomes concrete when you measure it against the questions that are really asked day to day. Four fields appear in almost every shop, regardless of the range: questions about the product itself, about availability, about shipping and delivery and about returns. On top come active purchase advice, which turns an open search into a manageable decision, and help right at the cart. Each of these fields is a place where, without advice, a share of visitors is lost and where an assistant counters at the right moment. The following six cards show what happens in each case in the conversation.

Answering Product Questions and Availability in the Chat

The most common question before a purchase is the one about the product itself: does the size fit, is the material right, is the device compatible, is the performance enough. An assistant that truly reads the catalogue answers this directly in the conversation and, where useful, shows the fitting product as a card with image, price and add-to-cart button. It invents no details but draws on the shop's real data, so the statement is accurate and legally clean. How these product cards are built is shown on the page about product cards in the chat; where the assistant draws its knowledge from is described in detail on the page about the knowledge base.

Availability is the second critical point. Nothing is more frustrating than deciding on an item and learning in the cart that the desired size is gone. Because the assistant reads current stock, it can answer the question of availability right away and, if something is sold out, suggest an available alternative directly instead of letting the visitor run into a dead end. This keeps the conversation moving. How closely product display and cart work together here is explored in the article product cards and cart in the chat.

Shipping and Returns Without Detours

No topic costs as many purchases at checkout as cost. Setting aside mere browsing, extra costs for shipping, taxes and fees that are too high are the most common abandonment reason in a survey of US online shoppers, at around 40 percent (Baymard Institute), at the top for years. On top of this, the total amount often cannot be seen in advance, which around 12 percent (Baymard Institute) of those who abandon give as a reason. An assistant defuses both before the cart is left: it names shipping costs and the free-shipping threshold openly in the conversation and sums up the total cost including shipping so no nasty surprise waits at the last step. If only a few euro are missing until free shipping, it can say so and suggest a fitting small item.

Returns are the second hurdle that slows many even before they buy. Someone unsure whether an item can be returned and what that costs either does not order at all or orders several sizes as a precaution, which in the end drives the return rate. The assistant explains return periods and conditions clearly and reliably, right at the point where the question arises. This removes uncertainty before ordering and can at the same time reduce unnecessary returns, because the expectation about fit or material is already clarified before the purchase.

Purchase Advice and Cart Help in Dialogue

Purchase advice is the point where an assistant stands out most clearly from search and filters. A search assumes the visitor knows what they are looking for; a filter assumes they know the right criteria. The assistant instead asks follow-up questions about occasion, budget and preferences and forms a small, curated selection from the answers rather than an endless list. This personalization pays off commercially: companies that excel at it generate around 40 percent (McKinsey) more revenue from these activities than the average. How such advice unfolds step by step is shown in the article guided selling in the chat and product finder; the matching feature set is bundled by the shop assistant.

Once the choice is made, the assistant helps right at the cart. A cart assistant adds items, changes quantities, removes positions, applies a voucher with permission and shows the subtotal, without the visitor jumping back and forth between chat and cart page. The cart thus turns from the last, often lonely step into a visible part of the conversation. If someone hesitates, the assistant summarizes the status and actively offers the next step instead of letting the cart quietly expire. How already filled carts are saved from being abandoned is explored in the article recovering cart abandonment with a shop chat.

The Effect on Conversion and Service Load

The first effect shows in conversion. Around 70 percent (Baymard Institute) of carts are abandoned, and how much room there is in completion is made clear by another figure: through better checkout design alone an average large shop can raise its conversion rate by as much as 35 percent (Baymard Institute). An assistant does not replace this design, but it smooths the friction that becomes visible in the conversation and catches doubts early. This weighs especially heavily on the smartphone, because a small screen, distraction and tedious input amplify every doubt. A chat that clears questions without a form or a page change pays off particularly on mobile.

The second effect concerns the service load. An assistant answers the same recurring questions about shipping, delivery time and availability around the clock and in any number of conversations at once, in the evening, at the weekend and at peak times. This noticeably relieves a small team and keeps the response equally fast in every case. The clear limit remains: where a case becomes sensitive or the assistant reaches its limits, it is meant to hand over to a human, with the full conversation history as context. How this handover works cleanly is described in the article AI assistant: handover to human agents, and in daily use it meshes closely with the support assistant.

Connecting to Shopware (Community Edition)

For product questions, availability and cart to work, the assistant has to read the shop and act in the conversation with permissions. For Shopware in the freely available Community Edition the assistant takes products, variants, prices and availability from the shop and syncs them on a schedule, shipping and return rules come from the stored settings and the knowledge base, and it carries out defined actions, for example assembling a cart in the chat or calculating the free-shipping threshold. Because it works on the real catalogue, the information shown stays current without content having to be maintained twice. The shop assistant thus fits into an existing shop instead of replacing it.

It is important that the assistant does not work past the checkout. The actual purchase, the payment and the legally required steps still run through the shop's regular checkout. The assistant prepares the cart in the chat and leads on into the shop with “Continue” instead of building its own parallel payment. Actions that change something happen only with clear permissions, so the assistant does no more than intended. Which functions make sense and how far they reach is defined through tool control and tailored to the individual shop in the project.

• Read products, variants, prices and availability directly from the Shopware catalogue
• Answer shipping and return rules from the stored details, not from a script
• Show fitting product cards in the chat from real data
• Extend the cart in the chat with items, change quantities and remove positions
• Calculate the free-shipping threshold and apply vouchers with permission
• Assemble a cart in the chat

Data Protection and the Honest Expectation

A chat that advises and sells inevitably processes information from visitors, from the product question to the data needed for the cart. Data protection is therefore not an afterthought but part of the build. Hosting and storage take place in Germany, complemented by a data processing agreement, a deletion concept and clear data sovereignty; the answers are generated by a specialised model provider, also outside the EU on the basis of standard contractual clauses. The assistant creates no hidden profile and does not use the conversations for external purposes. In German retail in particular, a visibly data-sparing approach builds trust, and trust is worth hard money at the moment of purchase. How this is regulated in detail is set out on the page about data protection and hosting.

For all its effect, an honest expectation belongs to the picture. An assistant can err, which is why it ties its statements to its own catalogue and knowledge data, is meant to hand sensitive cases over to a human and is continuously reviewed. A fixed increase in revenue or conversion cannot be assured, because it depends on the range, prices, target group and competition. What can be influenced are the factors that demonstrably work: less friction, clear advice at the right moment and fitting recommendations. What the chat achieves becomes measurable in the process. The conversation analytics show which questions come up often, where conversations tip over and where knowledge gaps exist, so the assistant improves month by month, because it builds on real conversations instead of assumptions. Which questions the assistant takes on in your range, and what product cards in the chat look like, is shown on the page for the assistant for online shops.

Frequent questions