AI chatbots for online shops: what they handle and what they don't

Tuesday, quarter past eleven at night. Someone ordered four days ago and fills in the contact form: "where is my parcel??". By the next morning it's sitting among a pile of near-identical messages. Someone on your team opens the backend, finds the order, copies the tracking number, writes a polite reply. Then the next one. Then the next.

If you run an online shop, you know this kind of message. None of it is hard. It just never stops, and it tends to arrive when nobody's at their desk. That's where an AI chatbot earns its keep. This article looks at what it can actually take over in a shop, where it runs out of road, and what's sensible to expect.

The four topics that dominate shop chats

When we go through incoming questions with shop owners, they nearly always fall into the same groups:

• Where's my order?
• Do you have this in my size or colour, and when will it arrive?
• How do I send something back?
• Which product is right for me?

The first three can largely be answered automatically. The fourth too, with a bit more groundwork. Then there's everything else: complaints, special requests, bulk orders. Those still need your team.

Order status: where the bot helps most

In most shops, "where's my order" is the question. With a connection to your shop system, the bot can answer it on its own. The customer gives their order number, the bot looks it up and replies with the current status.

In My-Chatify this works with Shopware. The bot connects to your Shopware catalogue, checks order status by order number and answers questions about delivery times.

Here's a real one. At Wilesco, the bot is connected to the Shopware catalogue, looks up order status and hands special cases to the team. Follow-up questions about order status have dropped by 82 %, with 1,240 conversations a week going through the bot.

That bit about special cases matters. If tracking says a parcel was delivered and the customer says it wasn't, you don't want a bot arguing the point. It passes the conversation to a person, who sees the whole history in the inbox.

Product advice with product cards

The second big group is product questions. "do u have the hiking boots in a 9 under 120" is a pretty typical message. A bot that only knows your website copy can only give a vague answer to that.

With the Shopware connection it's a different story. The bot knows your products, prices, variants and stock levels in real time. Customers can search and filter by size, colour and price right in the chat, and matching items show up as product cards with an image, price and buy button. From chat to basket is one click.

A short exchange might go like this:

Customer: do u have the hiking boots in a 9 under 120

Bot: I found two models in size 9 under 120, and both are in stock. [two product cards with image, price and button]

Customer: which one is waterproof

Bot: According to the product descriptions, the second one has a waterproof membrane. The first is listed as water-resistant.

What the bot won't do is invent features. If your product description says nothing about waterproofing, it has nothing to go on. So the quality of the advice depends directly on the quality of your product data. In nearly every shop project we come across items with thin descriptions. The bot makes those gaps visible, because the questions it couldn't answer show up in the analytics.

Returns

Returns questions automate well, as long as your rules are clearly written down. How long do I have? How do I get a returns label? Who pays for postage? When do I get my money back?

The bot takes these answers from your sources, such as your shipping and returns page or Q&A pairs you add. It's worth a critical look here. Does the returns page say the same thing as your terms and conditions, and as the leaflet that goes in the parcel? If not, customers will get different answers depending on which source the bot finds the better match.

Also think about when the bot should hand over. "How do I send this back" it can handle. "It arrived broken, here's a photo" belongs with your team.

What to realistically expect

A chatbot takes the repetitive work off your team. It answers the same questions at any hour, Sunday evenings included, and in whichever language the customer writes in. It doesn't replace the person who decides whether to make a goodwill gesture on a complaint.

A few things worth knowing up front:

• The bot is only as good as your data. Patchy product descriptions and inconsistent delivery information show up in the chat straight away.
• The first few weeks need some care. Check the unanswered questions regularly and fill the gaps.
• Some customers just want a human. That should always be possible.
• Results like Wilesco's depend on the shop, the range and the kind of questions coming in. You'll only see what changes for you once it's running.

Shopware, Shopify and other shop systems

With Shopware and Shopify, My-Chatify connects straight to your catalogue, so the bot searches your live products rather than a copy from last week. Order status lookups currently work with Shopware only.

On any other shop system the bot still takes on a good deal: shipping, returns, payment methods, opening hours for a physical shop, general product questions. It reads your product pages as a website source and keeps them current through automatic re-crawling. You just won't get live stock levels or order status.

Where to start

With whatever eats the most time already. For most shops that's order status, followed by shipping and returns. We walk through the setup step by step in how to set up an AI chatbot on your website.

You can try My-Chatify for free. Create an assistant, give it your shipping and returns page, and see in the Playground how it copes with the questions your customers ask. If you'd like to know what the Shopware connection would look like for your shop, our team in Dortmund is happy to talk it through.

All articles