A chatbot for your online store: when it helps and when it annoys
A chatbot can save your customer an email and save you hours of support. Or it can be one more window everyone closes. The difference is not the technology, it is what you use it for.
A chatbot on an online store helps when it answers recurring questions from your real data, such as order status, shipping, returns or product specifications, and when it can quickly hand the customer over to a person. It annoys when it pops up unasked, answers in generalities, or makes up answers it has nothing to base on.
Modern AI models can hold a natural conversation, so a chatbot no longer has to be a tree with three buttons. That is good news and a risk at the same time: a badly configured AI chatbot sounds convincing even when it tells the customer nonsense.
From our own practice we know that success is decided not by the model but by the source material and the limits. The same model gives good advice one time and confuses the customer the next, depending on what data it got and what it is allowed to say. That is why this article focuses more on when to deploy a chatbot and how to set it up than on choosing the technology.
When a chatbot on an online store helps
- Recurring support questions. "When will my parcel arrive?", "Can I return this?", "Do you deliver to pickup points?" If these questions make up a large part of your inbox, a chatbot handles them instantly and around the clock.
- Order status. If the chatbot is connected to the store and the carrier, it tells customers where their parcel is without you opening the admin.
- Choosing from a larger range. For technical goods it can advise by specifications, "which of these two vacuums is quieter?", as long as it has exact catalogue data.
- Outside business hours. In the evening and at weekends the customer gets an answer immediately, not on Monday.
When a chatbot annoys
- It pops up on its own and covers the content before the customer is even looking for anything.
- It answers in generalities, such as "Shipping depends on the delivery method selected", instead of a specific price and date.
- It makes things up. It promises a discount you do not offer or a specification the product does not have. You are responsible for what the chatbot says on your site.
- It cannot be bypassed. The customer wants to talk to a person and the chatbot sends them round in circles.
- It stands in for missing content. If shipping and returns terms are not clearly written on the site, a chatbot will not fix that. Fix the site first.
A chatbot or another solution?
Not every customer support problem needs a chatbot. Something simpler often helps:
| Problem | Better solution |
|---|---|
| Customers keep asking about shipping | clear price and delivery time right in the cart and on the product page |
| Customers cannot find products | better search that understands the customer's own words |
| Lots of "where is my order" emails | an automatic email with parcel tracking after dispatch |
| Complex advice when choosing | an AI chatbot over exact catalogue data |
| Recurring questions outside business hours | an AI chatbot with handover to a person the next day |
If customers mainly cannot find products, search is usually the better investment. We cover that in the article AI search in an online store.
What a good chatbot must be able to do
Answer only from your data
An AI chatbot should draw on your terms and conditions, shipping information, product catalogue and other texts you prepare for it. If the answer is not in that data, it should say so and offer a contact, not guess.
Be connected to your systems
Order status, stock availability and parcel tracking are the most useful answers, but they require a connection to the store and the carrier. Without that, the chatbot can only repeat what is on the website. That is a job for integrations and APIs.
Have clear limits
The chatbot must not give discounts, change orders or promise dates it cannot verify. These limits need to be set at deployment and tested by trying to talk it into things.
Hand the customer to a person
An always-visible option to write to a person, ideally without the customer having to retype their question. Complaints and claims should go straight to you.
Respect GDPR
Customers type their name, address and order number into the chat. You need to know where that data goes, how long it is kept and whether it is used to train someone else's models.
Off-the-shelf chatbot or a custom solution?
There are plenty of ready-made chatbots you add to a store by pasting a snippet of code. They can be enough to start with, especially if you only want to answer questions from the texts on your site. When choosing, check whether it can draw only on your material, whether it connects to your store and carrier, where conversations are stored and what happens to the data if you cancel. A custom solution makes sense when you need a connection to your own system, precise control over what the chatbot may say, or when you already pay for an off-the-shelf chatbot and the results are not right.
How to tell whether the chatbot really helps
After deployment, do not track the number of conversations, track what changed for customers and for you:
- Are there fewer emails and calls with the same questions?
- How many conversations ended with a handover to a person, and why?
- Which questions could the chatbot not answer? That is the list of what is missing from your site.
- Do customers complain about the chatbot, or use it without trouble?
In the first weeks, read at least some of the conversations. Nothing tells you more about your customers than what they ask in their own words.
How to start sensibly
Do not start with the chatbot, start with your support inbox. Go through the emails and calls from recent weeks and list which questions repeat. If it is the same questions over and over and you have the answers in your data, a chatbot makes sense. If it is a different problem every time that needs a person's decision, a chatbot will not help.
Then deploy the chatbot for a narrow set of questions, watch the real conversations and expand it gradually. Setup, integration and tuning are done by a senior with AI assistants at €38 / hour excl. VAT, and you get an hour estimate before we start. More in our AI for your business service, and if the store itself needs work first, see websites and online stores. For a broader view of where to start with AI in a company, read AI in your company – where to start.
Considering a chatbot for your store and not sure it is worth it? Write to us and we will tell you honestly whether you need one.
FAQ
Is a chatbot worth it for a small online store?
Yes, if support takes a lot of your time with recurring questions and the answers are in your data. If there are few questions or they differ every time, a clearer website and better search will help more.
Can an AI chatbot tell a customer something untrue?
It can, if it is not restricted to your data. A good chatbot answers only from your terms and catalogue, and when it has no answer it hands the customer to a person.
Can a chatbot tell customers their order status?
Yes, if it is connected to the store and the carrier via API. Without that connection it can only answer general questions from the texts on your website.
Is a chatbot GDPR compliant?
It can be. You need to know where conversation data goes, how long it is kept and whether the model provider uses it for training. We address this when designing the solution.
Related articles
AI agents in business: 5 tasks they can handle today
There is plenty of talk about AI agents, but most companies need to know one thing: what will save me time tomorrow morning. Here are five tasks where it already works.…
Read article →AI search in your e-shop and higher conversion
A shopper who uses your e-shop's search buys with a higher probability than one who clicks through categories. If search doesn't find what they want, they leave for a com…
Read article →AI in your company – where to start
AI is not a magic button that digitalizes your company for you. It's a tool that can save dozens of hours — if you apply it to the right problem. We'll show where it make…
Read article →The new era starts
with your brief.
Tell us what you need. A senior gets back to you, AI gets to work.