Your store gets the same handful of messages every day. Does this kurta run small? Do you deliver to my pincode? Is cash on delivery available? Where is my order? Each one takes a minute to answer, and each unanswered one is a shopper who may have closed the tab and bought somewhere else.
Taking that load off you is the usual reason to add a chatbot for ecommerce website visitors. It can do a lot of it, but only if you are clear about which questions it can answer from what it already knows and which ones need it to look into your store’s live data. That split decides how useful it will be, how much work the setup is, and how much can go wrong.
Where shoppers get stuck before they buy
Start with the shopper’s point of view. Most pre-purchase questions fall into a few groups:
- Fit and suitability: sizing, materials, compatibility, which variant to choose.
- Delivery: whether you ship to their area, roughly how long it takes, what shipping costs.
- Payment: whether COD is offered, which UPI apps and cards work, whether EMI is available.
- Returns and exchanges: the window, the condition rules, who pays for the return.
- Trust: whether the products are genuine, whether there is a real business behind the site.
Look at your own messages for a week and sort them into these groups. The biggest groups are where a bot earns its place.
Questions a store bot can answer from fixed content
Many of those questions have the same answer for everyone. A bot can answer them from your policy pages, product descriptions and a written FAQ, with no connection to your store’s database at all:
- Your return and exchange policy, in plain words.
- Whether you offer COD, and any rules attached to it. If you are still deciding those rules, see setting up cash on delivery in WooCommerce.
- Your size guide, explained conversationally (“if you are between sizes, go one up for this fabric”).
- Care instructions, warranty terms, what is in the box.
- Which regions you ship to, if that is a simple list.
This is the easy, low-risk layer, and it is where most small stores should begin. The setup is mostly writing: making sure the bot’s source content is complete and does not contradict itself.
Questions that need live store data
Other questions depend on the current state of your store:
- “Is the medium in blue back in stock?”
- “What is the price of this today?” (if you run sales often)
- “Where is my order?”
- “Can you deliver to 7001xx by Friday?” (if your courier’s serviceability changes)
To answer these, the bot needs a connection to your store. WooCommerce has a REST API that a chatbot integration can read, and Shopify has its own APIs and an app ecosystem for this. Many chatbot platforms offer a ready connector for one or both. The connector is the part that takes testing, because a bot that reads stock wrongly is worse than one that says “please check the product page”.
A reasonable rule: if a question changes answer more than once a week, either connect live data properly or have the bot point to the place on the site where the current answer lives.
Order tracking without leaking order details
“Where is my order?” is the question store owners most want to automate, and it is the one that needs the most care. An order record contains a name, phone number, address and what someone bought. A bot that shows that to anyone who types an order number is a privacy problem.
A safer pattern:
- Ask for the order number and the email or phone used at checkout.
- Only proceed if both match the same order.
- Show the status and the tracking link, not the full address or item list.
- For anything beyond status, such as a change of address, a cancellation or a damaged item, hand over to a person.
Often the better first step is to send proactive updates, so fewer people need to ask. Automatic WhatsApp order updates for WooCommerce cut “where is my order” messages at the source. For a broader look at what a bot should and should not handle with customer data, see the post on chatbots and customer data.
Helping people find the right product
An AI chatbot for an online store can also act as a guided search. A shopper says “I need a gift for my father, he likes cooking, under a certain budget” and the bot suggests a few products with links. This works when:
- your product titles and descriptions are detailed and accurate,
- categories and attributes (size, colour, material, use) are filled in consistently,
- the bot is told to recommend only products that exist and are in stock.
If your catalogue data is thin, a recommendation bot will expose it. Improving product descriptions helps your SEO and your human shoppers anyway, so it is not wasted work.
Things a store bot should never do
These are the limits worth setting first when planning an ai chatbot for small business website project for a shop. Write them into the bot’s instructions and test them:
- Never invent a discount or coupon. AI models can be talked into “offering” a deal. Only mention codes you have listed for it.
- Never promise a delivery date unless it reads that date from a reliable source.
- Never approve a refund, return or cancellation. It can explain the policy and start a request; a person decides.
- Never collect card details or UPI PINs in the chat. Payment happens at checkout only.
- Never argue. An angry customer goes to a human, fast.
WooCommerce and Shopify: practical notes
On WooCommerce
Most chatbot tools install as a WordPress plugin or a script added to the site. Watch for the chat widget covering the add-to-cart button or the sticky checkout bar on mobile, and for caching or script-optimisation plugins delaying the widget. If the bot reads order data, it needs API keys with the narrowest permissions that work, and those keys should be stored by the platform, not pasted into a page. When something in the store side breaks, that is normal WooCommerce support work rather than a chatbot problem.
On Shopify
Chat tools usually come as apps from the Shopify App Store. Check which permissions an app asks for when you install it, and remove apps you stop using, since they keep their access until uninstalled. Shopify also has its own inbox tool; compare it with third-party options before adding another subscription.
Language and tone for your shoppers
Shoppers write the way they talk. Expect Hinglish, short forms, voice-typed messages with odd spellings, and questions with no punctuation at all. Test your bot with messages like “size chart kaha hai” and “cod milega kya 7001xx” if that is how your customers write, not only with tidy English sentences.
Decide the bot’s tone too. A fashion store can be friendly and casual; a store selling medical supplies or electronics spares should be careful and precise. Whatever you choose, keep replies short. On a phone screen, three lines and a link beat a paragraph.
Finally, give the bot a plain way to say it does not know, in every language it uses. “I’m not sure about that one, let me connect you with our team” is better than a guess, whichever language it is in.
How to tell whether the bot is paying off
Skip vanity numbers like “conversations started”. Look at:
- Transcripts. Read a sample every week at first. Are the answers right? Where does the bot get stuck?
- Handover rate. If most chats end up with a person, the bot is not covering the right questions yet.
- Repeat questions in other channels. If WhatsApp and email still get the same COD question, the bot is not reaching people, or is not answering it clearly.
- Complaints about the bot. Even one “your bot told me X” that turns out wrong is worth fixing immediately.
Chatbot for ecommerce website setup: a sensible order of work
- Write down your top customer questions and the correct answers.
- Fix your policy pages so they say the same thing.
- Launch a bot that answers only those, with a clear route to a person.
- Add proactive order notifications to reduce tracking questions.
- Only then consider live data connections for stock or order status.
Working in this order keeps each step small, and you can see whether it helped before paying for the next one.
If you want help setting this up on your store, tell us your platform and your most common customer questions. Our chatbot setup is quoted on request once we know both.
Next in this series: How to Train a Chatbot on Your Website Content
