# Ecommerce Chatbot: What It Does and How to Set One Up

> What an ecommerce chatbot handles, how rule-based and AI chatbots differ, what to teach it and when to hand over to a person. Setup steps included.

Leonard Loo · CEO & Co-Founder, Volty

Updated Nov 5, 2026

6 min read

An ecommerce chatbot is software that chats with shoppers on your website, WhatsApp or social media, answers questions about products, delivery and returns, suggests items, and passes the conversation to a person when a customer is ready to buy something unusual, has a complaint or needs a refund. Done well, it answers the common questions at any hour and gives your team the conversations that need judgment.

## What a store chatbot handles

Most shopper questions fall into a small set. Group them, and you can see which ones a chatbot can take.

| Shopper question | Chatbot can answer | Hand to a person |
| --- | --- | --- |
| Does this come in another size or colour? | Yes, from the product page | If it is out of stock and the customer wants an alternative made |
| How long will delivery take? | Yes, from your shipping policy | If it is for a date that matters, such as an event |
| What is your return policy? | Yes, from your policy page | When the customer wants an exception |
| Which one should I buy? | Yes, by asking two or three questions | For custom or bulk orders |
| Where is my order? | Only with a feed of order data | Otherwise, always |
| I want a refund. | Collect the details | Always, for approval |

The pattern is that questions with one correct answer in your documents are the chatbot's, and questions that need a decision belong to your team. The chatbot's real value is that it stays available for the first group at two in the morning.

## Rule-based chatbot or AI agent

A rule-based chatbot shows buttons and replies from a script someone on your team wrote. It is predictable and fast to set up for five questions, and it fails on the sixth. Every change to a price, a policy or a product means rewriting the script.

An AI agent reads your own content, such as your product pages, shipping and returns policies and FAQs, and writes the answer in its own words. It can handle a question like "Is this jacket waterproof?" if your page says so, and it can say it does not know when the page is silent. What it does not do on its own is decide. You set what it may do and what must go to a person. Our guide to AI agents and chatbots compares the two in more detail.

## Where to put it

- Product pages and the cart, where hesitation happens. A chat bubble on these pages catches questions that would otherwise be an abandoned visit.
- WhatsApp, for customers who already have your number and expect updates there. See WhatsApp for ecommerce.
- Instagram and Messenger direct messages, because a lot of shopping questions begin under a post.
- Email, for customers who write to your support address.

Use one inbox for all four. Staff need one place to read a customer's history, and customers should not be asked for their name and order number again each time they switch channels.

Each channel has its own rules. Website chat can show cards, such as a product carousel or reply buttons, which makes browsing easier than plain text. On WhatsApp and the other direct message channels, Meta's messaging windows limit when a business can write first, so a chatbot there is mostly for customers who contact you. Plan the first reply for each channel accordingly.

## What to teach it

- Your product pages, with prices and variants kept current.
- Policies: shipping zones and times, returns, exchanges, warranty and payment methods.
- The twenty questions your staff answer most. Copy them from your last month of chats and write the correct answer once.
- Your tone, as two or three example replies you would be happy to see sent under your name.
- Hand-off rules: what it must pass to a person, who takes it, and what it tells the customer when it does.

## Set it up in five steps

- Pick one channel to start, usually your website chat or WhatsApp.
- Connect your store's website so the chatbot reads your real pages.
- Run it in a mode where a person approves replies, and read them for a week.
- Correct what is wrong and write the lesson in plain words. Fix the source page if the page was the reason.
- Let it reply by itself on the questions it gets right. Keep refunds, complaints and large orders with your team.

## Five conversations to design first

Before you launch, write out how each of these should go. Test them yourself on the live chat, then keep them as your regression list whenever you change a policy.

- Sizing: the shopper gives a height and a usual size, and the chatbot recommends a size from your size guide, with the link, and says what to do if it does not fit.
- Delivery: the shopper gives a postcode or country, and the chatbot gives the shipping time and cost from your policy. If the policy does not cover that place, it hands over rather than guessing.
- Return: the shopper asks to send an item back. The chatbot checks the item is inside the return window on your policy, explains the steps, and hands over for anything outside it.
- Gift: the shopper asks for a recommendation under a budget. The chatbot asks who it is for, then offers two or three products with links.
- Complaint: the shopper writes that an item arrived damaged. The chatbot apologises once, collects the order number and a photo, and hands over at once.

## Common mistakes

- Launching with a thin set of pages. A chatbot that hands over half its chats saves your team little time. Fill the gaps in your product and policy pages first.
- Hiding the way to a person. Always let the customer ask for a human, and say who will reply and roughly when.
- Setting it and forgetting it. Prices, stock and policies change. Read a sample of chats weekly and fix what you find at the source.
- Pushing every chat towards a sale. A shopper who asked about a return policy wants the policy.
- Running separate chatbots on each channel, with separate answers. Customers notice when the website and WhatsApp disagree.

## Measure it

- First reply time, in and out of office hours.
- Share of chats resolved without a person, and the share of those that the customer did not reopen.
- Chats handed over, and whether the summary saved your staff a question.
- Orders that came from a chat. Ask "how did you find us?" or tag chats that end in a sale.

One Volty customer, Darrell Er, Founder of Overmugged, said: "We saved 50 man hours each month, and improved conversion through faster replies to customers!"

## How to do this in Volty

Volty puts one AI agent behind your website chat, WhatsApp, Instagram, Messenger and email, answering from your own pages and documents. Add your store's address and Volty reads its product pages from the sitemap, up to 5,000 pages, re-read nightly, with its policies and FAQs, and sends the product link in its reply. Website chat goes onto your site with one script tag and can answer with cards: a product carousel, a product card, time slots, reply buttons. Volty brings in your staff when a customer is ready to buy, and each inbox can run in Off, Draft or Live mode. Read about Website and App AI, the Shopify integration, or look at pricing.

## Frequently asked questions

What is an ecommerce chatbot?

An ecommerce chatbot is software that answers an online store's shoppers in chat, on the store's website, WhatsApp, Instagram or Messenger. It handles the repeat questions about sizes, stock, delivery and returns, recommends products, and passes complaints, refunds and custom orders to a member of staff.

What is the difference between a rule-based chatbot and an AI agent for a store?

A rule-based chatbot follows menus you draw, so it breaks on any question outside them. An AI agent reads your product pages and policies and answers in its own words, including questions nobody wrote an answer for. It still needs limits, such as handing refunds to a person.

What should an ecommerce chatbot never do on its own?

Issue refunds, change an order that has shipped, promise a delivery date it cannot see, give medical or safety advice about a product, or argue with an upset customer. Set these as hand-off rules, so the chatbot names who is taking over and stops.

Where should I put an ecommerce chatbot?

Where your shoppers already ask questions: the product and checkout pages of your website, WhatsApp, and Instagram or Messenger direct messages. One inbox for all of them means a customer who moves from your site to WhatsApp does not have to start again.

How do I know if my chatbot is working?

Check four numbers each week: first reply time, the share of chats it resolved without a person, the share handed over with a complete summary, and the chats it got wrong. Read twenty real conversations every week for the first month.

## Recommended for you

How to Change WhatsApp to a Business Account (Step by Step) · How to Get WhatsApp API Access: Step-by-Step (2026) · What is RAG? How retrieval makes WhatsApp chatbots accurate · WhatsApp Business API messaging limits explained (2026) · WhatsApp Business and the PDPA: A Singapore Compliance Checklist · WhatsApp 24-hour rule: how the window works and what it costs · Best WhatsApp API Providers in 2026: 9 Options Compared · Conversational AI for customer service: how to design it well · GoHighLevel Alternatives: 6 CRM and Messaging Options

## Start your Sandbox Trial

Email us questions · Start your Sandbox Trial

Sign up for our 7-day trial. No credit card needed.

Future-proof your business with AI. Hear from our CEO

The Leonard and Volty Show · October 8, 2026

Meta now charges for WhatsApp replies

The Leonard and Volty Show · October 1, 2026

Too busy to hire?

The Leonard and Volty Show · September 21, 2026

From $500 chatbot jobs to agentic AI

The Leonard and Volty Show · September 10, 2026

Your AI is a new employee

The Leonard and Volty Show · September 4, 2026

Trust in the AI world

The Leonard and Volty Show · August 29, 2026

Does the AI model matter?

The Leonard and Volty Show · August 19, 2026

What are AI agents?

Meta now charges for WhatsApp replies. Here's what it means for your business, October 8, 2026

## Links

- [AI agents and chatbots](https://voltade.com/blog/ai-agent-vs-chatbot)
- [WhatsApp for ecommerce](https://voltade.com/blog/whatsapp-ecommerce)
- [Website and App AI](https://voltade.com/product/website-app-ai)
- [Shopify integration](https://voltade.com/integrations/shopify)
- [pricing](https://voltade.com/pricing)
- [How to Change WhatsApp to a Business Account (Step by Step)](https://voltade.com/blog/how-to-change-whatsapp-to-business-account)
- [How to Get WhatsApp API Access: Step-by-Step (2026)](https://voltade.com/blog/how-to-get-whatsapp-api)
- [What is RAG? How retrieval makes WhatsApp chatbots accurate](https://voltade.com/blog/improving-whatsapp-chatbot-understanding-with-rag)
- [WhatsApp Business API messaging limits explained (2026)](https://voltade.com/blog/mastering-whatsapp-business-api-your-guide-to-messaging-limits)
- [WhatsApp Business and the PDPA: A Singapore Compliance Checklist](https://voltade.com/blog/navigating-data-privacy-for-whatsapp-business-in-sg)
- [WhatsApp 24-hour rule: how the window works and what it costs](https://voltade.com/blog/navigating-the-whatsapp-24-hour-messaging-rule)
- [Best WhatsApp API Providers in 2026: 9 Options Compared](https://voltade.com/blog/best-whatsapp-api-providers)
- [Conversational AI for customer service: how to design it well](https://voltade.com/blog/conversational-ai-for-customer-service)
- [GoHighLevel Alternatives: 6 CRM and Messaging Options](https://voltade.com/blog/gohighlevel-alternatives)
