E-commerce chatbot: sales and support in one conversation

The two-bot problem in online shops

In most online shops the customer conversation is split in two. Before the purchase, questions go to the sales team or a «product bot»; after it — «where is my order?», «the size doesn't fit» — to support. For the customer, though, it is one shop and one conversation. If the channel that gave size advice yesterday doesn't recognise them today with a return question, the experience breaks and the agent asks everything again.

An e-commerce chatbot creates value when it joins sales and support on one customer history. This article maps the shop's whole customer lifecycle: what the bot does before purchase, at checkout and after purchase, what data each stage needs, where it hands over to a person, and which figures show the result.

A lifecycle map

A shop's conversations fall into three stages, each with its own questions, data and risks.

1. 1Before purchase«Is this in my size?», «which is better?», «how long is delivery?». Data needed: the catalogue, variants, stock, delivery rules.
2. 2Checkout«Can I pay by card?», «is there instalment?», «the promo code doesn't work». Data needed: payment methods, promotion terms. Risk: a payment error is not the bot's to fix.
3. 3After purchase«Where is my order?», «I want to exchange», «the item arrived damaged». Data needed: live order status, the return policy. Risk: personal data and money.

Each stage has its own in-depth article; this one is about the rules that connect them.

Before purchase: from choice to order

Here the bot's job is to help the customer find the right product — not to list the catalogue, but to narrow the choice with two or three questions. The detailed flow is in the e-commerce product chatbot.

• Answers come from the catalogue itself: name, price, variant, stock
• When a customer sends a photo, the matching product can be recognised — especially useful for clothing and furniture
• A variant that is out of stock is not offered; a similar alternative is shown
• Delivery time and cost are stated precisely for the region

Vexvon's Shopify app syncs products, collections and variants with their own price and stock, as well as delivery and return pages; when a price changes in Shopify, the bot quotes the new one. On other platforms the same result comes from loading the catalogue into the knowledge base or through the company's API.

Checkout: where the bot stops

Checkout questions are short, but a mistake is expensive: a lost order.

1. 1The bot answersPayment methods, the delivery fee, promo code terms, whether instalments exist — all from approved material.
2. 2The bot hands overThe payment fails, money was taken but no order appears, anything involving card details. The bot asks the customer not to type card details in chat and passes the conversation to a colleague.
3. 3The bot does notApply a promo code by hand, change a price, or promise a discount.

The flow for orders left unfinished after checkout is covered in abandoned cart recovery.

After purchase: status, exchange, return

Most post-purchase questions are about where the order is. The bot can answer correctly only with live data — without an integration with the order system or courier, the answer stays generic. The status flow is set out in the order status chatbot.

• Status: with the order number and customer verification, only that customer's order
• Exchange: the rule is explained, size and item chosen, the request passed to staff
• Return: terms come from the shop's own page; a photo and order number are collected
• Damaged item: the bot promises nothing and hands over at once

One history: the rule that joins the stages

What joins sales and support is not having the same bot but having the same customer history. In practice that is four rules:

1. 1One customer cardA customer who asks about size on Instagram and about status on WhatsApp appears on one card — joined by phone number.
2. 2Pre-purchase context is visible to supportWhen a return request comes in, the agent sees which size was recommended. If the cause was wrong advice, it goes back to the knowledge base as a fix.
3. 3Support signals return to salesA frequently returned product, a size often asked about — information added to the product page and catalogue.
4. 4One toneThe sales bot should not be eager and the support bot curt. The customer is talking to one shop.

Example: one customer's two weeks

1. 1Day 1, Instagram«Is this linen shirt available in M? I'm 178 cm.» The bot checks stock, explains the choice between M and L from the size chart and sends the link. The customer orders.
2. 2Day 3, WhatsApp«Where is my order?» The bot recognises the number and finds the order: «It's with the courier and will arrive today.»
3. 3Day 6, WhatsApp«It's a bit tight — can I exchange for L?» The bot states the rule, checks stock of L and passes the exchange request to staff. The agent also sees the Day 1 conversation.
4. 4ResultThe customer never had to explain themselves twice. The team notes in the size chart that this model runs small.

What to measure

1. 1Before purchaseThe share of bot conversations that lead to an order; product questions most often asked but unanswered.
2. 2CheckoutThe share of orders completed after checkout questions; handovers for payment problems.
3. 3After purchaseThe share of status questions closed fully by the bot; reasons for returns.
4. 4TogetherThe share of conversations handled entirely by the bot; leads and orders by channel.

Without one history these figures sit in separate reports. The CRM joins them on one customer card.

Common mistakes

• Running two separate bots and two customer lists for sales and support
• Loading the catalogue once and never updating it — the bot offers items that are out of stock
• Answering status questions with «it'll arrive soon» without an integration
• Writing return terms in the bot differently from the page
• Never reading return reasons — a size chart error stays for months

Limits

The bot does not know what is not in the catalogue — a product with thin material gets a thin answer. Live order data exists only if the integration is built. Decisions on payments, refunds, compensation and damaged goods always stay with a person. The bot should not make advertising claims about product quality, only state the facts in the description.

Conclusion: one shop, one conversation

The value of an e-commerce chatbot lies not in its separate functions but in joining them. When pre-purchase advice, the checkout answer and post-purchase support share one history, the customer does not repeat themselves and the team sees why products come back.

Capabilities for Shopify stores are on the Shopify integration page, and every industry article is in this category. To map your store's lifecycle together, get in touch.

Frequently asked questions

1. 1Which shops is an e-commerce chatbot suited to?Any online shop that gets many questions by message, especially clothing, furniture and electronics, where size, variant and delivery questions are frequent.
2. 2What is needed to set it up?A catalogue (or a Shopify store), delivery and return rules, a way to integrate order status, and handover rules.
3. 3Does the bot fully replace agents?No. Simple questions close in the bot; payment problems, exchange approval and complaints stay with people.