March 11, 2026
September 29, 2026
Editorial Disclaimer
This content is published for general information and editorial purposes only. It does not constitute financial, investment, or legal advice, nor should it be relied upon as such. Any mention of companies, platforms, or services does not imply endorsement or recommendation. We are not affiliated with, nor do we accept responsibility for, any third-party entities referenced. Financial markets and company circumstances can change rapidly. Readers should perform their own independent research and seek professional advice before making any financial or investment decisions.
Sales dip in March, so the owner of a small homeware shop runs a 15% off weekend. It works, sort of. Orders go up. Margin goes down.
The same week, 40 people asked whether she delivers to Scotland. Nobody answered them.
She is not careless. She was packing orders, answering a supplier, and doing the hundred other jobs that come with running a shop. The questions just arrived at the wrong time, on the wrong day, in the wrong inbox.
This is the pattern an AI chatbot for ecommerce is built to catch, and it is worth catching before anyone reaches for the discount button again.
When sales slow, the price is the easiest lever to pull. It is also usually the wrong one.
Take a product with a 40% gross margin. Knock 5% off the price and the margin drops to 35%. To earn the same gross profit as before, you now need to sell roughly 14% more units. That is our own illustrative sum, but the arithmetic is not up for debate.
Discount pricing feels generous. On a spreadsheet it looks more like a tax on your own business, which is why Robin Waite's guide to pricing your services argues against discounting just because someone asks.
There is a second cost, and it is slower. Run enough 15% off weekends and your customers learn the schedule. They stop buying at full price because they know a code is coming.
That is not loyalty. It is a waiting list for your next sale. The shoppers you most want, the ones who would happily pay full price, are the ones a habit of discounting teaches to hold off. There are customer retention strategies that work better than discounts, and most of them cost nothing at all.
When shoppers leave, they usually tell you why, if you ask. Baymard Institute's cart abandonment research found that, setting aside people who were only browsing, 40% of US shoppers abandoned because extra costs such as delivery were too high, 20% because delivery was too slow, and 19% because they did not trust the site with their card details.
Look at that list again. Two of the three are about delivery and its cost. The third is a question about trust. None of them says "your product is too expensive." An abandoned cart is often just an unanswered question with a basket attached. Answer the question and you reduce cart abandonment at the source.
If the answer-first approach is so cheap, why does almost every small shop reach for the discount instead? Two reasons, and neither is laziness.
Run a promotion and you can watch the orders arrive. It feels like doing something. An unanswered question leaves no trace at all. The shopper closes the tab, and nothing in your dashboard says "this one wanted to know about delivery to Aberdeen."
So the owner fixes the problem they can see, which is price, and never finds out about the one they cannot.
In a shop with two or three people, ecommerce customer service belongs to whoever has a spare minute. Packing orders wins. Supplier calls win. The inbox gets checked at lunch and again at nine at night, if you are lucky.
That is not a people problem. It is a coverage problem, and it is the part of the business an ecommerce chatbot is actually good at covering.
The leak is rarely on the product page. It is in the gap between a question and its answer.
How much is delivery to Aberdeen. Does the medium come up small. Can I return it if the colour is wrong.
Every one of those answers already exists somewhere on your site. The shopper just cannot find it at 10 p.m. on a Sunday, which is when they are actually shopping. An AI chatbot for ecommerce stores answers them there and then, from your own delivery, sizing, and returns pages.
Delivery is the big one. Shoppers rarely object to paying for it. They object to finding out at the last step. Tell them the price and the timing when they ask, and most of them carry on to checkout without a second thought.
Plenty of small shops have a chat widget installed. Far fewer have anyone watching it after six. A chat widget that says "we will get back to you within 24 hours" is not really ecommerce customer service. It is a very polite voicemail.
"Do you still have the navy one in medium? The site says sold out." That message is someone trying to give you money. Often the item is back in stock, or due next week, and the product page simply has not caught up.
If nobody replies, the shopper assumes the answer is no and buys elsewhere. If someone replies with the restock date and offers to let them know, you keep the sale. The difference is one message.
A first-time buyer does not know you. They want to know who ships the parcel, what happens if it arrives damaged, and whether a real person reads the inbox.
Those questions sound small, but they carry the whole sale. A vague reply makes them nervous. A specific one, with the returns window and how to start a return, does more for a shop they found yesterday than any trust badge in the footer.
The fix is not to never discount. It is to stop discounting to solve problems that were never about price.
Before your next promotion, look at last month's unanswered questions. If most of them are about delivery and returns, a discount strategy will not fix that. Answering them will. It is the cheapest way to reduce cart abandonment, and the only one that does not touch your margin.
This is ecommerce conversion rate optimisation at its least glamorous: nobody redesigns anything. You just make the answers reachable, and that alone can improve conversion rate without a single penny off.
None of this means discounts are always wrong. Clearing end-of-season stock, rewarding repeat customers, and launching a new line are all sensible reasons to cut a price for a while.
The test is simple. If you can say exactly which problem the discount solves, run it. If the honest answer is "sales are a bit slow and I am not sure why," look at the unanswered questions first. They are usually cheaper to fix.
It should answer from your own pages. It should never invent a discount you did not approve, promise a delivery date it cannot see, or guess at a returns rule. When it does not know, it should say so and pass the shopper to a person.
A good rule of thumb: if you would not want a new starter saying it on their first day without checking, the chatbot should not say it either. Confident and wrong costs more than slow and right, especially where refunds are involved.
Before the next promotion, give this a week. It costs nothing but an afternoon.
Export every question shoppers sent you last month, from the chat widget, email, Instagram DMs, and WhatsApp. Do not just count them. Read them.
Reading them is the useful part. You will spot the questions that come up every single week, and you will probably wince at how many of them nobody answered.
Delivery, sizing, returns, and stock. Almost everything will land in one of those four. Whatever is left is usually a genuine one-off that deserves a person anyway.
For each pile, find the page that answers it. Most will have one. The problem was never that the answer did not exist. It was that nobody could find it at the moment they needed it.
That means the chat, the DMs, and the product page itself. This is where an AI chatbot for ecommerce earns its keep, because it answers from those pages on every channel, at any hour, without anyone rewriting the delivery rules for the hundredth time. A Shopify chatbot or a WooCommerce one does the same job; what matters is where its answers come from.
If sales are still slow once the questions are being answered, you have learned something useful. Price may genuinely be the issue. Now a discount strategy is a decision, not a reflex.
You will not know whether this works unless you measure it, and the numbers worth watching are refreshingly small.
This is the number that matters most. If it falls, the leak is closing. If it stays flat, the answers are not reaching shoppers yet, and you know exactly which pages to look at.
Count the promotions you ran this quarter and compare them with last quarter. If the answer-first approach is working, you will need fewer of them to hit the same sales, which is the whole point.
The questions people ask are free product research. A spike in "does this come in a larger size" is a stock decision waiting to be made. Most sales AI tools will show you the volume. The useful ones also show you what went unanswered, which is where the next round of ecommerce conversion optimisation should start.
We build one of these, so here is how our AI chatbot for ecommerce holds up against everything above.
Agentency is trained on your catalogue copy, size guides, shipping rules, and returns policy, and every answer shows its source. The returns policy is quoted the same way to every shopper, and changing the document changes every future answer.
It answers tracking questions from your shipping rules and can call your order API through a Call Action for the live record. It installs with a script tag or the WordPress or Shopify plugin, so a Shopify AI chatbot is an afternoon's work. You do not need a developer, and you do not need to rebuild your store around it.
For restock alerts, a pre-chat form collects the shopper's email or phone number and adds them to your Customers list. That is the AI sales assistant part: a lost sale becomes a customer you can reach, without a discount.
The same agent answers on your storefront, WhatsApp, Instagram DMs, and Messenger, all from one knowledge base, and it replies in the shopper's language. The person who asked about the navy shirt in a DM gets the same answer as the one who asked on the website.
Refund exceptions escalate with the transcript, order reference, and shopper details attached. The conversations view shows which questions went unanswered and which missing pages cost you sales.
Your prices were probably fine. The delivery question to Scotland was the problem.
Before the next 15% off weekend, try answering the questions first. An AI chatbot for ecommerce does that around the clock, which is the kind of AI for sales that protects your margin instead of spending it.
Think of it as the colleague who never gets tired of the delivery question. Part shop assistant, part AI sales assistant, and entirely uninterested in giving away your margin to fix a problem that was never about price.
Agentency is an AI chatbot builder for businesses that want every customer question answered, day and night, without the guesswork. Teams add the documents and pages they already maintain, and the Agentency agent answers from that material only, with the source shown. When the answer is not there, it says so and hands the conversation to a person with the full transcript attached. One agent covers the website widget, WhatsApp, Instagram, Messenger, Telegram, Slack, and more than ten channels in total, and it replies in the visitor's language. Businesses in ecommerce, real estate, restaurants, IT support, and internal knowledge use Agentency to answer, resolve, and sell around the clock.
Omar El Bahr is a Senior Digital Growth Specialist at Agentency, where he leads SEO, content strategy, and organic growth across international markets. He is a Forbes Communications Council contributor and has written for Entrepreneur on business communication and digital strategy.
Show the full cost, delivery included, before checkout. Answer delivery, sizing, and returns questions the moment they are asked. Then fix the pages behind the questions nobody answered.
Start with the repetitive questions: order status, delivery, sizing, and returns. Let AI answer those from your own pages, and keep exceptions, complaints, and anything involving money with a person.
They can, when they answer the question that was stopping the sale. A chatbot that only sends links to your help centre will not. One that answers from your own delivery and returns pages, at 10 p.m., often will.
For a small shop, the main ones are coverage and consistency. Questions get answered in the evenings and at weekends, when most people shop, and every shopper hears the same returns rule. The team keeps the conversations that need judgment.
Remove surprises. Show delivery costs early, answer sizing and returns questions before checkout, and make sure stock information is current. Most small shops get more from fixing those than from redesigning the product page.
