How to Automate Repeat Customer Questions in Your Online Store

Most of the messages an online store receives are the same few questions asked again and again: where is my order, which size should I pick, can I return this, is it back in stock. Answering them by hand is slow, and slow answers cost sales. This guide shows how to automate those repeat questions without handing your customers to a bot that guesses: how to find out which questions you actually get, how to write the answers once, which kind of tool to choose, how to test before going live, and when a person must take over.

It is a product-neutral guide. Tekeralab does not sell a customer-support agent today; the last section explains where we stand.

TL;DR

• Start from your own messages, not a template. Collect two to four weeks of customer messages and sort them into groups. Most stores find that a handful of groups cover most of the volume.
• Write each answer once, in one place. A short knowledge base of policies and question–answer pairs is what every automation tool, simple or AI-based, depends on.
• Automate the questions that have one correct answer: delivery times, return rules, sizes, payment options. Keep people on anything that needs judgement or involves money: damaged items, refunds outside your policy, angry customers.
• Test with real past questions before switching on, and make "talk to a person" easy at every step.
• Measure your own numbers: first-response time, how many questions are answered without you, and how many are handed over.

Step 1: Find out which questions you actually get

Before choosing a tool, look at your inbox. Copy or export the last two to four weeks of messages from every channel you use: WhatsApp, Instagram direct messages, your marketplace's question tab and email. Then sort them by hand. A spreadsheet with three columns is enough: the question as the customer wrote it, its group, and whether answering it needed information about a specific order.

Typical groups for a small online shop:

Count the messages in each group. The point is not a precise statistic; it is to see where your time actually goes, and to start with the two or three groups that are both frequent and simple.

Step 2: Write the answers once

Every way of automating answers, from a saved quick reply to an AI assistant, is only as good as the written answers behind it. Keep them in one document and keep that document current:

• Policies: delivery times by region, shipping costs, the return and exchange window, how refunds are paid, warranty terms. Write them in plain language, the way you would say them to a customer.
• Product facts: materials, size charts, care instructions, what is and is not in the box. If your catalog already holds these, point to it instead of copying.
• Question–answer pairs for the groups you found in step 1. For example:
  • Q: "Do you ship to rural addresses?" → A: "Yes. Delivery takes 2–4 business days and the courier calls before arrival." (Use your real times.)
  • Q: "Can I change the size after ordering?" → A: "Yes, within 24 hours of ordering. Reply here with your order number."
• Hand-off rules: which situations must always reach a person, and how quickly you promise to answer them.

When a policy changes, change it here first. An automatic answer that quotes an old return window is worse than a slow answer.

Step 3: Choose how to answer

There are four common ways to handle repeat questions. They are not exclusive; many shops combine two.

On answer quality: your own replies and saved replies are as accurate as the answers you wrote, keyword rules work only for the wording they were written for, and an AI assistant is reliable only when it is limited to your own documents and tested.

A practical path for most small shops: start with saved replies for your top groups, so you answer faster even by hand. Move to automatic answers only for the groups where the answer never depends on judgement, and only after the knowledge base from step 2 exists.

If WhatsApp is your main channel, the channel's rules matter as much as the tool: the Business app and the Business Platform (API) do different things, and messages you send after the 24-hour customer service window has closed need approved templates. Our WhatsApp sales guide on app vs API, the 24-hour window, costs and consent covers those rules.

Step 4: Test before customers see it

Do not switch on automatic answers blind.

1. Take 20–30 real questions from your export, including awkward ones: spelling mistakes, two questions in one message, a question in another language.
2. Run them through your setup and read every answer as if you were the customer. Is it correct? Is it in your tone? Does it promise anything you cannot deliver?
3. Check the hand-off: send a complaint, a refund request outside your policy and an angry message. Each one must reach a person, and the customer must be told that a person will reply.
4. When an answer is wrong, fix the knowledge base, not just that one reply. Then run the same questions again.

With an AI assistant, test one more thing: ask questions your documents do not answer. The right behaviour is "I'll check with the team" and a hand-off, never an invented delivery date, price or policy.

Step 5: Keep people in the loop

Automation should make your answers faster, not make you harder to reach.

• Say when a reply is automatic and offer a person in one tap or one word. Meta's WhatsApp Business Messaging Policy, for example, asks businesses that automate conversations to offer a clear path to a human.
• Never automate decisions about money or fault: refunds outside your policy, damaged goods, compensation. Automation can collect the order number and photos; a person decides.
• Read a sample of automatic answers every week, and add the new questions you see to the knowledge base.
• Think about personal data. Names, order numbers, addresses and phone numbers in customer messages are personal data. If an outside service processes your messages, check where it stores them and whether your privacy notice covers it. In Turkey that means KVKK (Law No. 6698); this is not legal advice.

Step 6: Measure what changes

Track a few numbers from your own data, before and after:

• First-response time: how long a customer waits for the first useful reply, inside and outside business hours.
• Answered without you: the share of conversations closed by a saved or automatic reply.
• Hand-off rate: how many conversations reached a person, and whether they should have.
• Wrong answers: every automatic answer you had to correct. One wrong policy answer matters more than ten fast correct ones.
• Questions to orders: how many conversations that started with a question ended in an order.

Fast replies matter most when a buyer is hesitating, often on a phone. Our 2026 marketplace seller data looks at mobile conversion and cart abandonment, and is a useful companion when you decide which questions to answer first.

Where Tekeralab fits

Tekeralab does not sell a chatbot or a customer-support agent today. Our live product, Smart Publisher, schedules social media posts. Answering customer messages is one of the roles we are designing for Smart Employee, our planned AI employee, but it is not available as a product yet; we explain what exists and what does not in how we are designing Smart Employee.

We know this workload first-hand: before Tekeralab, we ran a marketplace where answering the same buyer questions every day was part of the job. That story is in the real products behind Tekeralab.

Frequently asked questions

Which customer questions should I automate first?

The ones that are frequent and have one correct answer that does not depend on judgement: delivery times, return rules, sizes, materials and payment options. Leave complaints, damaged items and refunds outside your policy to a person.

Will customers mind getting an automatic reply?

Most customers want a fast, correct answer. What frustrates them is a reply that ignores their question, or a bot they cannot get past. Say when a reply is automatic and make it easy to reach a person.

Can an AI assistant give wrong information?

Yes. AI models can give confident answers that are not true. Limit the assistant to your own written policies and product facts, test it with real questions, including ones your documents do not answer, and make it hand over to a person when it is unsure.

Do I need the WhatsApp Business API to automate answers?

Not always. The WhatsApp Business app has quick replies and greeting and away messages, which is enough for many small shops. Automatic answers connected to your store, several people on one number, or messages you send after the 24-hour window has closed need the Business Platform (API) and approved templates.

Does Tekeralab offer a customer-support agent?

Not today. Tekeralab's live product is Smart Publisher, which schedules social media posts. Customer support is one of the roles we are designing for Smart Employee, and we will say so plainly when it is available.