Skip to content

AI in retail: tasks for the shop and the online store

By the KI können editorial team8 min read

Written with AI and checked against the sources by a second AI. As of November 15, 2026. More in the notes on AI.

In retail, the checkout, the stock and the customers often leave little time for texts and plans. An AI can draft product descriptions, notices and reply templates for you. What it doesn’t know is your goods, your prices and your rules. You give it those and check the result yourself at the end.

In a small shop, a shop assistant with a tablet stands behind the counter, a customer looks at the shelf, crates of goods in front.

Key points

  • The AI writes product texts from your real details. Material, measurements and ingredients come from the label or the supplier.
  • For customer questions, notices, onboarding and social media, it delivers good drafts and templates.
  • Check prices, unit prices, discounts, allergens and legal texts at the source, not in the chat.
  • Your customers’ names, addresses and orders don’t belong in a personal AI account.
  • Never write reviews yourself or have them written. Realistic-looking AI images get a label.

Which tasks fit in the shop and the online store

An AI like ChatGPT, Claude or Gemini is good at phrasing and sorting. But it has never held your goods in its hands. Everything it writes about material, size or origin must come from your details. Otherwise it adds what is typical for similar products, and that sounds right even when it is wrong.

Tasks in retail and what you check
TaskWhat you give itWhat you check
Product textData sheet, label, your notesMaterial, measurements, care, contents
NoticeOccasion, date, timesDates and times
Reply template for customer questionsYour rules on delivery, exchange, paymentWhether every promise is right
Onboarding planTasks, routines, contact peopleWhether it fits your shop
Social media planRange, dates, toneWhether everything is real
Order calculatorYour calculation ruleThe result for a past week

In the trial lesson you put your first instruction together and get feedback. No account, in a few minutes. Start the trial lesson

Product texts from real details

Rosa runs a small online store for hand-dyed yarn. Every new color needs a description, and after the twentieth she runs out of ideas. She gives the AI the details from the label and her own notes on the color and has it write a text that doesn’t make anything up.

Sample instruction
Write a product description for my online store for hand-dyed yarn.
Details from the label: 75 percent virgin wool, 25 percent polyamide, 100 grams, about 420 meters per skein, needle size 2.5 to 3.5 mm, cold hand wash.
My notes on the color: dark moss green with lighter patches, every skein a little different.
Structure: one sentence on the impression of the color, then the details as a list, then one sentence on what it suits, such as socks or shawls.
90 words at most, informal, calm, no superlatives.
Don’t invent properties like “doesn’t itch” or “sustainable”.
If you are missing something, write [missing: …].

Before listing the product, Rosa checks every detail against the label: fiber content, weight, length, care instructions. She also deletes sentences like “perfect for allergy sufferers” or “eco-friendly” if the AI adds them anyway. You may only make such claims if you can prove them.

Notices and replies to customer questions

Lucia is a sales assistant in a bakery with three branches. For annual closures, changed opening hours or Easter pre-orders she regularly needs notices. She gives the AI the occasion, dates and times and has it lay out a short text in large print. Afterwards she checks the dates against the calendar digit by digit.

Ben runs a bike shop with a workshop and a small online store. Many e-mails ask the same things: how long does a service take, can I pick up an online order in the shop, how does an exchange work? He has the AI write templates that contain his rules and placeholders for everything that changes.

Sample instruction
Write three reply templates for customer e-mails to my bike shop.
1. Question about how long a service takes. Rule: book via [link to booking page], usually takes [number] working days, pick-up until 6 p.m.
2. Question whether online orders can be picked up in the shop. Rule: yes, after our pick-up e-mail, bring ID.
3. Question about exchanges. Rule: don’t answer this yourself; insert [text from our withdrawal information and exchange rules].
Each template: greeting with [name], five sentences at most, formal, one clear next step at the end.
No promises that are not in my rules.

The third template matters most. Withdrawal, exchange and statutory warranty are legal questions. Take the answer from your checked legal texts instead of letting the AI phrase it. Fill in the name placeholders only in your e-mail program.

Task plans and onboarding new colleagues

Thomas is a trained retail salesperson and manages a shoe shop. In the fall, two part-time helpers start. So that their first week doesn’t get lost in shouted instructions, he has the AI draft an onboarding plan, which he then checks with his team.

Sample instruction
Draft an onboarding plan for the first week of a part-time helper in a shoe shop.
She works Tuesdays, Thursdays and Saturdays, five hours each.
Tasks she should learn: checkout and gift cards, receiving and pricing goods, measuring feet and advising, accepting exchanges by our rule, closing the shop.
For each day: two or three learning goals, who goes with her (as a role, such as “store manager”), and a short question to end the day.
As a table that fits on one page.
Don’t add your own rules for the checkout or exchanges; I will add those.

Thomas puts the plan next to the real routines in the shop. What the AI cannot know about the checkout or exchanges, he adds himself. From the same plan, the AI can then make a checklist to tick off.

Rosters without names

An AI can also make suggestions for shift and task plans, such as how to share Saturday shifts fairly. Use codes instead of names, and leave out sick notes and private reasons for requests entirely. Whether a plan fits working hours, the collective agreement and break rules, check yourself or ask whoever is responsible for it in your business.

Ideas for social media

Ida is a florist with her own shop. She wants to show something regularly, but not think every morning about what. She has the AI draft a plan for a month: one fixed day a week, topics from her range and the dates in the month, with a suggestion for picture and text for each post. She takes the photos and posts them herself.

With pictures, a second look pays off. A picture of your goods should show the goods as they are. If an AI image shows a bouquet that looks real but doesn’t exist like that, it misleads customers. In some cases the EU AI Act then also requires a label. A short line such as “AI-generated image” is honest in any case.

A small order calculator

An AI can also build you a small web page that calculates something for you, as a single HTML file you open in the browser. Lucia from the bakery wants to know how many loaves to order for Saturday. The rule comes from her, not from the AI.

Sample instruction
Build an order calculator as a single HTML file that runs in the browser without internet.
Inputs: number sold on the last four days of the same weekday, leftovers in the evening on those four days.
Calculation: average number sold plus 5 percent, rounded up to whole pieces.
Also show the average leftovers in the evening.
Empty fields or negative values: a clear error message instead of a result.
Large fields, so it is easy to use on a tablet at the counter.

Before Lucia orders by the calculator, she enters the figures from a past week and works the result out once by hand. If something is wrong, she describes exactly which input gives which wrong result and what would be right.

Where to look closely

  • Prices and unit prices: for goods sold by weight, volume, length or area, the price usually has to be accompanied by the unit price, such as per kilogram (in German law: Grundpreis). Both come from your checkout system or costing, not from the AI.
  • Discounts: if you advertise a price reduction, you usually have to state the lowest price of the 30 days before it. Don’t let the AI invent “was” prices.
  • Legal texts: legal notice, withdrawal information, terms and conditions and guarantees come from checked sources, such as your trade association or a lawyer.
  • Customer data: names, addresses, orders and phone numbers don’t belong in a personal AI account. Use placeholders like [name] and add the real details afterwards.
  • AI chat on your website: if an AI answers your customers’ questions directly, it must be clear to them that they are writing with an AI.
  • Health or environmental claims: “helps with” or “eco-friendly” only if you can prove it.

How to practice this

Whether it’s a product text or an onboarding plan, what matters is giving the AI the right facts, forbidding it to make things up and checking the result carefully. In the app “KI können” you practice this in short lessons with examples from retail and put the instructions to work in your own AI. The trial lesson needs no account. You can find every topic in the course overview.

Sources

Information as of November 15, 2026. Laws and offers can change; this article is not legal advice.