Can a Small Shop Use AI Without Having an Online Store?

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Can a Small Shop Use AI Without Having an Online Store?
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Can a Small Shop Use AI Without Having an Online Store?

Yes. A shop with no website or online store can use AI on its till reports, its Google Business Profile, shelf signs and window cards, customer messages on WhatsApp or Instagram, supplier emails and staff product knowledge. The best starting point is the till: most point-of-sale systems export sales data, and some now answer plain-English questions about it.

Without an online store, AI won't sell anything for you. Its value in a physical shop is saving time and helping people find you. Being found is increasingly done by AI too: when someone asks Google or a chat assistant for a gift shop open on Sunday, the answer draws on your Business Profile and reviews. For many small shops, an incomplete Google profile is a bigger gap than a missing website.

Follow me on Instagram@sagnikteaches

Your till already holds the most useful data

Every card and cash sale through a modern till is a record of what sold, when and for how much. That's enough for AI to answer the questions shopkeepers usually answer by feel: which days are worth an extra pair of hands, which lines have stopped selling, what sells together.

Connect on LinkedInSagnik Bhattacharya

Some systems now let you ask directly. Square, for example, lets sellers ask Square AI questions about their sales, customers, staff and transactions in plain language from the dashboard, where the feature is available. With other systems, export a sales report by item and date as a spreadsheet and upload it to a chat assistant on a plan that doesn't train on your data.

Subscribe on YouTube@codingliquids

For an illustrative card and gift shop, a first prompt might be:

Attached: 12 months of sales by item from our till (columns: date,
item, category, quantity, net_sales). We're a small card and gift shop.
1. Which 3 weekdays and which 3 months have the highest sales?
2. Which 10 items sold fewer than 5 units in the last 6 months?
3. Which categories grew or shrank most, comparing the last 3 months
   with the same 3 months a year earlier?
Show the figures you used. Don't guess at anything not in the file.

An illustrative answer, trimmed:

1. Busiest days: Saturday ($1,840 avg), Friday ($1,210), Thursday ($980).
   Busiest months: December, November, May.
2. Slow items (under 5 units in 6 months): Luxury candle - fig (2),
   Leather notebook A4 (3), Birthday card range "Botanics" (4 of 12
   designs)...
3. Growth: Wrapping and ribbons +38%. Decline: Jigsaws -41%.
   Note: "Cards - misc" appears only from March; earlier sales may
   have been recorded under individual card names.

What to check: the May spike is probably Mother's Day in your trading calendar, which matters for planning next year; the note about "Cards - misc" is the kind of data quirk that can make a category look like it grew when it was only renamed, so look at it before acting. And the jigsaw decline might be a supplier stock problem rather than demand. AI gives you the questions to ask; the till can't tell you why. Forecasting stock and reordering with AI takes the same data a step further.

Getting found when people ask Google or an AI assistant

For a shop without a website, the Google Business Profile does the work a homepage would. It's also a main source for AI answers about local businesses. Google phased out the public questions-and-answers section on Business Profiles in late 2025, and answers about places now lean on the profile's own details, photos and reviews. That makes accuracy and completeness more important than ever.

What to do, with AI helping at each step:

  • Fill every field. Categories, a full description of what you sell (brands, product types, services like gift wrapping or key cutting), accessibility details and all opening hours, including holidays. AI can draft the description from a list of what you stock; you correct it.
  • Show what's on the shelves. Google's Local Inventory app can sync products from compatible tills to your profile automatically. It needs a verified profile, a compatible till or a wired barcode scanner, and products with manufacturer barcodes, and it's only available in some countries; handmade or custom-barcoded goods won't match. You can also add products manually with Google's product editor.
  • Answer Google's inventory questions. Eligible shops may be asked quick yes-or-no questions about products they carry. Answer them accurately; misrepresenting stock breaks Google's policies.
  • Reply to reviews. AI drafts, you add a specific detail. Replies show people, and Google, that the shop is active.

The description field is a good first job. Give a chat assistant a plain list (cards for every occasion, local artists' prints, candles, kitchenware, children's gifts, free gift wrapping, step-free entrance) and an illustrative draft comes back: "The best gift shop in town, offering an eco-friendly range of unique, handpicked treasures for every occasion." Three fixes. "The best in town" is a claim you can't back, and Google's description guidelines rule out promotional content, citing "best bagels in town" as the kind of line not to write. "Eco-friendly" is a claim about products you haven't checked. And "treasures" says nothing a searcher would type. The edited version: "Independent card and gift shop: cards for every occasion, prints by local artists, candles, kitchenware and children's gifts. Free gift wrapping. Step-free entrance." Plainer, and every word is something a person or an AI answer might match.

A realistic mistake to avoid: a shop asks AI to write a Google post for a public holiday weekend, the draft says "Open all weekend, 10am to 4pm", and nobody checks it against the actual plan to close on Monday. Customers arrive to a locked door and leave reviews about it. Every AI-written post or description gets checked against your real hours and stock before it's published. Put the check in the routine rather than relying on memory: whoever posts reads the hours line aloud against the shop's own calendar first. Optimising your Google Business Profile with AI goes field by field.

Words on the shop floor

Shelf talkers (the small cards that sell a product on the shelf), window cards, event posters and gift-guide tables are where a physical shop's voice lives. They're also a job most owners put off because writing 30 of them takes an evening. AI takes it to 20 minutes, provided you give it the facts.

An illustrative before and after for an independent hardware shop:

BEFORE (handwritten, faded):
"Good sealant. Ask staff."

PROMPT: Write a shelf card, max 25 words, for [brand] clear
anti-mould bathroom sealant, 310ml cartridge, $9.50. Needs a sealant
gun (we sell them, $6). Tone: practical, friendly, no exclamation
marks.

AFTER (AI draft, edited):
"For showers, baths and sinks. Stays clear, resists mould.
310ml, fits a standard gun (ours are $6, aisle 3). Ask us how
to get a neat bead."

The edit added "aisle 3" and the offer of advice, which is the part that turns a shelf card into a sale. AI supplied the structure; the shopkeeper supplied what only a shopkeeper knows.

Behind the counter: product knowledge for staff

A Saturday assistant who can't answer "which paint for a bathroom ceiling?" or "is this book part of a series?" loses sales. AI can turn supplier PDFs, product data sheets and your own notes into a one-page crib sheet per category. Here's an illustrative filled-in section for an independent bookshop:

CRIB SHEET: CHILDREN'S BOOKS (updated 1 Sep)

"Something for a 7-year-old who loves animals"
  -> Chapter books shelf B2: [series A] (start at book 1), [series B]
"Reluctant reader, age 9-10"
  -> Short chapters + pictures, shelf B3. Graphic novels shelf B5.
"Is this part of a series?"
  -> Check the spine number; if none, check the till's product notes.
Gift wrapping: free, takes 5 minutes, paper under the counter.
Special orders: most arrive in 2-4 working days. Take name, number,
  ISBN if they have it. Never promise a date.

Upload the supplier information, ask the assistant to summarise it into that format, then have your most experienced staff member correct it. The correction step is essential, because AI summaries of product sheets sometimes merge two products or drop a warning. Laminate the result and keep it by the till.

Answering messages without a website

Customers who can't browse your stock online message instead: "Do you have X?", "Are you open Sunday?", "Can you put one aside?". The free WhatsApp Business app lets you set a greeting message, an away message and quick replies; Instagram has saved replies. AI helps you write a good set once, then draft individual replies for anything unusual.

An illustrative filled-in set of quick replies for the gift shop:

  • /hours: "We're open Tue-Sat 9:30-5:30 and Sun 11-4, closed Mondays. Late opening Thursdays in December until 7."
  • /stock: "Let me check the shelf and get back to you within the hour. If you'd like one put aside, we can hold it until closing tomorrow."
  • /wrap: "Yes, free gift wrapping on anything bought in the shop. Allow 5 minutes, or longer on busy Saturdays."
  • /delivery: "We don't post orders yet, but we can take payment over the phone and hold it for collection."

Note the stock reply. Without a live connection to your stock system, no AI or quick reply should ever say "yes, we've got it". WhatsApp Business app versus the API explains what changes if you want AI to reply automatically rather than just draft.

Suppliers, rotas and the admin nobody sees

A lot of a shopkeeper's week is admin that has nothing to do with a website: supplier emails, reorders, chasing credit notes, working out next month's rota around school holidays. AI is good at the writing and summarising parts:

  • drafting a firm but polite email about a short delivery, from your notes;
  • summarising a new supplier's terms (minimum order, carriage-paid threshold, returns) into five lines you can compare;
  • turning a list of staff availability into a draft rota you then adjust;
  • writing the monthly email to your loyalty list, if you keep one, from bullet points about new stock and events.

The short-delivery email shows how little you need to give the AI. Notes typed on the phone at the stockroom door: "order 4471, 24 mugs ordered, 18 came, 2 chipped, invoice charges 24, want credit for 8, 3rd time this year". An illustrative draft back: "Dear [supplier], I'm writing about order 4471. We ordered 24 mugs but received 18, and two of those arrived chipped. The invoice charges for 24. Please could you issue a credit note for 8 mugs, or send the six missing plus two replacements, by the end of next week? This is the third short delivery this year, so I'd also welcome a quick call about how orders are checked before dispatch." What to fix: nothing in the facts, but decide whether you want a credit or replacements before sending, because "or" invites the supplier to pick the cheaper one for them. Attach a photo of the chipped mugs.

Summarising new supplier terms is the job where a table beats a paragraph. An illustrative five-line comparison of two candle suppliers, built from their terms PDFs:

TermSupplier ASupplier B
Minimum first order$150$250
Minimum repeat order$100$150
Carriage paid over$300 (else $15)$250 (else $12)
Payment termsCard on order30 days after invoice
Returns of unsold stockNoneSwap up to 10% after 60 days

At a typical $180 reorder, neither supplier is carriage-paid: A charges $15 an order and B $12, so over six orders a year that's $90 against $72. The bigger differences are in the other rows. B gives you 30 days to pay and lets you swap slow stock, while A takes card on order and takes nothing back. And rounding a B order up to $250 makes carriage free. Check the table against the PDFs before deciding, because summaries sometimes carry a figure from one supplier into the other's column.

Rotas are where AI drafts need the closest check. Give a chat assistant each person's availability and the opening hours, and it will produce a tidy week. In a realistic first attempt, it put two people on Monday, the day the shop is closed, because Monday appeared in one person's availability list, and gave the Saturday assistant a Friday shift she'd marked as college. Tell it the closed days explicitly, list hard "never" times separately from preferences, and read the result against everyone's notes before you share it.

For promotions specifically, whether AI can help set prices and promotions covers what the till data can and can't support.

A first week with AI in a card and gift shop

Review replies from that week show the one edit that matters. A customer writes: "Lovely little shop, found a perfect present for my sister's 40th and they wrapped it beautifully." The AI's draft: "Thank you so much for your kind words! We're delighted you found the perfect gift. We look forward to welcoming you back soon!" Nothing wrong with it, and it could sit under any review of any shop. The edited reply adds one detail from the review and one from the shop: "Thank you! Forties are our favourite birthdays to wrap for. Hope your sister loved it, and the new autumn cards are in if you're back for the next one." Thirty seconds of editing, and it reads as a person who was there.

Here's an illustrative first week for a two-person card and gift shop with no website, a Square-style till and a Google profile last updated three years ago. Total cost: $0 to $20 a month, depending on whether the owner uses a paid chat assistant plan.

  1. Monday evening, 45 minutes: export 12 months of sales and run the till prompt above. Note the three busiest days and the ten slowest lines.
  2. Tuesday, 30 minutes: update every Google profile field with AI-drafted text, correct it, add 15 new photos taken on a phone. Check holiday hours.
  3. Wednesday, 20 minutes: reply to the 11 unanswered reviews with AI drafts, each edited to mention something specific.
  4. Thursday, 30 minutes: write 20 shelf talkers for the best-selling lines and 5 for the slow lines worth one last push.
  5. Friday, 20 minutes: set up WhatsApp Business quick replies and a greeting message.
  6. Saturday: ask the Saturday assistant what customers asked that she couldn't answer. That list becomes next week's crib sheet.

About two and a half hours in total. The measurable results come over the following months: more profile views and direction requests in the Google profile's performance figures, fewer "are you open?" messages, and a decision on the slow lines at the next buying trip.

Write the starting figures down on day one so you have something to compare against. An illustrative three-month check for the gift shop: direction requests up from 41 a month to 63; calls from the profile up from 18 to 22; "are you open?" messages down from about nine a week to three, since the hours and holiday times are now right; and four of the ten slow lines cleared through a front-of-shop table with new shelf cards. None of those numbers proves AI did it on its own, but together they tell you the week was worth repeating each season.

Where AI oversteps in a physical shop

  • Writing reviews or "customer" posts. Never. Fake reviews break platform rules and consumer law in many places.
  • Promising stock or delivery dates it can't see. Always "I'll check and come back to you".
  • Setting prices automatically. Use AI to show you margins and slow sellers; keep pricing decisions yours.
  • Loyalty list data. Don't paste customer names and emails into a chat assistant to "personalise" a newsletter; write one good email and send it through your email tool.
  • Replacing the conversation at the counter. The advice and recognition a customer gets in person is the reason they don't buy online. AI should free up time for more of it, not less.

If you later decide to sell online, the same groundwork (clean product names, good descriptions, accurate stock) is most of the job. Until then, a shop with a tidy till, a complete Google profile and good words on the shelves is using AI well.

Shopkeepers also ask

Do I need a website before I bother with AI?

No. A complete Google Business Profile, good photos and replies to reviews do more for most small shops than a basic website, and AI can help with all three. A website becomes worth it when you want to take orders or appear for searches about specific products. If you do go online later, AI makes product listings much faster.

Is it safe to put my till data into a chat assistant?

Sales totals by product and date are usually fine on a plan that doesn't train on your data. Remove anything that identifies customers, such as names, emails, phone numbers or loyalty IDs, before uploading. Business plans from the main providers don't use your content for training by default, and consumer plans have a setting to switch it off.

Can AI tell customers what's in stock?

Only if it can see your stock levels. Without a connection to your till or stock system, an AI replying to messages should never confirm stock; it should say you'll check and reply, or invite the customer to call. Google's Local Inventory app can show in-store products on your profile if your till is compatible and your products have manufacturer barcodes.

Further reads

Sources: Square press releases on Square AI; Google Business Profile Help pages on showing in-store products, answering inventory questions and business description guidelines; reporting on the retirement of Business Profile Q&A.

Want AI working in your shop, website or not?

On a 1:1 call we'll look at your till, your Google profile and where your time goes each week, and pick the first two jobs worth handing to AI.

Book a 1:1 call with me