Grocery store WhatsApp ordering, handled by an AI that knows your shop

It's twenty past six. The till queue is five deep, someone needs the back door opening for the bread delivery, and your phone has buzzed eleven times since five o'clock. Three of those buzzes are regulars asking for a delivery tomorrow. One wants to know if you've got fresh coriander. One is a voice note you can't play at the counter. Grocery store WhatsApp ordering is meant to make life easier, and it does for your customers. For whoever is holding the phone, it often just means a second queue.

This page explains how Klugon's AI assistant fits into that second queue. It answers your shop's WhatsApp, works out what the customer wants, checks it against the rules you've given it (delivery area, minimum order, slots, what you do and don't stock) and writes it up as a clean order. It then hands that order to your team. A person in your shop checks the basket, confirms it with the customer and does the picking and delivery. The AI doesn't take payment in the chat and doesn't promise anything you haven't told it to.

That last part matters, so here's the honest version up front. Klugon is not a grocery e-commerce platform with live stock, substitutions logic and card payments. It is an assistant that takes the typing off your hands and leaves the decisions with you. For most independent grocers, convenience stores and specialist food shops, that's the part that eats the hours.

What grocery store WhatsApp ordering looks like in a real shop

Most independent grocers didn't choose WhatsApp as an ordering channel. Customers chose it for them. A regular asks once whether you could drop a few bits round, you say yes, and a year later a good share of your weekly trade arrives as messages.

Those messages come in a handful of shapes, and they repeat:

• The list order. "Can I get 2 pints semi-skimmed, a large Hovis, 6 eggs, bananas, 2 tins of chopped tomatoes and whatever chicken you've got." Usually typed, sometimes a photo of a handwritten list, sometimes a voice note. More on those last two below, because the AI reads text only.
• The stock question. "Have you got okra in?" "Is the halal lamb back?" "Do you do gluten-free bread?" These often turn into orders if someone answers within a few minutes, and turn into a trip to the supermarket if nobody does.
• The logistics question. "Do you deliver to Hillsborough?" "What time's the last delivery?" "Is there a minimum?" "Can I collect instead?"
• The repeat. "Same as last week please." This is the one that sounds easiest and is hardest to do from memory at the till.
• The change. "Can you add a bag of ice to my order?" "Swap the semi for whole." "Actually I'll pick it up."
• The complaint or query. A missing item, a bruised box of tomatoes, a charge they don't recognise.

When the messages arrive

Independent food shops tend to have a pattern that looks nothing like an office. There's an early burst before people leave for work, when the commuter crowd checks whether you're open and whether the papers have arrived. There's a second wave in the late afternoon, between school pick-up and tea, when people realise what's missing from the fridge. Then a long evening tail of tomorrow's orders, typed on the sofa at nine or ten at night, when your shop is either shut or down to one person.

Weekends are different again. Saturday mornings bring the big family shops. Sunday trading hours in England and Wales limit large stores, which is exactly when the corner shop's phone gets busy.

The calendar that drives your volume

Grocery demand follows the year closely, and the message volume follows it:

• Christmas week is the big one: turkeys and joints ordered weeks ahead, veg boxes, party food, and the 23 December panic for cranberry sauce.
• Ramadan and Eid change everything for shops serving Muslim communities: dates, fruit, fresh meat and the iftar rush, with messages arriving late into the night.
• Diwali brings orders for sweets, ghee, flour and gift boxes.
• Bank holiday weekends and barbecue weather mean charcoal, burgers, ice and drinks, often ordered the morning of.
• Back to school and the first cold snap each bring their own runs on lunchbox items and soup ingredients.
• Snow, storms and illness push older and housebound customers to ask for delivery when they'd normally walk in.

A human can cope with any one of these. The problem is that the busiest days for messages are also the busiest days on the shop floor.

What the AI does with each message

When a customer messages your WhatsApp Business number, Klugon's assistant reads the message and replies using what you've taught it. It works from three things: your product list, your knowledge base and your instructions.

It answers from your product list

You can add products to Klugon with names, prices, descriptions and photos. The AI can search them when a customer asks for something. For a grocer, that means it can say "yes, we stock Warburtons Toastie, £1.65" rather than "I'll check". It can also suggest related products when it makes sense, such as asking whether someone ordering a joint for Sunday wants potatoes and a bag of carrots too.

A caveat for big ranges. Products are added in the dashboard one at a time. There is no import from your EPOS and no stock feed. A full corner shop range runs into thousands of lines, so most grocers will not want to type it all in. The practical approach is:

• Add the lines people actually ask about by name: your fresh meat cuts, specialist and world-food items, bakery, and anything seasonal.
• Upload a price list as a CSV or PDF to the knowledge base, so the AI can answer "do you have…" for the rest.
• Tell the AI that availability is confirmed by staff when the order is picked, so it never promises a specific item is on the shelf.

It applies your delivery and collection rules

This is where most of the back-and-forth sits, so it's worth being precise in your instructions. You write the rules once; the AI applies them every time. For example:

• Delivery postcodes or areas you cover, and any you don't.
• Minimum order value for delivery, and the delivery charge.
• Delivery days and slots ("Tuesday, Thursday and Saturday, 10 am to 1 pm or 4 pm to 7 pm").
• The cut-off for next-day delivery ("orders received after 8 pm go out the day after tomorrow").
• Collection hours, and how long after confirmation an order is usually ready.
• What you don't deliver: hot food, alcohol, tobacco, or anything else you've decided to keep in-store only.
• Whether you'll substitute, and how ("we'll message you if we're out of something; if you'd rather we don't substitute, say so").

It writes the order up so staff can act on it

Once the customer has said what they want, the AI repeats the basket back in a tidy list, checks the delivery address or collection time, and asks if anything's missing. Then it saves the customer as a lead with their name and number, and flags the conversation for your team.

Flagging a conversation does three things in Klugon: it marks the thread "escalated" in your inbox with a "Needs a person" banner, logs the handover, and sends a notification to Slack or through Zapier if you've connected either. From Zapier you can push the order wherever your team already works, such as a Google Sheet that acts as the picking list.

It does not send the order to your EPOS, reserve stock or put anything in a delivery system. A person picks it up from there.

Voice notes and photos of lists

Here's a limit worth knowing before you go live. The AI reads typed messages. When a customer sends a voice note or a photo of a handwritten list, the AI receives a placeholder saying a voice message or image arrived, not the contents. It can't listen to the note or read the photo.

That's a real gap for grocers, because older customers love voice notes and busy parents love photographing the fridge list. The fix is simple and costs one message: tell the AI to reply "sorry, I can only read typed messages, could you type your list for me?" In testing your own setup, send it a voice note and check that it does exactly that. Your staff can't play the note from Klugon either, so if a regular can't or won't type, the AI should ask them to ring the shop instead.

It answers the questions that aren't orders

A good share of grocery messages are questions with fixed answers. The AI handles these without bothering anyone:

• Opening hours, including bank holidays and Christmas.
• Whether you do delivery, where, when and for how much.
• Whether you sell a certain product or range, from your product list and knowledge base.
• Whether you take cards, and whether there's a cashback limit.
• Parking, the bus stop, step-free access.
• Whether you sell stamps, top-ups, lottery tickets or parcel collection services, if you list them.

The Insights page in Klugon shows the most-asked questions over the last 30 days, and a list of "questions without a good answer". For a grocer, that second list is useful as a stock signal. If ten people asked about paneer this month and the AI had nothing to say, that's a product you might want on the shelf.

Grocery store WhatsApp ordering: what stays with your team

Klugon's AI is good at writing orders up. It shouldn't be the last word on any of the following, and you should tell it so in its instructions.

Confirming the order

The AI collects the order. Your team confirms it. That means a person reads the flagged conversation, checks the shelves, and replies in the Klugon inbox to say "all in stock, delivering Thursday between 4 and 7, total £32.40". Owners can reply from the inbox on WhatsApp, Messenger, Instagram and Telegram.

Two practical points about that reply:

• Meta's 24-hour window. WhatsApp lets a business send ordinary replies only within 24 hours of the customer's last message, according to Meta's WhatsApp pricing documentation. Outside that window you need pre-approved template messages, which Klugon doesn't send. In practice, an order placed at 9 pm needs confirming by 9 pm the next day, or the customer needs to message again.
• The AI keeps replying. Handover doesn't pause the assistant. If the customer sends another message while your team is dealing with the order, the AI will answer it too. Tell the AI in its instructions to say "a member of the team will confirm your order shortly" once a basket is complete, so it doesn't start a second conversation about the same shop.

Payment

Klugon doesn't take payment for grocery orders in the chat, and we don't recommend setting it up for that. Most independent grocers already handle this one of three ways: pay on delivery by card machine, pay on collection, or a payment link the shop sends from its own card provider when confirming. Keep doing whichever you already do. The AI can tell customers which method you use.

Age-restricted products

Alcohol, tobacco, vapes, lottery, energy drinks at some shops and knives all come with rules, and selling alcohol in England and Wales requires a licence under the government's alcohol licensing guidance, which includes having an age verification policy. A chat can't verify anyone's age. If you deliver alcohol, keep the decision and the ID check with the driver and your licence holder, and tell the AI to note age-restricted items as "ID will be checked on delivery" rather than confirming them.

Allergen questions

This is the one to be most careful about. The Food Standards Agency's best-practice guidance on distance selling and pre-ordering says allergen information for food sold remotely should be available to the customer before the order is placed and again when it's delivered. That applies to things like your deli counter, home-made samosas or bakery items that aren't prepacked.

An AI paraphrasing your ingredients from memory is not a safe way to meet that. Our advice is:

• Give the AI a written allergen sheet for your non-prepacked lines, uploaded to the knowledge base, so it can point customers to it.
• Tell it never to say an item is "safe" for an allergy, and to flag any allergy question for a person.
• Keep labels or stickers with allergen information on the delivered goods, as you would anyway.

Substitutions

Every grocer has a view on substitutions, and every customer has a stronger one. Some people are happy with any brand of chopped tomatoes; others will send back the whole delivery because you swapped Hellmann's for own-brand. The AI can't see your shelves, so it can't make that call at picking time.

What it can do is ask the question while the order is being written up. Tell it to finish every basket with something like "If we're out of anything, would you like us to swap it for something similar, or leave it out?" and to note the answer in the order. Then your picker has the customer's preference in writing, rather than a guess. If a customer says "no swaps on the baby formula, ever", that sits in the conversation history for next time too.

For the items where a wrong swap really matters (baby milk, gluten-free lines, halal or kosher meat, anything someone has said is for a medical diet) tell the AI never to suggest an alternative itself, and leave the decision to staff.

Complaints and refunds

"Half my eggs were broken" needs a human who can decide whether to refund, replace or apologise. Tell the AI to apologise, collect the details and a photo if offered, and flag it.

Anything urgent or personal

Grocery shops that deliver to older customers sometimes become an informal welfare check. If a message reads like someone is unwell, confused or in trouble, the AI should not try to help on its own. Tell it to flag the conversation straight away and, if it sounds like an emergency, to tell the customer to call 999. Klugon can't call anyone and can't alert you by phone; your Slack or Zapier notification is the alert, so set that up to reach someone who's actually looking.

Sample conversations

These are samples written for this page. The shops are made up, the customers are made up, and nothing here is a real recording or real transcript.

A sample WhatsApp order, evening

The shop is Hollins Lane Grocers, an invented convenience store with a deli counter. Delivery is Tuesday, Thursday and Saturday to three postcodes, £3 delivery, minimum order £20.

What happened behind the scenes: the AI saved the customer as a lead, flagged the conversation, and the shop's Slack channel received a "handover requested" post. At 8.15 the next morning, the owner opened the conversation in the Klugon inbox and replied with the confirmation and total. The reply went out inside Meta's 24-hour window because the customer's last message was the night before.

A sample stock question that turns into a collection

The shop is Okafor's Afro-Caribbean Foods, an invented specialist grocer that doesn't deliver but does click and collect.

The AI suggested a related product because the owner had turned that on and listed plantain and peppers as related items. It didn't promise stock.

A sample website chat about delivery areas

Some customers find you through Google, land on your website and ask there before they ever message on WhatsApp. The Klugon chat widget handles that.

Note the request for a phone number. Your team can reply to WhatsApp conversations from the Klugon inbox, but not to website chat, so a website order needs a number the team can ring or message.

Setting up WhatsApp ordering for your grocery store

This is what setup looks like for an independent grocer. Allow an afternoon for the Klugon side and more for the Meta side if you haven't used it before.

Before you start

• Your WhatsApp Business number. Klugon connects to Meta's WhatsApp Cloud API, not the free WhatsApp Business app on your phone. You'll need a Meta app, a WhatsApp Business phone number ID and an access token. If your current shop number is registered on the WhatsApp Business app, moving it to the Cloud API is a Meta process; read Meta's documentation before you start, as it affects how you use that number on your phone.
• Your delivery rules, written down plainly: areas, minimums, charges, slots, cut-offs, collection hours.
• A product list of the lines people ask about by name, with prices.
• A price list or range list as a CSV or PDF for the knowledge base.
• Your allergen sheet for non-prepacked foods.
• Somewhere for alerts: a Slack channel your team actually watches, or a Zapier account to send flagged orders to a Google Sheet or another tool.

The checklist

1. Sign up and start the 7-day trial of the social agent (WhatsApp, Instagram, Messenger). A card is required and is charged when the trial ends unless you cancel.
2. Create an assistant. Name it after your shop and write its instructions: who you are, where you deliver, how ordering works, and the line "orders are confirmed by a member of staff".
3. Add your key products, with prices and photos where they help.
4. Upload your price list, allergen sheet and delivery terms to the knowledge base, or paste them in as text.
5. Add any links the AI should share, such as a Google Maps link to the shop.
6. Write the rules for handover: allergy questions, complaints, age-restricted items, anything urgent.
7. Connect WhatsApp in the integrations page with your phone number ID and access token. The WhatsApp number add-on is billed at $19 a month when connected.
8. Connect Slack or Zapier so flagged orders reach the person who confirms them.
9. Test it. Send it a dozen real orders from your own phone, including awkward ones: a list with no quantities, an address outside your area, an allergy question, a voice note.
10. Decide who replies from the inbox and when. Someone needs to confirm overnight orders the next morning, inside Meta's 24-hour window.

Optional extras worth considering:

• The chat widget on your website, pasted in as one script tag. If you're on Shopify, WordPress, Wix or Squarespace, you can add the widget by pasting that script tag; there's no plugin or product sync.
• A QR code. Every Klugon business gets a public page with a chat, and a printable QR code pointing to it. Stick it by the till or on your delivery bags.
• The phone line. If many customers still ring rather than message, the AI voice agent can answer calls too, with the same rules. See the section on costs below.

For more on the Meta side, our guide to setting up WhatsApp booking walks through the Cloud API steps in detail. The process is the same whether the messages are bookings or orders.

What it costs, worked through

Klugon's prices are in US dollars. The figures below are from our pricing as of October 2026.

The Klugon side

For a grocer taking orders mainly on WhatsApp:

• Social agent, Starter: $79 a month (or $63 a month billed yearly), which covers 300 conversations a month across WhatsApp, Instagram and Messenger. One conversation is one customer thread.
• WhatsApp number add-on: $19 a month.
• Social agent, Business: $249 a month for 1,500 conversations, and needed if you want Slack, Zapier or the calendar and CRM integrations. For an order flow that relies on Slack or Zapier alerts, this is the realistic tier.

So a grocer using WhatsApp with Slack alerts is looking at $268 a month on monthly billing, or $218 a month billed yearly.

Meta also charges for some WhatsApp messages. Its pricing documentation says that from 1 October 2026 each business phone number gets 1,000 free service messages a month, with charges after that, and that messages from customers to the business are never charged. Check Meta's page for current rates in your country, as these change.

A worked example

Here are the assumptions. Change them to match your shop.

• You get 25 order-related WhatsApp threads a day, six days a week, so about 650 a month.
• Each one takes a member of staff about 6 minutes to read, reply to, clarify and write down when done by hand, mostly in interrupted bursts at the till.
• That's 65 hours a month of staff time.
• You pay that time at £13 an hour including employer costs.

That's roughly £845 a month of staff time on writing orders. With the AI doing the reading, clarifying and writing up, a person still spends about 2 minutes per order confirming and checking. That's around 22 hours, or about £285. The difference is roughly £560 a month, against $268 for Klugon.

The numbers are not the whole story. The bigger cost for most grocers isn't wages, it's the orders that don't happen because the 9 pm message got answered at 11 the next morning, by which time the customer had done an online supermarket shop.

Compared with the alternatives

• A part-time order clerk. If you're thinking of paying someone 15 hours a week to manage WhatsApp, the cost at the assumption above is around £845 a month, and they'll still be off when the evening orders come in.
• A marketplace app. Delivery marketplaces bring new customers but usually charge commission per order. Klugon is a flat monthly price and the customer stays yours.
• Doing nothing. It's free, and the till queue gets longer.

Privacy and your customers' data

Customer names, addresses and phone numbers in your WhatsApp are personal data. A few points specific to grocers:

• Delivery addresses and notes sometimes include things like "leave with neighbour, I'm in hospital" or "door code 4471". Treat the inbox like a till drawer: only staff who need it should see it.
• Health details can creep in through dietary requests ("diabetic, no sugar please"). That's special category data under UK GDPR. You don't need to collect it, so tell the AI not to ask about health conditions and only record the product preference.
• Marketing. If you want to send offers, consent needs to be explicit. Klugon's lead capture records marketing consent only when a customer gives it. The ICO's direct marketing guidance covers what you can send and when.

Klugon is operated by Softaar Technologies UG, a German company, under the GDPR. Some processing happens with sub-processors in the US under standard contractual clauses, as listed in the privacy policy. You can export conversations, leads and customers as CSV and delete individual records from the dashboard. Full erasure of your account data is by email request.

How Klugon compares with other ways grocers take WhatsApp orders

Grocers usually end up with one of these setups. Each has a place.

If you've looked at button-based chatbot tools, our ManyChat alternative comparison explains the difference between flow builders and an assistant that reads what people actually type.

Klugon is a sensible fit for a shop that already gets orders by WhatsApp and wants them written up and answered faster, without changing how picking, payment and delivery work. It isn't the right tool if you want customers to pay online for a full basket with live stock, which needs a proper online shop.

Klugon also works across other kinds of shop in the AI assistant for retail hub. If you run a mixed business, the pages for florists, pharmacies and gift shops cover the specifics of each.

How to tell if it's working

After a month, look at these in Klugon's analytics and insights pages:

• Conversations by hour. If the busiest hours are evenings and Sunday mornings, that's trade you were previously answering late or not at all.
• The handover rate. "Handled by the assistant" shows how many conversations needed a person. For grocery orders, every order should be handed over to be confirmed, so the useful number is how many non-order questions the AI answered alone.
• Questions without a good answer. Go through this list weekly. Each one is either a gap in your knowledge base or a product you don't stock.
• Leads. Every new ordering customer is saved as a lead. The leads list gives you a phone book of delivery customers that you didn't have before.
• Your own till figures. Compare delivery and collection takings for the month before and after.

Frequently asked questions

Does the AI take payment for grocery orders on WhatsApp?

No. The AI collects the order and your team confirms it. Payment happens however you already handle it, usually by card or cash on delivery or collection, or a payment link your team sends from your own card provider.

Can it check what's in stock?

Not live. Klugon doesn't connect to your EPOS or stock system. It answers from the product list and price list you give it, and your team confirms availability when they pick the order. Tell the AI to say availability is confirmed by staff.

Can customers order with voice notes or a photo of their list?

Not through the AI. It reads typed text only. A voice note or photo reaches it as a placeholder ("[voice message]" or "[image]"), so it can't hear or see the list. Tell it to reply with something like "sorry, I can only read typed messages, could you type your list?" Most customers will.

Will the AI send an order confirmation?

The AI replies in the same chat to say the order has been passed to the team. The actual confirmation with total and delivery time comes from your staff, replying in the Klugon inbox. Klugon doesn't send automatic confirmations, reminders or texts.

What happens if a customer messages at midnight?

The AI answers straight away, writes up the order and flags it. Your team confirms it in the morning. Keep inside Meta's 24-hour window, or the customer will need to message again before you can reply.

Can I use my existing shop WhatsApp number?

You connect a WhatsApp Business number through Meta's Cloud API. If your number is on the WhatsApp Business app today, check Meta's documentation on moving it, because it changes how you use that number on your phone. Social Starter connects one WhatsApp number; Business connects up to 3 and Scale up to 10, at $19 a month each. All the numbers are answered by the same assistant setup.

Can it take orders by phone as well?

Yes, if you add the voice agent. It answers calls with the same instructions and every call appears in your inbox. It can't transfer a call to a person; it flags the conversation so your team can call the customer back.

How does it handle allergy questions?

Tell it to share your written allergen information and flag the conversation for a person, and never to say a product is safe for someone's allergy. The Food Standards Agency expects allergen information for remote sales to be available before ordering and on delivery.

Do I need a website?

No. WhatsApp ordering works without one. Every Klugon business also gets a public page with a chat and a printable QR code, which some shops put by the till or on their delivery bags.

Can I try it before connecting WhatsApp?

Yes. During the 7-day trial you can test the assistant in the browser and send it sample orders, then connect WhatsApp once you're happy with how it writes them up.

If your shop's phone has become a second till, it's worth seeing what Klugon's AI receptionist does with a week of your real orders. Plans and add-ons are listed on the pricing page, and the WhatsApp AI chatbot page covers the channel in more depth.