Kiosks AI for Cafés and Takeaways: Self-Service Ordering Costs

A single self-service kiosk for a café or takeaway costs about $1,999 for the hardware plus roughly $99 a month in software on Square's list prices, so around $3,200 in the first year. It is worth having when queues at peak cost you sales and your menu is simple. Where orders are complex or the line is short, a counter order still wins.

The price tag is the easy part. What decides the outcome is menu set-up, how many orders a kiosk can really take in your layout, and whether the "AI" part, which mostly means suggested add-ons and voice ordering, pays for itself. The sections below price the options, do a break-even sum, and list the cases where a person is still faster.

Follow me on Instagram@sagnikteaches

What a kiosk costs: the list-price build-up

I used Square's own kiosk and restaurant pricing pages as a reference because they publish figures. Other vendors often quote on request, so use these as a yardstick, not the only price. All figures below are USD list prices checked in October 2026.

Connect on LinkedInSagnik Bhattacharya

The hardware page lists a 21.5-inch display, three directional microphones with noise filtering, a proximity sensor that wakes the screen, Wi-Fi or wired Ethernet, tap, dip and PIN payments built in, and mount options for floor, counter, wall or VESA (a standard screen-mount pattern). It carries a 30-day return window and a two-year limited warranty. Square also notes it is English only, and available only in some markets, so check availability for yours.

Subscribe on YouTube@sagnikteaches

Year-one cost for one kiosk

• Hardware: $1,999 (or about $2,328 over 24 months through financing: $97 times 24).
• Software: $99 times 12 is $1,188.
• Kitchen display if you do not have one: $30 times 12 is $360, plus its hardware.
• The Square Plus plan if you are on Square Free today: $49 times 12 is $588. Skip this line if you already pay for Plus or Premium.
• Menu set-up time: 10 to 20 hours of your own time, which I value at $0 if you do it yourself and $300 to $600 if you pay someone.
• Total, for a café already on Square Plus and without a kitchen screen: about $3,200. Add about $1,000 for a screen and set-up help, and about $600 more if you have to move up from the free plan.

Processing fees are not a new cost if the customer would have paid by card at the till anyway, so do not count them against the kiosk. Count only the extra hardware and software.

The cheaper route: a tablet and Kiosk Lite

Kiosk Lite on a tablet you already own is $360 to $600 a year in software. The trade-offs are a weaker microphone setup, no proximity wake, and a need to secure the tablet. For a café with a low-traffic counter wall, it is a sensible way to test demand before buying the big screen. Many owners also find a small tablet at the end of the counter is enough, and that a floor-standing kiosk only suits a space with a clear queue line.

The break-even sum with your own numbers

Take an illustrative takeaway doing 90 orders in the busy lunch hour from 12 to 1pm, with an average ticket of $14. Two staff work the till, and the line reaches the door. Staff estimate that on a typical week, five or six people a day walk away without ordering.

1. Lost sales: 5 walk-aways a day times $14 is $70 a day. Over 26 trading days that is $1,820 a month.
2. Add-on effect: suppose the kiosk's suggested add-ons lift the average ticket on kiosk orders by 50 cents. If it handles 40 orders a day, that is $20 a day, or $520 a month. This is an assumption, not a promise: Square says its suggestions use sales data and the guest's cart, but I did not find an independent figure.
3. Gross extra: about $2,340 a month.
4. Margin: at a 65% gross margin on food and drink, the extra contribution is about $1,520 a month.
5. Cost: $99 software plus about $97 financing is $196 a month.

On those assumptions, the kiosk pays back in about two months and looks easy. Now halve the walk-aways to 2 a day and set the add-on lift to zero. Extra gross is $730, contribution is about $475 and the cost is $196. Still positive, but the margin for error is small. The kiosk is a bad idea only when the line is short, so test your own numbers: stand at the door during the peak for three lunches and count the people who look at the queue and leave.

The unknown in this sum is whether the kiosk adds orders or just moves them from the till. If you simply shift 40 orders from the till to the kiosk and no extra people buy, you save time but no money. That can still be worth it, because the staff you save at the till can make drinks faster. But the value is then in throughput, not in revenue, and you should price it that way.

Menu set-up: the real work

Square says menu, prices and modifiers transfer from your Square account, and price, sold-out and menu changes apply to every kiosk. That is a strength if your Square menu is in good shape and a trap if it is not. A till menu built for staff who know it by heart is not the same as a menu a stranger can use.

What to fix before you switch on

• Item names. "Flat white" is clear, but "Brekkie 2" is not. Rename for customers.
• Photos. A kiosk without photos is slow. Take plain, well-lit pictures of the top 20 items.
• Modifiers. Milk choices, sizes, extra shots and sauces should be in a sensible order, with the default pre-selected.
• Allergens. Add a clear route to allergen information, and a prompt to speak to staff for serious allergies.
• Categories. No more than six or seven at the top level, with best sellers first.

Using AI to prepare the menu

A chat assistant is good at the dull parts of this. Paste your item list (names only, no prices) and ask:

Here is our current till menu. Rewrite each item name so a first-time customer understands it.
Keep names under 30 characters. Suggest one 12-word description for each of the top 10 sellers.
Do not invent ingredients. List any items where you are unsure what they are, so I can check.

An illustrative result: "Brekkie 2" becomes "Bacon and Egg Roll" only if you told it what is in it, otherwise it should list it as unsure. What to fix: models like to invent flavour words ("hand-crafted", "artisan") and ingredients. Delete both. For ideas on how to use your sales data to shape the list, see the tutorial on planning a seasonal drinks menu from sales data.

What the "AI" in a kiosk really is

For most café kiosks it means two things. The first is suggestion: Square describes upsell prompts drawn from your best sellers by time of day and from what is already in the cart. The second is conversation: guests can order by voice or touch, ask menu questions and filter items, for example asking for kid-friendly options. Neither needs a giant system, and neither replaces the menu work above.

Where it can fail:

• Noisy rooms. Voice ordering depends on microphones in a loud café. Test it during the busiest hour, not at 3pm.
• Accents and languages. Square lists English only. If many of your customers order in another language, a touch screen with translated labels may beat voice.
• Allergy questions. A kiosk answering menu questions must not be treated as an allergen authority. Keep a human route and wording.
• Odd upsells. A suggestion of a second coffee at 4pm is annoying. Review the prompts weekly and switch off the ones that irritate.

Throughput: how many orders can one kiosk take?

The break-even sum assumes the kiosk takes 40 orders in the busy hour. Is that realistic? Time it. A confident guest orders a coffee and a pastry in about 45 to 60 seconds on a clear screen, and a first-timer takes two minutes. If your average is 75 seconds, one kiosk can in theory handle 48 orders an hour, but only if someone is always standing at it. In practice, the kiosk is empty while guests queue for the till out of habit, then busy in bursts. Two things move the numbers: signage that points people at the screen, and a staff member who says "you can order on the screen, it is quicker" during the first fortnight. Without that nudge, adoption in a small café often stalls at a fraction of the orders, and the extra lane sits idle.

A simple measure: after the first fortnight, take the share of orders placed on the kiosk during the 12 to 1pm hour. If it is under 20%, the problem is placement or habit, not the machine. Move it to where the queue forms, not where it looks tidy.

Three businesses, three answers

A 30-seat café with a lunch rush

Illustrative: 70 orders in the peak hour, a mixed menu of drinks, sandwiches and cakes, and half the guests are regulars. A kiosk at the end of the counter for sandwiches and drinks takes pressure off the till, and the regulars keep chatting at the counter. Verdict: a Kiosk Lite tablet first, a full kiosk if adoption passes 30% of peak orders.

A takeaway with 120 orders on a Friday night

Mostly repeat orders, long modifier lists and a phone that rings constantly. A kiosk handles the walk-in line while a phone AI deals with calls. The menu needs the most work here, because a wrong modifier means a remake. Verdict: a kiosk plus tidy menu is worth a pilot, and the kitchen screen is not optional.

A coffee cart with a short line

Twenty orders an hour, one person serving, a tiny footprint. A kiosk adds cost and takes space, and the owner's chat is the product. Verdict: do not buy. If speed matters, a tablet for pre-orders on your own site is cheaper.

A mistake that shows up in week one

Here is an illustrative launch that goes wrong. A café puts its kiosk live with the full till menu, including a "Custom item" button staff used for odd charges. Guests tap it, type nothing sensible and walk away confused. Orders with an empty modifier go to the kitchen with no clue what to make. The fix is simple: remove staff-only items from the kiosk menu, and make milk choice a required step so a ticket never leaves without it. When you test, give the kiosk to someone who has never used the till and watch where their finger hesitates. Every hesitation is a menu fix, and fixing five of them often matters more than any suggested add-on.

Using AI to review the first fortnight

After two weeks, export sales by hour and by item for kiosk and counter orders separately, with no customer details. Then ask a chat assistant to compare them. Prompt:

Below are two lists: items ordered at the counter and items ordered on the kiosk during the same two weeks.
Compare average items per order, the most common add-ons on each, and which menu items are rarely ordered on the kiosk.
Suggest three menu or screen changes to test. Do not invent numbers; say if the data is too thin.

A typical, illustrative answer: kiosk orders average 2.3 items against 1.9 at the counter, cake is rarely ordered on the kiosk because it sits on the third screen, and an extra-shot prompt is accepted often at the counter but rarely on the kiosk. What to fix: treat it as a list of things to test, not as proof. Two weeks is thin, and a rainy week skews everything. Run one change at a time, such as moving cake to the first screen, and compare again.

Hidden costs and how to trim them

Beyond the table above, a few costs surprise owners in the first year. Cleaning and wear are real: a touch screen at a food counter is wiped many times a day, and fingerprints on a dark screen look grubby within the hour. Budget for cloths and a daily wipe in the opening checklist. Connectivity matters too, because a kiosk that drops its Wi-Fi mid-order loses the sale, so a wired Ethernet connection (which Square lists as an option) is worth the cable run. Then there is staff time: someone has to refresh photos, mark items sold out and check the prompts, which is perhaps an hour a week.

To trim the cost, start with one kiosk, not two. Use the software only in your busy months if your vendor lets you pause it, and ask whether a seasonal pause is allowed before you sign, since terms differ. Negotiate the stand: a cheaper counter mount can save money against the quoted $1,999 build, which varies with stand type. And sell the old tablet: if you move to a dedicated unit, the lite route's tablet can become a pre-order screen or a second till.

When the counter still wins

Here are five cases where I would keep ordering at the counter, or at least not rush to buy.

1. A short line. If your peak queue is three people, a kiosk adds a lane nobody needs. A regular smiling at the counter also sells more than a screen does.
2. Complex or custom orders. A catering order, a half-and-half pizza or a multi-allergy request is quicker to say than to tap.
3. Regulars who like talking. A neighbourhood café runs on the chat. If the line is part of the appeal, keep it.
4. Older or less confident customers. Some will hover at a screen. Keep a counter for them and a member of staff who can step out to help.
5. Tiny premises. A kiosk needs floor space and a queue path. If the screen blocks the door, it costs more than it earns.

A takeaway with a long phone queue may be better served by an AI phone line, while one losing margin to delivery apps should read what a takeaway actually pays on each order channel before buying any hardware.

A three-week pilot that costs almost nothing

You do not have to buy the kiosk to learn whether it will work. Run this first.

1. Week 1, baseline. Count orders by hour, average ticket, and walk-aways at peak. A tally sheet by the door is enough.
2. Week 2, tablet test. If your vendor offers a lite version, put a tablet at the end of the counter with a sign. Record how many guests use it and how many ask for help.
3. Week 3, decide. Compare queue length, average ticket and staff time at peak. If the tablet took 20 or more orders a day and queues shrank, a dedicated kiosk is justified. If almost nobody used it, you have saved yourself $1,999.

A short checklist before signing

• Is the vendor's software fee per device per month, and does it stay the same if you add a second kiosk?
• What happens to the contract if you cancel: do you keep the hardware, and can you use it with another system?
• Does the kitchen get orders on a screen or printer, and who pays for that?
• Can you switch off voice ordering or AI suggestions without losing the kiosk?
• How are refunds and mis-tapped orders handled, and who has the override at the counter?

If you run a takeaway and already use an EPOS, start with whatever AI features it ships and compare them with the EPOS comparison for takeaways. A kiosk is rarely the first thing to buy. It is the thing to buy once the till, the menu and the kitchen flow are in order, and the queue is still the problem.

Kiosk questions from owners

How much does a self-service kiosk cost for a café?

On Square's list prices, one kiosk is $1,999 for the hardware with a counter stand and $99 a month in software per device, on top of a Square Plus ($49 a month) or Premium ($149) plan. A tablet-based Kiosk Lite is $50 or $30 a month per device, depending on the plan. Check the vendor page for your market.

Does a kiosk replace staff?

Rarely. It moves the order-taking from the till to the screen, so the person at the counter can make drinks or food instead. Most small cafés keep the same headcount and serve more at peak, which is the benefit to measure.

Will an AI kiosk really increase the average order?

Vendors say suggestions based on your sales data and the guest's basket lift add-ons, but I found no independent figure to rely on. Measure your own average order for two weeks before and after, and compare like for like days.

Can customers still order at the counter?

Yes, and you should keep that option. Some customers prefer to talk, have complex orders or need help with allergies. A kiosk is an extra lane, not a replacement for the counter.

Further reads

• AI Customer Service Kiosk for a Shop Counter: Worth the Cost? — The same question for a retail shop counter.
• AI for Takeaways: What Is Worth Automating and What Is Not — What else in a takeaway is worth automating first.
• Takeaway Delivery App Menus: Rewrite Descriptions and Photos With AI — Menu descriptions and photos, the same work a kiosk needs.
• How Cafés Use AI to Fill Quiet Hours With Targeted Offers — Fill quiet hours, where a kiosk does little.
• AI Receptionist Kiosk or Phone AI: What a Walk-In Business Needs — Kiosk or phone AI for a walk-in business.
• AI Chatbot Online for Hospitality: Questions a Café Can Safely Automate — Which customer questions a café can safely hand to AI.
• AI Drive-Thru Ordering: What a Fast-Food Franchisee Controls — Who really decides on AI drive-thru ordering in a franchise: three models, the clauses to read, questions to put in writing and a lane-level sum.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Square's Kiosk hardware page and Square for Restaurants pricing page, re-checked 27 October 2026.