Square AI Kiosk Adds Conversation to Self-Service Ordering

Square launched a 21.5-inch Square AI kiosk that lets restaurant customers build an order by speaking, tapping, or moving between both methods. The conflict is clear. Square wants its machine to behave more like an attentive employee, but restaurants need consistent accuracy during their busiest and noisiest hours.

The new hardware can answer menu questions, suggest items, process modifications, and work through substitutions, according to Square. Presence detection greets approaching customers, while integrated contactless and chip payments complete the transaction without a separate reader.

This is more than a larger replacement for Square’s earlier iPad-based kiosk. Square is putting conversational AI directly inside its restaurant commerce system, where menus, orders, loyalty accounts, payments, printers, and kitchen displays already connect. Competitors such as SoundHound AI have pursued similar voice ordering across kiosks, phones, and drive-thrus. Square’s advantage is that it controls more of the transaction path.

That integration also raises the standard Square must meet. A fluent answer is not enough when a customer is asking about allergens, changing several modifiers, or ordering during a crowded lunch rush. The Square AI kiosk must translate conversation into the correct structured order every time.

The Square AI Kiosk Turns a Menu Screen Into an Ordering Agent

Square’s central change is not the larger screen. It is the decision to let conversation control a live restaurant order.

Square unveiled the standalone kiosk on October 14, 2026, as new hardware for quick-service restaurants. Its vertical touchscreen displays menus, item images, cart details, and prompts. A customer can still navigate those elements manually, just as with a conventional self-ordering terminal.

Voice creates a second path through the same order. Square says customers can ask for popular items, request recommendations, identify vegetarian choices, and search for food without particular ingredients. They can also request modifications or substitutions using ordinary language.

A customer might ask for a drink without dairy, remove onions from an item, or replace one side with another. The system must interpret that request, identify the applicable menu rules, and update the cart. The screen then gives the customer a visible record of what the software understood.

Square calls the underlying system an “agentic, conversational voice layer.” In this setting, agentic means the software does more than retrieve menu text. It takes actions inside the ordering workflow, such as selecting an item, applying a modifier, or suggesting an add-on.

The company’s kiosk announcement says the software can use contextual information when making suggestions. That context includes what is popular at a location, what sells at that time, and which items commonly appear together.

This makes the recommendation system operational rather than decorative. If someone orders a sandwich, the kiosk can suggest a frequently paired side. If a customer asks what is popular during breakfast, it can narrow the menu without forcing that person through several categories.

Customers can alternate between voice and touch during one session. That flexibility matters because speech is not always the fastest input. A customer might ask the system to find pescatarian dishes, tap one result, speak a modification, and then review the completed cart visually.

The device also supports check-in through Square Loyalty or a Neighborhoods profile. That connection can let customers earn or redeem rewards. It also gives Square a route toward more personalized ordering, although customers retain the option to order without connecting a profile.

Payments happen through built-in contactless and chip readers. An integrated scanner supports gift cards, with packaged-goods scanning listed as an upcoming capability. These components keep the order and payment sequence inside one piece of hardware.

Square’s setup documentation clarifies another important change. The new device is a standalone system and does not require the iPad used by Square’s earlier kiosk product. Restaurants can install it on a counter, floor stand, wall mount, or compatible VESA mount.

That distinction matters for larger operators. A purpose-built device can offer a more consistent deployment than a consumer tablet paired with several accessories. It also gives Square more control over microphones, speakers, payment hardware, display behavior, and future software updates.

The tradeoff is dependence on the connected system. Square says voice ordering and payments require a stable internet connection. Offline and cash payments are not supported, and the device is intended for indoor use. A restaurant therefore needs another checkout option when connectivity fails or a customer wants to pay with cash.

The first impression is a conversational screen. The deeper product is a tightly connected endpoint inside Square’s restaurant platform. That is what moves the announcement beyond a routine hardware refresh.

Why Square Is Bringing Voice Ordering Into the Restaurant

Square is targeting the moment when self-service stops saving labor because an employee must rescue the customer.

Traditional kiosks are efficient when customers know exactly what they want. They become less efficient when a menu is long, substitutions are complicated, or dietary questions require help. Each confused customer can pull a worker away from food preparation or counter service.

Square is trying to close that assistance gap without abandoning self-service. Its kiosk keeps the visual certainty of a touchscreen while adding a conversational layer for questions and changes. The company argues that this combination can keep lines moving and reduce pressure on front-of-house teams.

That claim fits the realities of quick-service operations. Demand arrives in concentrated periods, but restaurants cannot instantly add trained staff for a lunch rush. A kiosk can absorb additional orders, yet only if customers can complete those orders without intervention.

Voice ordering is particularly useful when the desired item is difficult to locate. Baltej Sra, CEO of Boba Nation Grand, described seeing a customer use voice to find a specific menu item quickly. The company tested the kiosk at some locations and plans to extend it across the rest of its stores, according to Square.

That example is narrow, but it shows the intended mechanism. The voice system is not simply replacing taps with spoken commands. It is supposed to reduce the search cost created by large menus.

The same mechanism can affect order value. A human cashier recognizes natural moments to suggest a topping, drink, or side. A basic kiosk typically presents fixed prompts. Square says its software can make suggestions using the customer’s current cart and local ordering patterns.

Restaurants should treat that as a company claim until broader deployment produces independent results. Square has not published controlled data showing how much the new kiosk changes completion time, intervention rates, average order value, or order accuracy.

The company has also not named major restaurant chains adopting the device. In an interview about the launch, Square described kiosks as a growth area but did not announce specific chain agreements. That leaves the initial market signal dependent on smaller test locations and future customer disclosures.

Still, Square enters with a meaningful structural advantage. Its platform can connect the kiosk to the menu source, point-of-sale system, kitchen displays, printers, customer profiles, and payment processing. An order does not need to cross several disconnected vendor systems before reaching the kitchen.

This integration is important because restaurant voice AI depends on current operational data. The agent needs to know which modifiers are permitted, which products are unavailable, and which fulfillment choices apply at that location. A natural conversation becomes dangerous if the system’s menu knowledge is stale.

Square’s dashboard lets operators assign menus and settings to particular locations. Restaurants can organize items, define modifiers, choose fulfillment methods, and control order notifications. The kiosk then converts customer input into the structure already used by the restaurant.

That reduces one class of integration work. It does not eliminate the harder challenge of interpreting speech. Background noise, accents, overlapping voices, unusual product names, and rapidly changing requests can still expose errors.

The business pressure therefore runs in both directions. Restaurant operators want more order capacity without creating another system for employees to monitor. Square wants to make its payments and restaurant software harder to replace by attaching an AI-assisted ordering surface to them.

If the kiosk performs well, Square gains a more visible role at the start of the transaction. If it performs poorly, the failure appears directly in front of the customer. The hardware places Square’s AI claims where mistakes are immediate and measurable.

Square AI Kiosk Versus the Touch-Only Ordering Model

The primary contest is between a guided conversation and the rigid decision tree that has defined most self-service ordering.

A conventional kiosk asks the customer to adapt to the menu’s architecture. The user selects a category, chooses an item, navigates modifier screens, responds to upsell prompts, and reviews the cart. The flow is predictable for the restaurant but can become tedious for the customer.

Square reverses that relationship. Its voice layer is supposed to interpret the customer’s goal and translate it into the menu’s structure. Someone can begin with a constraint, such as avoiding a particular ingredient, rather than knowing the category and item name in advance.

This difference becomes clearer with a complicated request. A customer might want a meal without dairy, prefer a smaller drink, reject one default topping, and ask whether another side is available. A touch-only interface can support all those actions, but each one may sit on a separate screen.

A conversational system can collect the same requirements through dialogue. It can ask a follow-up question when a request is incomplete, then present the result for confirmation. That sounds closer to counter service because the software maintains context across several turns.

The touchscreen remains essential. Voice interfaces can hide available choices, while visual menus make options and totals easy to compare. Square’s multimodal design, meaning an interface that combines speech and touch, lets each method cover the other’s weakness.

Presence detection starts that interaction before the first tap. The device can recognize that someone has approached and display a welcome prompt. It does not mean the kiosk identifies the person. Personalization requires an optional loyalty or profile connection.

The visual review step can also limit the damage from a misunderstood phrase. Customers can inspect the cart before paying and correct mistakes by voice or touch. That feedback loop is more suitable for commerce than a voice-only exchange.

However, the user experience will depend on response time. A menu agent that pauses after every request can feel slower than tapping, even when it eventually produces the right order. Square has not disclosed latency targets or independent accuracy benchmarks for the launch system.

It also has not explained which speech models power the product, whether processing occurs locally or remotely, or how long voice recordings and transcripts are retained. Those details matter to enterprise buyers evaluating privacy, reliability, and regulatory obligations.

The company instead emphasizes the product-level outcome. Thomas Templeton, Block’s hardware lead, told Fast Company that voice lets the kiosk handle more complicated questions and makes interaction feel more human.

That framing identifies the actual wager. Square is not asking restaurants to believe that speech recognition is new. It is arguing that current AI can connect open-ended language to structured menu actions reliably enough for front-line deployment.

The Square AI kiosk also differs from a general chatbot. Its useful operating space is deliberately constrained. It does not need to discuss any topic. It needs to understand one seller’s menu, allowed modifiers, availability rules, loyalty system, and transaction flow.

A narrower domain can improve reliability because the range of acceptable actions is limited. Yet restaurant menus contain their own complexity. Combo rules, time-specific items, regional products, allergies, promotions, and substitutions can create thousands of valid and invalid combinations.

The touchscreen model handles that complexity through explicit options. The conversational model must infer intent without inventing choices. Its success depends on respecting the menu’s rules, not merely producing a plausible response.

Square’s strongest design decision is keeping both interfaces active. Restaurants do not have to force every customer into voice ordering. Customers who value speed and familiarity can tap, while those who need assistance can speak.

That approach also softens adoption risk. A restaurant can introduce the AI capability without redesigning its entire ordering experience around speech. If voice struggles in a particular situation, the screen still provides a route to completion.

The real test is whether the system reduces total friction. Voice ordering adds value only when the time saved on discovery and customization exceeds the time spent listening, clarifying, and correcting. Square will need operational evidence, not just fluent demonstrations, to settle that question.

Voice AI Competitors Already Understand the Hard Parts

Square is entering an established restaurant automation race, not creating a new category by itself.

SoundHound AI already markets conversational ordering across phones, drive-thrus, and kiosks. Its OASYS restaurant platform combines voice interactions with menu and transaction systems. The company has also worked with point-of-sale vendors, including Square and Toast.

At the 2026 National Restaurant Association Show, SoundHound demonstrated a “Dynamic Kiosk” that combines voice, touch, and visual interaction. Its restaurant AI demo closely resembles the basic interaction Square now describes.

That comparison matters because the differentiator is not voice alone. The competition centers on deployment, restaurant integrations, accuracy, menu management, and the ability to recover when the software does not understand a customer.

SoundHound approaches the market as a specialized voice AI provider that connects with several commerce systems. Square approaches it as a commerce platform adding conversational intelligence to its own hardware and software.

The specialist can spread its technology across more restaurant channels and infrastructure providers. The platform owner can optimize the entire path from menu selection through payment and kitchen routing. Each route creates different strengths.

Toast also supports restaurant automation through its partner network. Its SoundHound integration focuses on phone ordering and sends completed orders into the Toast system. The partnership shows that restaurant operators may assemble voice ordering from several vendors rather than buy every component from one company.

Drive-thru deployments provide a useful warning. McDonald’s ended a test of automated ordering technology developed with IBM in 2024. The companies did not treat that outcome as the end of restaurant voice AI, but the decision showed how hard broad deployment can be.

The McDonald’s test covered more than 100 restaurants before the chain decided to remove the system. McDonald’s said it still saw an opportunity for voice ordering and planned to evaluate other approaches.

A drive-thru is harsher than an indoor kiosk. Engines, weather, poor microphones, and customers speaking from different positions complicate audio capture. A Square kiosk has more control over distance, hardware, and visual confirmation.

Even so, both environments share the central problem. The AI must map natural speech to a constrained menu while customers hesitate, change their minds, use informal names, or speak over companions. A system can perform well in a demonstration and still struggle across thousands of real interactions.

Restaurant operators also care about escalation. The important question is not whether the AI ever fails. It is whether the device detects uncertainty, asks the right question, and hands the customer to a person without losing the order.

Square has not provided detailed information about confidence thresholds or human handoff. Its support materials advise restaurants to test each menu and device configuration before accepting live orders. That instruction acknowledges how local setup affects performance.

The operational burden may shift rather than disappear. Employees could spend less time entering routine orders but more time maintaining modifiers, checking availability, reviewing exceptions, and helping customers when the agent reaches an edge case.

Accessibility creates another requirement. Voice can help customers who find layered touch menus difficult, but it cannot be the only usable mode. Speech disabilities, hearing loss, language availability, and privacy preferences make touch and staff assistance necessary.

The kiosk’s indoor placement also limits direct comparison with drive-thru systems. Square is initially positioning the product for the front counter and dining space. Competitors with phone and drive-thru products cover more ordering channels, while Square offers tighter control over the in-store transaction.

Square can still exert pressure because it already has relationships with merchants using its point-of-sale and payment products. Those sellers may prefer a native device over adding another AI vendor and managing another integration.

Conversely, large chains may resist dependence on one commerce platform. They often use specialized ordering, loyalty, payment, and kitchen systems across many markets. A voice provider designed to work across those systems may fit them better.

The announcement therefore creates two races. Square must prove conversational ordering works better than touch-only self-service. It must also show that its integrated approach can compete with specialists serving several restaurant technology stacks.

Menu Accuracy Is the Test That Marketing Cannot Skip

The Square AI kiosk succeeds only if it converts flexible speech into rigid operational truth.

A restaurant order is a set of commitments. The kitchen must receive the correct item, size, preparation, modifiers, and allergy-related instructions. The customer must see the right total, while inventory and reporting systems must record the transaction correctly.

Generative AI often works by producing likely language. Restaurant systems require deterministic actions, meaning the same validated request should result in a permitted menu operation. Connecting those two behaviors safely is the product’s core engineering challenge.

Square says the agent fully understands a seller’s menu. That wording should be treated as a company claim, not an independently established result. The company has not released benchmark results across accents, languages, menu sizes, or noisy dining environments.

Allergy questions deserve particular caution. The kiosk can surface menu information supplied by the restaurant, but that does not make the system a medical authority. Ingredients, cross-contact risks, preparation practices, and supplier changes can fall outside a digital menu description.

Restaurants will need clear boundaries around those answers. The kiosk should distinguish stored ingredient information from guarantees it cannot make. When the available data is incomplete, escalation to an employee is safer than a confident response.

Menu synchronization is another risk. A voice agent can only make valid recommendations when the underlying data reflects current availability. If an item sells out but remains active in the menu, the AI may guide customers toward something the kitchen cannot prepare.

Square’s integrated stack can reduce synchronization failures because the kiosk uses menus configured through Square Dashboard. Operators still need disciplined processes for updating items, modifiers, and location-specific availability.

Connectivity creates a separate failure point. Square’s documentation says a stable connection is required for voice ordering and payments. Offline transactions are not supported. Restaurants must plan for service interruptions, even if their local menu appears on the screen.

Privacy questions remain underexplained. Presence detection can welcome approaching customers, while optional account check-in adds loyalty data to the session. Square has not detailed voice retention, transcript access, model training practices, or the handling of incidental speech around the device.

Those questions do not prove a privacy problem exists. They identify information that buyers should request before deploying microphones in customer-facing spaces. Signage and clear controls can also help customers understand when voice features are active.

The physical environment will shape performance. Music, blenders, kitchen equipment, nearby conversations, and several kiosks operating together can create difficult audio conditions. Square controls the device microphones, but it does not control every restaurant layout.

Language support is another open question. Square’s setup guide lets operators configure device language, but the launch materials do not provide a detailed list of supported conversational languages. Multilingual performance can become decisive in restaurants serving diverse communities.

There is also a social question. Some customers will not want to speak their orders in a public space, especially when dietary preferences or personal information are involved. Others may prefer voice because it removes several layers of tapping.

The product does not need universal voice adoption to create value. It needs enough customers to complete orders faster, with fewer interventions and acceptable accuracy. The ability to switch between voice and touch makes that threshold more attainable.

Restaurants should measure the system through operational outcomes. Useful metrics include completed orders, correction frequency, employee interventions, abandoned sessions, order accuracy, payment completion, and average transaction time.

Average order value also matters, but it should not dominate the evaluation. Aggressive automated suggestions can increase the cart while frustrating customers. Repeat visits and satisfaction offer a better check on whether contextual upselling feels useful or intrusive.

Square’s launch materials emphasize speed, personalization, and staff relief. Those benefits remain plausible but unproven at scale. The first deployments must show that conversational assistance removes more work than it creates.

That evidence will determine whether the Square AI kiosk becomes a new default interface or an optional feature used in limited scenarios. The hardware is ready for restaurants. The reliability case is still being built in public.

What Comes Next for Square Voice Ordering

Three signals will show whether Square has built a durable ordering platform or an impressive demonstration.

The first signal is deployment beyond initial test locations. Named restaurant groups, multi-location rollouts, and repeat hardware orders would show that operators see enough value to expand after direct experience. A few polished pilots would provide much weaker evidence.

The most useful disclosures will include store counts and deployment schedules. Expansion within an existing customer matters more than a test announcement because it suggests the economics and support burden survived real use.

The second signal is measurable operational performance. Square or its customers should report order completion time, correction rates, staff interventions, abandoned sessions, and accuracy across different menu types. Independent studies would carry more weight than selected testimonials.

Performance should also be compared with touch-only kiosks at similar locations. Without that baseline, faster ordering or larger carts could reflect store traffic, menu design, or customer mix rather than conversational AI.

The third signal is how Square expands and governs the voice layer. Buyers should watch for clearer language support, privacy documentation, escalation controls, offline resilience, and tools for reviewing misunderstood requests.

Competitor responses will provide another clue. SoundHound and other restaurant AI vendors already offer voice across several channels. If they deepen kiosk integrations or compete through broader platform compatibility, Square’s closed-loop advantage will face a direct test.

Square has chosen a sensible starting point. Indoor kiosks offer controlled hardware, a visible confirmation screen, and direct access to restaurant data. Those conditions give conversational ordering a better chance than voice-only systems operating in unpredictable environments.

Yet the product’s ambition is larger than adding a microphone. Square wants its kiosk to understand intent, guide decisions, update a cart, make recommendations, identify a customer, and finish a payment. Every additional action increases both its usefulness and the consequences of error.

For restaurant buyers, the next step is practical evaluation. Test the Square AI kiosk against real menu complexity, peak-hour noise, unavailable items, uncommon accents, and repeated modifications. Then measure how often customers finish without staff assistance.

For everyone watching AI move into physical commerce, the central question is equally concrete: does conversation make self-service feel more like service, or does it place a friendly voice over the same old friction?