QSR Back of House

Yum Brands and Nvidia are working together to speed up Yum’s development of artificial intelligence, the companies announced Tuesday. Nvidia’s technology will support Yum’s proprietary tech system, Byte. Yum is the tech firm’s first restaurant partner.

The company piloted several AI tools at some Pizza Hut and Taco Bell locations earlier this year, and plans a broader rollout at about 500 restaurants across these brands, KFC and Habit Burger during the second quarter. 

Yum’s strategy in recent years has hinged on building up its proprietary tech systems through acquisitions and in-house development, with the aim of keeping ahead of its major QSR competitors.

The restaurant powerhouse said the technologies developed in partnership with Nvidia will focus on three areas: drive-thru and call-center voice AI, computer vision to analyze operations and AI analytics at the restaurant level. The partnership includes Nvidia’s software and may eventually yield “AI agents that plan, reason and act to assist across restaurants,” according to the press release.

The voice ordering tech, which uses Nvidia Riva and Nvidia NIM microservices, can handle complex menus and speech patterns, according to the press release. 

The computer vision tech includes a component of real-time labor surveillance meant to “optimize drive-thru efficiency and back-of-house labor management through real-time analytics and alerts,” the press release said. 

Yum did not respond immediately to a request for comment on the ethical issues raised by algorithmic and AI surveillance of workers.

The restaurant-level analytics, which Yum refers to as Accelerated Restaurant Intelligence, will be used to create action plans for restaurant managers at struggling locations, using best practices from stronger-performing units. 

By working with Nvidia, Yum aims to reduce the cost and increase the speed and quality of the tech tools available to franchisees. The technology also benefits from the scale of Yum’s 61,000 stores and the vast amounts of information generated by its transactions, according to the press release.

Joe Park, Yum’s chief digital and technology officer, said the program would “enable us to harness the rich consumer and operational data sets on our Byte by Yum! integrated platform.”

Tools built through the program will be proprietary to Yum, the press release said. Ultimately, Yum wants to integrate “technology into every touch point, across every restaurant, around the world.”

Last year, Yum announced it would deploy voice-AI at hundreds of Taco Bell drive-thrus. The chain also expanded the number of restaurants using its restaurant coaching mobile app to 20,000 Pizza Hut and KFC locations in more than 120 countries, according to Yum’s most recent earnings call. 

The company’s kitchen and delivery system, called Dragontail before it was folded into Byte, is live in 8,000 stores. Yum plans to expand it across its U.S. systems for Taco Bell, Pizza Hut and KFC in 2025. Likewise, the restaurant company expects to deploy back-of-house management tech across its Taco Bell U.S. store system this year, according to the earnings call.

Miso Robotics has developed a smaller, more efficient robotic fry station that takes less time to install than previous versions of the Flippy fry cook, according to a press release. White Castle began piloting the new Flippy late 2024.

The robotics company said that Jack in the Box and White Castle have scheduled additional Flippy installations early this year. As of 2025, Miso had only installed the robotic arm in 13 restaurants across the two brands, according to a document filed with the Securities and Exchange Commission.

Miso said new investments enabled the company to pilot its new models at some White Castles late last year. 

Miso’s new model of Flippy is faster and takes up about half the floor space of the previous prototype, according to the press release. Miso said its robot can be installed in a few hours in existing kitchens. This is about 75% less time than previous versions, reducing the disruption to operations necessitated by installation. 

Flippy’s more compact footprint allows the robot to fit more easily into existing kitchens, reducing potential remodel costs and avoiding some permitting problems, according to the press release.

The company claims that its fry cook, which costs $5,400 a month, is cheaper than hiring an additional worker; this is not necessarily accurate. Flippy’s hourly cost to operate depends on how many hours a day a restaurant uses the machine.

A restaurant that uses Flippy 12 hours a day is paying $14.51 an hour, slightly more than the median wage for fast food cooks, which is $14.31, according to the Bureau of Labor Statistics. The hourly cost of operation falls to $7.25 — the federal minimum wage — for a restaurant that uses the robot 24 hours a day. 

However, Miso claims Flippy can process about 100 baskets in an hour, which the company said is more than human employees generally process. The machine can cook various fried foods, including french fries, onion rings, chicken and tacos.

Dennis Lock, Miso’s head of sales,said in the press release that the technology is not necessarily about replacing individual workers. Flippy can lead to “cost reductions and revenue increases per month through labor redeployment to higher value tasks, faster speed of service, and reduced food waste,” Lock said. 

Recently, the company signed a new investment deal with EcoLab and a partnership with Nvidia, a tech company specializing in hardware for AI. The company also credited the development of the new Flippy to the leadership of Rich Hull, who joined the company as CEO in 2023.

In 2022, White Castle said it would deploy Flippy at 100 of its restaurants, and re-upped that commitment in October 2023, according to the SEC disclosure, despite the slow roll out of Flippy.

Sponsored content

By Sift

Digital food and delivery businesses experience continuous growth and change, and with that comes a side of surging fraud threatening profits and customer trust. The digital dining space is notoriously dynamic, with fraud that fluctuates by the season. Tracking those changes along with other fraud trends and data is a necessary part of effective risk management, providing insight into performance and allowing for proactive, predictive mitigation.

That’s an especially valuable edge for QSRs, which typically process transactions (and entire user journeys) in a matter of minutes or seconds. Sift’s Fraud Industry Benchmarking Resource (FIBR), an interactive tool enabling businesses to compare their own fraud rates against industry benchmarks, delivers exactly those insights, with options to compare payment fraud, account takeover, and chargeback rates by industry and region.

Rising Refund and Promotion Abuse

The food and delivery sector’s Q2 payment fraud rate stands at 3.1% in FIBR, with fraudulent chargebacks at 0.048%. Both numbers are trending down compared to Q1 of this year, but represent a typical valley in QSR fraud: once the elevated fraud surrounding start-of-year promotions and account creation subsides, and security teams have recalibrated, fraud tends to slow down. At least, until the summer months, when new promotions and releases go live.

A significant portion of fraud in the food delivery space stems from refund and promotion abuse: MRC’s 2025 Global eCommerce Report found that first-party fraud, where customers claim issues like missing or poor quality items to receive unwarranted refunds, sits in the top five threats experienced by 90% of providers. In fact, panelists in the 'Global Payments & Fraud Survey' session at MRC Vegas 2025 speculated that the global first-party fraud rate is very high, between 60-80% of total fraudulent chargeback disputes. In 2024 alone, these types of disputes cost the industry an estimated $103 billion.

Similarly, chargebacks present a dual problem for food and delivery merchants: they result in direct financial losses for the business, and they damage relationships with payment processors. The average chargeback rate for food delivery and quick-service restaurants (QSRs) is at 0.048%, which is just shy of the industry average—but providers have to be wary of ballooning disputes, since exceeding certain thresholds leads to increased fees and stricter scrutiny from card networks.

Account takeover and fake account creation further strain the system. Anyone can easily create multiple accounts to exploit new user discounts, and in many cases, a single bad actor can be responsible for significant loss. Fraudsters have been known to leverage hundreds of fake accounts to exploit new user promotions, return hundreds of dollars in stolen deliveries for a healthy profit, and consume thousands of dollars worth of promotions in a matter of weeks. At that speed and scale, food and delivery merchants—already forced to process orders in seconds—often consider fraud a necessary cost of business, even when losses are mounting with no end in sight.

ATO, Credential Stuffing, and Credit Card Fraud Dominate

Account takeover and fake account creation further strain the system. Digital criminals use stolen credentials to access user accounts and place fraudulent orders, drain loyalty points and gift cards, and hijack customer promotions. Anyone can easily create multiple accounts to exploit a business, and a single bad actor can be responsible for significant loss: in some cases, one fraudster has created hundreds of fake accounts to consume thousands of dollars worth of promotions in just a few weeks. At that speed and scale, food and delivery merchants—already forced to process orders in seconds—often consider fraud a necessary cost of business, even when losses are mounting with no end in sight.

Stolen credentials are regularly sold on the dark web in conjunction with compromised payment methods and other data, making for a lucrative marketplace of exposed information. The lack of two-factor authentication on many QSR platforms, coupled with the speed of transactions, makes the issue especially difficult to manage.

Credential stuffing—automated login attempts using stolen credentials—is a major driver of these account takeovers. Bots can test thousands of credential combinations across multiple platforms in minutes, exploiting weak or reused passwords to gain access. Fraudsters not only test the login credentials, but once inside, can also check balances, credit cards on file, and the date of an account's most recent order. They can place high-volume orders, reroute deliveries, or steal stored payment data. For food and delivery platforms that lack real-time bot mitigation or behavior-based detection, credential stuffing attacks can overwhelm infrastructure and create long-term trust issues with legitimate customers.

Credit card fraud currently accounts for 99.6% of fraud incidents in the food & delivery space according to FIBR, highlighting downstream fraud as an expensive symptom of ATO. While manual review rates have decreased slightly in previous months to 0.026%, they’re not an ideal indicator of true risk in a space where human oversight for every transaction is an impossibility. That’s a reality that won’t change, and because fraud tactics will never stop evolving, businesses have to take a proactive approach to risk management that’s both flexible and sustainable.

The Currency of Trust

Refining security measures is only one pillar of customer trust. It's equally important to deliver user experiences that meet high expectations for speed and ease. Overly stringent security protocols deter trusted customers and lead to cart or brand abandonment, and the protection they offer against fraud is rarely outweighed by the cost of low order acceptance, dropped transactions, and decreasing lifetime value.

These challenges are inspiring businesses to turn to advanced trust and safety platforms, AI-powered ML models and algorithms, and solutions that can detect unusual behavioral patterns in real-time. The ability to pinpoint risk with precision also allows businesses to accept more transactions with confidence. Along with other strategies like implementing resilient identity verification processes, requiring PINs for order pickups or deliveries, and educating customers about common scams, food & delivery brands reduce the risk of fraud with every step taken to adapt alongside emerging threats.

But what truly sets successful food and delivery platforms apart is their ability to balance strong protection with instant gratification—identifying legitimate customers without friction, and stopping bad actors before they cause financial and reputational damage. Proactively tracking and benchmarking fraud rates clarifies where and when risk is highest, helping food & delivery platforms make smarter decisions about friction and overall operations and in turn, strengthening brand equity and customer trust.

Visit FIBR to compare your company’s payment fraud, account takeover, and chargeback rates to Sift global industry benchmarks.

The National Restaurant Association Show in Chicago brought together experts and executives from across the industry to discuss the latest hospitality trends, ranging from artificial intelligence to a return to on-premise dining. But there were some questions that went unanswered, at least to the satisfaction of Restaurant Dive. 

Here are three questions Restaurant Dive has been mulling over since we left Chicago.

There are robots at NRA, but why don’t we see more in restaurants?

From servers to robotic fry cooks and automated baristas, the NRA show floor, as always, was replete with new tech that promised to save labor.

Yet, aside from instances when Restaurant Dive has sought them out, restaurant robots still seem rare in the wild. But there is also scale to contend with — there are, after all, around 724,000 restaurants in America, according to the U.S. Bureau of Labor Statistics. 

As early as 2010 — 15 years ago — The New York Times profiled robots as if they were poised to take jobs in the industry. Yet in 2025, there are several million more restaurant workers than in 2010, and nominal wages are higher than ever. 

Some full-service dining players, like Bruce Nelson, a fractional CFO across several restaurant groups in Minnesota, think that AI could work in the back of house to analyze data. But major use of robots to serve customers runs counter to the ethos of hospitality, Nelson contends.

“When it comes right down to it, we are a human business, and we always will be a human business,” Nelson said. “I don't want to go to a restaurant and have a robot feed me.”

An answer to the question might come from the robots themselves, and their developers. Richtech Robotics, now on its fourth NRA show, showed off a new version of the company’s barista bot, Adam.

“Train me once and I can craft coffee, shakes, cocktails, make boba tea, grill burgers, cook noodles or [do] any countertop task you ask of me,” Richtech’s robot intoned at its booth.

But Adam costs $3,500 a month for “robot-as-a-service,” according to Timothy Tanksley, Richtech’s director of marketing. The average production and nonsupervisory restaurant worker earned $19.19 per hour in February, according to the BLS. 

Adam would have to operate 182.3 hours a month — 42 hours a week — for its monthly cost to equal the labor cost of a worker. 

That assumes a robot is a one-to-one replacement for a worker, and that it costs nothing to install. Yet Adam, per Tanksley, is meant to function alongside baristas, not replace them. 

“As soon as an order comes in, he's making it, he's sending it out, that kind of thing. And that really allows the humans to talk to guests,” Tanksley said. 

Major brands are signaling that winning price-sensitive consumers depends on convincing them to treat cafes as places to hang out and socialize, with baristas as well as customers. Does it make sense to hire the equivalent of a full-time employee just to make it so existing employees can talk with consumers a bit more? If not, where does a robot fit in?

Maybe nowhere. Richtech’s Adam is deployed in fewer than 10 coffee shops and a handful of Walmart units across the U.S. 

Is a recession around the corner?

The biggest question at the show was also one nobody could answer definitively. What is the macro outlook for the economy? 

Traffic is down, but not by much, said Victor Fernandez, chief insights officer at Black Box Intelligence. David Portalatin, Circana’s senior vice president and industry advisor for food and foodservice, echoed this sentiment. Both cautioned that this is a continuation of previous trends — if tariffs are going to have a serious negative impact, it hasn’t shown up yet.

Or maybe traffic isn’t down at all — for some restaurants, at least. Both Jonathan Gillespie, a partner at Chicago fine dining restaurant Adalina, and Nelson, said their concepts have either seen recovery after a drop in the winter, or they outright haven’t seen traffic fall.

“Restaurants are the canary in the coal mine,” Nelson said. Restaurants are highly discretionary, after all — the consumer can always eat at home. 

And so far, the data is noisy. Earnings season bore this out. Many brands saw traffic fall, but some saw shocking surges, and others — including Starbucks — saw their losses shrink. With tariffs starting to bite and more potentially on the table, we may have an answer sooner rather than later.

Can the mocktail replace the cocktail?

Gen Z consumers are likely drinking less alcohol. Good for society? Maybe. Good for restaurants? Probably not — alcohol is a high-margin, high-price good. 

If consumers aren’t pulling back from liquor out of health concerns, they may be doing so to manage their spending, Fernandez said. 

About a quarter of companies Fernandez works with say they’re seeing guests skip alcohol, or beverages altogether, in the current macro environment.

Nonalcoholic beverages are filling the gap, somewhat. Nelson said restaurants he works with are offering non-alcoholic cocktails for a $6 or $8 price point, while an alcoholic drink costs roughly twice that. 

Many consumers, especially Gen Z, are willing to shell out for beverages, to the point that afternoon beverages are an emergent occasion. McDonald’s is looking to capture some of that interest with its forthcoming beverage tests. Still, sugary iced coffee and refreshers loaded with boba may fulfill a different need than mocktails.

In many markets, consumers may be paying at parity for non-alcoholic drinks, said Casey Gamblin, area beverage manager at Ithaka Hospitality Partners.

“NA beverage pairings versus spirit forward beverage pairings are being priced exactly the same, I'm seeing that in cities like here in Chicago and New York City, LA,” Gamblin said. 

Some consumers are certainly willing to pay $14 for a glass of bitters, soda water and botanicals. But how many? And for how long? And will the humble club soda with lemon, ordered from an obliging bartender, remain an appealing option for sober consumers who still want to fit in at bars and restaurants?

Restaurants can speed up employee recruitment using technology like applicant tracking systems, automated interview scheduling and mobile application options, according to the National Restaurant Association’s Workforce Technology Research Insights report.

The organization, with the help of Paradox , interviewed leaders from 16 restaurant companies about their hiring processes and retention strategies to find what employers are prioritizing as the labor market cools.

Thirty-seven percent of operators said they plan to adopt labor management and recruitment systems while 28% said they were interested in artificial intelligence-driven solutions, NRA said. 

Employers are exercising greater discretion in hiring, the NRA found, while looking for ways to save time for managers. Key measures of labor mobility have flattened over the last two years,  with job openings still high and sectoral unemployment rising.

“The process of reviewing applications, conducting interviews, and onboarding new hires consumes valuable time that could otherwise be spent running the restaurant,” the NRA said. 

Large numbers of operators, the report claimed, are looking to invest in automated labor management, recruitment and scheduling systems. 

Southern Rock Restaurants, a major McAlister’s Deli franchisee, said technology now made it possible to hire workers within 24 hours, while the operator’s hiring time averaged 14 days when using a traditional, manual recruitment approach.

The operator uses a tracking system that schedules interviews automatically, the report said. A vast majority of applicants (86%) apply to Southern Rock through mobile devices, with the bulk of applications submitted off-hours. This is in line with general workforce trends, as 54% of job seekers applying at night or on the weekend, according to the NRA report. 

Southern Rock “provides QR codes and text-to-apply options in their stores, enabling customers to submit applications conveniently via cell phone. They also use QR codes for employee referral programs, offering bonuses for successful referrals,” the NRA said. 

The employee referral program blends the speed and convenience of technology with the reliability of employee recommendations. 

“Two-thirds of operators [say] that their own employees are their best ambassadors for attracting new recruits,” the NRA found.

Despite the availability of new technologies, some employers prefer to stick with older hiring practices. One family-owned dining group stuck with time-consuming manual processes because of the expressed preferences of its managers, but was still successful at filling its shifts, the NRA said.

Finding candidates is only half the battle, operators told the NRA, with many focusing on trying to improve retention. One-third of candidates who accepted an offer dropped off before their start dates and over half didn’t make it to their first 90 days, NRA said, citing data from Paradox. The Paradox data includes data from restaurant and non-restaurant firms.

Operators interested in retention are investing in regular management communication and proactive support for new workers, the NRA said. One way to achieve that is a standardized training program.

A multi-brand restaurant company uses a “20-day promise” as part of its onboarding program. The 20-day promise communicates expectations to workers, including shift-by-shift training benchmarks. Transparency like that helps ease workers into their new roles, the report said. 

Papa Johns announced Thursday a partnership with Google Cloud to use Google data analytics, machine learning and artificial intelligence chatbots in its ordering and delivery processes.

The company expects the partnership to increase ticket size, frequency and customer satisfaction scores while reducing the cost of customer service.

The pizza chain has created an “innovation team,” called PJX, to take advantage of the opportunities offered by the partnership.

Papa Johns’ PJX team will pursue at least five key projects as the tech arms race heats up among QSR heavyweights. The most interesting of those strategic priorities is the deployment of a Google Cloud-based point-of-sale system, which Papa Johns said will set the stage for AI-driven dispatching, delivery route optimization and automation of unspecified restaurant processes.

PJX will try to anticipate customer needs and use Google BigQuery, Vertex AI and Gemini to “proactively suggest orders through push notifications or email, based on learned customer preferences and anticipated needs,” the company said.

The pizza brand will also feed its internal data into Google’s generative AI tools to improve personalization of its web and app experiences by “presenting unique discount codes or advertisements based on previous orders, customer preferences, location and more,” in real time.

Additionally, Papa Johns said the partnership would help it predict ordering behaviors and generate AI-powered marketing campaigns.

Papa Johns will turn over much of its customer experience management to AI bots. The pizza chain plans to integrate AI-voice ordering into its app and “build an AI-powered chatbot that can handle common customer inquiries, seamlessly escalating complex issues to live agents.”

Whether the use of chatbots will actually improve consumer experiences is open to question. AI chatbots have a documented tendency to invent inaccurate information.

The computational demands of AI have driven up greenhouse gas emissions at Google, Microsoft and other tech giants. Bloomberg has reported the energy intensity of AI data centers is endangering corporate climate commitments. According to Papa Johns’ 2024 Corporate Responsibility Report, the pizza brand is committed to sustainably managing greenhouse gas emissions and other environmental concerns. 

Papa Johns did not immediately respond to a request for comment on AI’s accuracy and emissions problems, or on whether generative AI represented a qualitative improvement on other forms of data analytics.

Same-store sales at the pizza brand have been uneven in recent years, ranging from positive 3% in Q3 2023 to negative 6% in Q3 2024. To combat this, CEO Todd Penegor, who joined the brand from Wendy’s in mid-2024, is pushing the brand to focus on its core pizzas and to improve its market density through development.

If the AI partnership is successful at raising tickets and transactions, or if it lowers the cost of customer service for the brand, it could help Papa Johns focus on its core strategies. 

Papa Johns is one of many restaurant companies looking to cash in on the AI boom. Yum Brands recently signed a deal with Nvidia, a major AI developer, to build more analytics and AI tools. Wendy’s is speeding up the deployment of drive-thru voice AI. And Hi Auto, a QSR-focused voice AI tech firm, recently announced a $15 million funding round to help it develop and deploy its AI tools.