The Cashier Who Isn't There: From Digital Offshoring to Inevitable AI Replacement?
When a Goldman Sachs employee walks into Yaso Kitchen in New Jersey to order Chinese dumplings, they expect to find a cashier behind the counter. Instead, they find Amber, a Filipino woman who greets them from a screen mounted on a tablet. The initial reaction is confusion: "I thought it was an ad, like the ones in taxis," the customer told the press. But Amber is really working, eight hours a day, from Manila. It's the first remote shift of her life.
The scene is repeated in dozens of restaurants across Manhattan, Queens, and Jersey City. Happy Cashier, a New York startup founded by Chi Zhang, has placed these virtual cashiers in chains like Sansan Chicken and Yaso Kitchen. The value proposition is brutally simple: with New York's minimum wage at $16 an hour, why hire local staff when you can pay Filipino workers about $3 an hour? An eighty percent saving on labor costs.
Zhang knows the industry from the inside. After opening a restaurant in Brooklyn in 2015, he had to close it during the pandemic partly due to the difficulty of hiring workers. The labor shortage in the restaurant industry is not a fantasy: it's the same pressure that has pushed chains like Chipotle and Sweetgreen towards automation. But Zhang chose a different path, drawing from the 1.3 million Filipinos employed in the BPO sector (Business Process Outsourcing), the world's largest with a turnover of $35.4 billion in 2023.
The technical infrastructure is surprisingly simple. Happy Cashier integrates with MenuSifu POS, a system already widespread in Asian American restaurants. Orders placed at the point of sale instantly appear on the remote cashier's screen, who manages payments and customer questions via a Zoom connection. The whole setup requires little more than a tablet, a webcam, and a stable broadband connection. No investment in sensors, computer vision, or sophisticated hardware is needed. It's old-school offshoring dressed up as digital innovation.
The Economics of Wage Arbitrage
Filipino workers earn an average of $3 per hour, a figure Zhang defends by stating he pays "150% more than the average for cashiers in the Philippines". The logic is ironclad: you respect local labor laws (those of the Philippines), but you exploit the wage differential between two markets. A worker costing $3 an hour in Manila does the same job that would cost $16 in New York. As Canadian lawyer Jonathan Pinkus clarified in the similar case of Freshii, a chain that in 2022 used Nicaraguan cashiers paid $3.75 an hour in Ontario: "If you're sending jobs to people in a different country, you're only obligated to meet the labor standards of that country. Being virtually present in Ontario doesn't change anything."
The phenomenon is not new. In 2005, Resource Group had already placed remote receptionists via video in Washington D.C. offices. What has changed is the scale and normalization. In September 2024, Happy Cashier aimed to expand its service to over 100 restaurants, with growth accelerating after its presentation at the National Restaurant Association Show in Chicago in May of the same year.
The reactions have been polarized. When the practice emerged on social media in the spring of 2024, the wave of criticism was immediate. Teófilo Reyes, head of Restaurant Opportunities United, called the situation "extremely alarming". Jonathan Bowles of the Center for an Urban Future warned that the fast-food workforce is already shrinking and "new technologies could further transform the industry."
Yet the data tells a more complex story. In July 2024, New York City had 285,700 people employed in the restaurant sector, compared to 265,400 in July 2023, an increase of twenty thousand jobs. Bowles himself observed: "The industry is growing despite the outsourcing of jobs. That doesn't mean that entry-level jobs haven't decreased, they might have. But it does mean that restaurants are investing in cost-cutting solutions so they can grow." It's the eternal paradox: automation eliminates jobs but can make businesses sustainable that would otherwise close, creating other roles.
The Intermediate Step
This is where the story takes a turn reminiscent of "Sorry to Bother You," Boots Riley's 2018 film where African American workers use a recorded "white voice" to sell over the phone. The digital offshoring of cashiers is not innovation: it's the final stage before complete automation. As Homescreen bluntly wrote in its analysis: "Today it's a Filipino woman behind a screen controlling a POS system, but it's not crazy to believe that probably in the next six to twelve months this could be an AI avatar doing all the same things."
Cashierless checkout technology already exists. Amazon Just Walk Out, launched in 2018, uses computer vision, sensors, and deep learning to automatically track what customers take from shelves and charge them when they leave. In 2024, Amazon launched more Just Walk Out stores than in any previous year, reaching over 300 third-party deployments in the USA, UK, Australia, Canada, and France, mainly in stadiums, hospitals, university campuses, and distribution centers.
The point is that these systems require significant infrastructure investment: ceiling cameras, shelf sensors, controlled entry points. The Information reported that the Just Walk Out team has over 1,000 employees, mainly in India, who label footage and transactions to train the AI models. Even Amazon's "fully automatic" AI depends on human moderators. It's a lesson found in Spike Jonze's "Her": the most convincing intimacy is often the most mediated.
But for Zhang's small restaurants, the remote system via Zoom is the perfect solution in the time gap between the physical cashier and the fully synthetic one. It's cheap, can be implemented in days, and above all, it maintains the illusion of human interaction that many customers still seek. The problem is that remote cashiers, as a professional category, are on borrowed time. Advanced voice models like GPT-5 can already handle natural conversations, understand menus, manage order changes, and process payments. The global market for AI in call centers was worth $1.95 billion in 2024 and is projected to reach $10.07 billion by 2032.
Klarna, a Swedish buy-now-pay-later company, announced in February 2024 that its AI chatbot was doing the equivalent work of 700 full-time agents. However, in May 2025, the company backtracked, returning to hiring people after admitting it had pushed too far towards automation, with the CEO acknowledging that an excessive focus on costs had reduced service quality. The results were mixed: the company saved money but had to rehire some employees to handle complex cases like identity theft. Nevertheless, the direction is clear. In the Philippine healthcare sector, 200,000 workers in medical call centers at the end of 2024 were already facing competition from voice AI assistants that schedule appointments, fill prescriptions, and triage patients.
Voices from the Front Line
Positions on the transition to AI in service jobs follow predictable but no less relevant lines. The AFL-CIO launched the "Workers First Initiative on AI" in October 2025, the first comprehensive agenda of the American labor movement on artificial intelligence principles. President Liz Shuler stated: "We reject the false choice between American competitiveness on the world stage and respect for workers' rights and dignity."
The central question is who manages the transition. As early as 2018, the AFL-CIO Commission on the Future of Work had identified "fissured work" and outsourcing models as the main strategy for companies to circumvent employment responsibilities and reduce labor costs. The digital offshoring of cashiers is exactly this: fragmenting the work process to exploit global wage differentials. But with AI, the next step removes the worker from the equation entirely.
Bernard Marr, an analyst specializing in ethical AI, pointed out in his 2026 trends report that many argue employers have an ethical responsibility to implement reskilling and training initiatives. But testimonies from the field tell a different story: on Reddit, workers reported being fired after training AI systems, raising ethical concerns about the transparency of corporate intentions.
On the European regulatory front, the EU AI Act, which came into force in August 2024, classifies AI systems used in the workplace as potentially "high-risk." Since February 2025, emotion recognition systems in the workplace have been banned, while from August 2026, the rules on high-risk systems will become fully operational, including AI tools for hiring and performance evaluation. Lawmakers in 2026 are considering what level of human supervision should be required and what penalties to apply when organizations allow machines to act irresponsibly.
But America is different. Americans in 2026 show a deep skepticism towards AI, with fears centered on job displacement, ethical errors, and social disruption. Yet regulatory protections are fragmented, delegated to states and cities. The Biden administration had issued an executive order in October 2023 requiring federal agencies to balance the benefits and legal risks of AI, but practical implementation remains nebulous.
The Last Mile of Automation
Technological convergence is already underway. Presto, a company specializing in restaurant automation, is already making moves in this space, developing systems that combine touch tablets with voice AI. MenuSifu, the platform that supports Happy Cashier, has announced a partnership with Hestia to introduce automatic cooking machines, creating a "complete digital solution 1+2+4+N" that includes virtual cashiers and kitchen automation.
The timeline is clear. According to a study from the University of California San Diego published in 2025, cashierless technology not only reduces staff costs but also significantly changes consumer purchasing behavior. Cashierless stores saw a thirteen percent increase in demand in the five to ten minutes before classes started on university campuses. Condom sales increased by 164%: eliminating the embarrassment of human interaction has measurable effects.
Deloitte predicts that 25% of companies already using generative AI will implement AI agents by the end of 2025, a figure set to double by 2027. The conversational AI market will grow from $17.05 billion in 2025 to $49.8 billion by 2031, with a compound annual growth rate of 24.7%.
For the remote workers in the Philippines, the irony is cruel. Ruth Elio, an occupational nurse who supervises a medical call center in Manila, recounted workers handling medical emergencies for lonely elderly Americans, even though it wasn't part of their official duties. They develop headaches and back pain from sitting for hours. Now these same workers, who made digital offshoring possible, are at risk of being the first to be replaced by synthetic voices that don't need breaks, shifts, or salaries.
Zhang of Happy Cashier stated he aims for five hundred restaurants before "making a big leap." He did not specify what this leap is, but the context is telling. The question is not whether remote cashiers will be replaced by AI, but when. And the answer, looking at technology adoption curves and cost differentials, is probably within the next two years.
The cashier you see on a screen today is already a ghost. Not the ghost of Casper or a Pixar animated film, but the kind of ghost that appears in "Severance," the Apple series where work is fragmented to the point of losing all existential continuity. Amber from Manila works for restaurants she will never see, serving customers who mistake her for an advertisement, waiting for a language model to learn to imitate her voice well enough to make her superfluous. It's deferred automation, outsourcing as the larval stage of the planned obsolescence of humans in low-margin services.
Technology has no morals of its own, but the choices on how to implement it do. We can build systems that augment human capability or systems that simply eliminate costly humans. So far, in the restaurants of New York as in the call centers of Manila, the second option is winning by a landslide.