Can centuries of change in the workplace be packed into one decade? That is what OpenAI CEO Sam Altman thinks and he predicts the first to go will be customer service. “I’m confident that a lot of current customer support that happens over a phone or computer, those people will lose their jobs, and that’ll be better done by an AI,” Altman told “The Tucker Carlson Show.”
The technical foundation for making this transition is already established. Large language models (LLMs) like GPT-4 can understand natural language, read and synthesize data, and respond in human-like fluency. Such systems are top-notch at tier-one support high-volume, low-complexity transactions such as account balances or delivery tracking work that constitutes a large part of call center loads. Salesforce CEO Marc Benioff has recently asserted that AI agents are now “indistinguishable” from human representatives, while Swedish fintech Klarna said its AI assistant was doing the equivalent of 700 full-time agents’ work with the same or improved customer satisfaction ratings.
The appeal to businesses is obvious: AI-powered customer service scales on the fly, works 24/7, and can be directly integrated with backend systems to close requests in real time. But the rollout isn’t frictionless. Most initial rollouts have been revealed to contain technical flaws McDonald’s dropped its Automated Order Taker in more than 100 restaurants after high-profile mistakes were viral sensations. Gartner surveys indicate close to two-thirds of customers don’t want to talk to AI for service, and the reasons include difficulty contacting a human and lack of trust in automated answers.
Altman’s warning is broader than call centers. He is “unsure” about the long-term effect on programming careers, even as coding tools with AI, such as GitHub Copilot and Amazon CodeWhisperer, already perform routine development tasks automatically. Such systems can write boilerplate code, fix errors, and even suggest architectural solutions, possibly lowering demand for junior developers. A 2025 World Economic Forum report estimates that 40% of programming work could be automated by 2040.
It is the speed of change that makes this moment different from previous industrial transformations. Previously, approximately half of all occupations have undergone substantial change approximately every 75 years. Altman speculates AI may create a “punctuated equilibria moment” in which that magnitude of change occurs within a few years. McKinsey analysts put the estimate that existing AI and automation tools are already capable of doing employees’ work activities that consume as much as 70% of employees’ time, with customer service, data entry, and routine programming among the most vulnerable categories.
Whether this will be a repeat of the Industrial Revolution or an unprecedented discontinuity is controversial. Some, such as RethinkX’s Adam Dorr, liken workers’ fate to that of horses after the invention of the automobile swift obsolescence. Others, including Wharton’s Ethan Mollick, highlight instances in the past where automation improved productivity and produced higher-quality jobs in the long run, even if sometimes after wrenching transitions. The cotton gin, for instance, added jobs in processing cotton by reducing costs and triggering demand to spike, but also raised competition and decreased wages.
The underlying technology is evolving at breakneck speed. Generative AI’s disembodied nature deployable through existing browsers and apps means adoption barriers are low. Unlike industrial robots, which require capital-intensive installation, AI chatbots can be integrated into existing workflows with minimal infrastructure. This ease of diffusion is why ChatGPT reached 1 billion monthly visits just four months after launch, a rate of adoption far faster than the personal computer or smartphone.
But the engineering problem isn’t simply one of capability it’s reliability at scale. Successful AI customer service systems need to deal with context switching, reconcile disparate data sources, and sustain conversational coherence across long-lived interactions. They need to cope with edge cases in which there is incomplete or conflicting information, an arena in which human judgment continues to dominate algorithms. As Jason Maynard, the CTO at Zendesk, points out, Tier-one operations are already being automated successfully, but tier-two and tier-three issues those with complexity or ambiguity remain difficult for AI to resolve.
The stakes are high. McKinsey estimates as many as 12 million workers in the U.S. and Europe might have to switch professions by 2030 because of automation, with office support and customer service being the most impacted. For business, the math is motivated by cost pressures and competitive forces when one company reports it eliminated hundreds of service positions through AI, competitors are under instant shareholder pressure to do the same. For employees, the shift will depend on reskilling into positions where AI complements instead of substitutes human skills.
Altman’s projection indicates that the comfortable rhythm of technological upheaval slow accumulation, incremental adoption, generational replacement might no longer hold. If the pace of AI continues, the engineering choices made in the next half decade could shape not only the effectiveness of customer service systems or the productivity of software developers but the organization of whole labor markets for decades to come.
