Will AI really decimate human jobs? Tech industry insiders are split

CNN

21-07-2025

People have worried that computers will take their jobs for basically as long as they've been around, but those fears have felt more realistic than ever over the past year as artificial intelligence has begun to overhaul the way people work.
Anthropic CEO Dario Amodei amplified those concerns in May when he warned AI could spike unemployment, particularly among white-collar jobs, to 20% over the next one to five years. Certain companies are already enlisting AI to do some of the work previously done by people; Meta, Microsoft and Salesforce are increasingly using AI to code among other tasks. And CEOs of companies ranging from Amazon to JPMorgan have warned their human workforces will shrink because of AI.
However, some of those predictions deserve a healthy dose of skepticism. 'AI is so good, it's going to put humans out of jobs' is a strong marketing message for companies selling the technology, and a potentially convenient excuse for an executive already mulling a workforce downsize.
The answer to whether AI really spells trouble for human workers isn't so black and white. That's according to more than half a dozen tech industry insiders CNN spoke with over the past month, who are mixed on just how much and how fast AI will upend the job market.
Nvidia CEO Jensen Huang, for example, told CNN AI will kill jobs only if 'the world runs out of ideas.' And Google DeepMind CEO Demis Hassabis told CNN that an AI 'jobpocalypse' is among his minor concerns when it comes to the technology's impact.
Nonetheless, tech companies themselves have cut hundreds, in some cases thousands, of roles this year as they implement AI to take on a growing share of software development and other tasks.
There does seem to be broad consensus that the nature of work, including how it's done and which tasks humans do, is going to change, and perhaps more rapidly than with any previous technological transformation the world has seen before.
More than half of Americans say they're worried about AI's impact on the workplace, and a third believe AI will lead to fewer job opportunities for them over time, according to a Pew Research Center survey published in February. But the troubling AI failures that seem to make headlines every few weeks — from Elon Musk's Grok chatbot spouting antisemitic tropes after what the company called a faulty update to newspapers running AI-generated summer reading lists containing made up books — can make it hard to believe a wide-scale reliance on a robot workforce is just around the corner.
'I think that there will be some displacement. I think there'll be new job categories that emerge,' said Gaurab Bansal, executive director of the nonprofit startup consultancy Responsible Innovation Labs. 'I think we're looking at a complex reshaping, rather than a straightforward elimination.'
And we're just at the start.
'I think we're entering a decade-ish, maybe more, period of uncertainty,' Bansal said.
Small circles in Silicon Valley have for years been discussing how AI will upend the labor market — along with proposals for mitigating that impact, such as universal basic income, an approach that OpenAI CEO Sam Altman has backed.
But the conversation has burst onto the public scene in recent months as tech giants introduce 'agentic AI' tools.
Unlike chatbots, where in each interaction users put in one question and get back one relatively simple response, agentic AI systems can handle more complex, multi-step tasks without step-by-step prodding. Think: coding a website based on a user's idea of what it should look like or do, or researching a topic and compiling it into a presentation.
OpenAI last week launched an agent mode for ChatGPT that can accomplish tasks on behalf of users, while Anthropic in May introduced a model it claims can work independently for almost an entire workday.
'Now, you can give (the AI system) a goal, and it'll automatically break that down into a sequence of steps that it needs to perform the goal,' Swami Sivasubramanian, vice president of agentic AI at Amazon Web Services told CNN. 'You suddenly have a reasoning, thinking system that can use various tools, so now the possibilities are really up,' in terms of how they can be used in the workplace.
Amazon, for example, used an AI developer agent to upgrade 30,000 software applications across the company last year. The project would have taken an estimated 4,500 software developers a year to complete, but the AI completed the task in just six months, according to the company. The switch saved Amazon around $250 million in capital expenditures, Sivasubramanian said.
Microsoft CEO Satya Nadella also said earlier this year that 20% to 30% of the company's code was being generated by AI. Meta CEO Mark Zuckerberg said around half of the company's code development would be done by AI by next year. And Salesforce CEO Marc Benioff told Bloomberg the company uses AI for about 30% to 50% of the company's work.
Unlike earlier technology transformations, like the automation that's taken place in industrial settings, AI is relatively easy to adopt to improve or take over human tasks.
'AI workers are just software, and so you don't need to buy expensive physical machinery,' said Steven Adler, a former OpenAI researcher. 'The AI workers can also be upgraded easily when there's a stronger version. Particularly for companies used to getting software from the cloud … there's less friction to (deploying) a 'virtual coworker' product than ever before.'
While certain categories of work, including coding and data analytics, are ripe for more significant disruption, many experts argue other roles will undergo changes rather than being eliminated altogether. Most workers, they say, will use AI to automate repetitive tasks, leaving them more time to work on creative or relational aspects of their job.
A doctor, for example, might have an AI assistant take notes during a patient's visit and fill out their chart autonomously, giving that doctor more time for face-to-face discussions with the patient.
'Most tasks for most jobs can't be automated,' Meta's Chief AI Scientist Yann LeCun said in a LinkedIn post last month.
Put another way, AI is 'good, but not perfect, for a subset' of tasks that human workers currently do, but it can't replace most people outright, said Yacine Jernite, machine learning and society lead at open-source AI firm Hugging Face.
Companies and governments are investing in training workers for the AI era, including an AI training academy for teachers that a group of tech giants is partnering with teachers' unions to build in New York City. The White House last month launched a pledge for companies committed to investing in AI training and education for America's youth, which has been signed by 68 firms.
Dennis Woodside, CEO of the IT and customer service management software company Freshworks, said his firm is shifting workers that once reactively responded to customer support requests — a job more easily done by AI — to more hands-on work with clients.
Some in the industry believe that AI will make some jobs obsolete, but it will also create new categories of work we can't currently imagine.
'We saw that in the Internet era, where trillions of dollars of economic value were created each year and some of the biggest, most valuable, most successful companies that were started back then are prominent today in the economy,' said Dan Priest, chief AI officer at professional services firm PwC. 'Net-net there should be a positive benefit to jobs growth, it's just going to be different jobs.'
Still, Jernite said he worries leaders in Corporate America might feel pressured to start making changes simply because they've heard that AI could replace jobs.
'And people follow that direction, they fire people, and it doesn't have anything to do with what the technology can or cannot do, but how it's perceived,' Jernite said. 'Then they rehire those people when they figure out that's not how you leverage this really wonderful technology in a way that's responsible and sustainable.'
The transition could be uncomfortable. Adler, the former OpenAI researcher, said he expects many white-collar workers will start to see lower wages as AI augments their jobs.
'AI will make an individual worker more productive and will help more people to be capable of doing a given job,' Adler said. 'The net effect is an oversupply of labor, which pushes wages down unless there's a big surge in labor demand.'
Ultimately, Bansal said he believes policymakers must create a new economic framework for the AI era to ensure it's not just powerful tech companies benefiting at the expense of regular workers with fewer opportunities and more pressure to be productive. And, ideally, they need to move a lot faster than they did in addressing challenges posed by earlier waves of technology.
'We need a new social contract for this era,' he said. 'The bargain we have between workers and sort of the economy is born of a different technological era.'
CNN's Lisa Eadicicco contributed to this report.