In 2023, Goldman Sachs estimated 300 million jobs across the world could be exposed to artificial intelligence.
More recently, a separate analysis by the investment bank found that AI could displace six to seven percent of the US workforce if it is widely adopted, while a McKinsey paper found that 30 percent of the hours currently worked across the US could be automated by 2030.
A report by Boston Consulting Group (BCG) meanwhile took it a step further, declaring that “five years from now – or perhaps further in the future – 10 to 15 percent of jobs in the US could be eliminated.” The firm concluded that “remaking the workforce is a competitive imperative.” The shift has been described as the cognitive, white-collar equivalent of the Industrial Revolution.
Major tech firms have already begun making large-scale job cuts while simultaneously pumping large sums into AI. In February, Jack Dorsey announced his fintech firm, Block, would be slashing its workforce from 10,000 to 6,000 people, directly citing intelligence.
“We are already seeing that the intelligence tools we are creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company,” wrote Dorsey. “I had two options: cut gradually over months or years as this shift plays out, or be honest about where we are and act on it now.”
In April, Meta announced it planned to cut around 8,000 (10 percent) of its employees to “offset the other investments we are making,” in the words of chief people officer Janelle Gale. Last year, Microsoft made more than 15,000 job cuts while committing $80bn to AI investments. Amazon Chief Executive Andy Jassy has meanwhile said he expects the conglomerate to “reduce its total corporate workforce as we get efficiency gains from using AI extensively across the company.”
But as these firms bet on an AI-driven future with slimmer teams and higher productivity, there are multiple questions; how realistic is this vision, are consumers ready to put their trust in bots, and what might it mean for the future workforce if concepts that were once purely the stuff of sci-fi come to fruition?
All hail the AI agents
Underpinning much of the speculated AI-driven workforce of the future is agentic AI, or AI agents – that is, agents that can act on content produced by generative AI to plan, make decisions and execute tasks. Going beyond the limits of generative AI by taking a more active, human-like role, they can come in digital form – such as chatbots and coding agents – or physical embodiments, including robots, drones and self-driving cars.
AI could wipe out half of all entry-level white collar jobs
These tech-forward ‘agents’ are already making an early-stage appearance; one example is OpenClaw, which launched in November 2025 and, as per its website, “clears your inbox, sends emails, manages your calendar, checks you in for flights. All from WhatsApp, Telegram, or any chat app you already use.” Firms in both the tech sphere and beyond are starting to capitalise on the hype.
In May, Citigroup announced Arc, a new platform that will “allow the bank to build and scale AI agents across the firm,” in the words of a recent statement. “These agents enhance human judgment by taking on tasks such as research, synthesis, preparation and execution, reducing manual effort and accelerating how teams operate day to day,” it reads.
The same month, AI lab Anthropic released a series of new agents (or AI systems) for the financial services industry that can carry out tasks with limited human intervention – including building a pitchbook, auditing statements and drafting credit memos.
JPMorgan is going full steam ahead in its plans for agentic AI. Chief analytics officer, Derek Waldron, told the CNBC last year that in its future vision, “every process is powered by AI agents, and every client experience has an AI concierge.” Amazon chief executive Andy Jassy outlined a similar premonition, declaring in a statement in 2025: “There will be billions of these agents, across every company and in every imaginable field.”
An ongoing need for humans
The general messaging from many experts in the industry is that the rise of these agents would entail a combination of human and AI work, providing workers with more time to do (currently) human-only tasks. “Even as these systems become more capable, it remains critical that humans stay in the loop, setting direction, applying judgment and remaining accountable for outcomes,” Sida Peng, Research Economist at Microsoft, told World Finance.
Even highly skilled and highly valued positions are now under threat
“What that means for the workforce is that people spend less time on repetitive execution and more time directing work, making decisions and managing outcomes across teams of agents and humans.”
It’s not just about automating existing human tasks but augmenting them, according to Peng. “One of the biggest opportunities AI presents is the ability to expand what people and organisations are capable of,” he says. “In our Work Trend Index research, 58 percent of workers said they are producing work they could not have done a year ago, rising to 80 percent among Frontier Professionals [AI-first employees using agents to ‘augment or automate’].”
The rise of the ‘feeling economy’
This certainly sounds promising – and the fundamental shift in how we work could mean mass new job creation. A recent McKinsey paper, The agentic organisation: Contours of the next paradigm for the AI era, points to ‘M-shaped supervisors’ and ‘T-shaped experts.’ “New profiles are emerging, such as agent orchestrators, who design and supervise agent workflows; hybrid managers, who lead blended human-agent teams; and AI coaches, who help employees integrate AI into daily work,” wrote the researchers.
Another study, Artificial intelligence, workers, and future of work skills, meanwhile highlights ‘trainer’ roles, where human workers teach AI systems how to perform (including curating data); ‘explainer’ roles, requiring communication and technical skills to explain AI output for end users; and ‘sustainer’ roles to help make sure AI systems in organisations operate fairly and transparently. This shift could mean ‘human’ skills become increasingly in demand, according to many experts. A recent independent paper, How AI influences the experience of work: upsides and downsides for workers, describes this as a new ‘feeling economy,’ with interpersonal skills, empathy and social and emotional intelligence at the core.
Pencheng Shi, Associate Dean for Research and Program Director at the Rochester Institute of Technology, agrees. “What I have seen is that employers are starting to look for people who are much better in softer skills,” he says. “These include leadership, organisation, communication and really understanding the human side.”
He believes competing with AI could bring other benefits too – challenging us to go beyond what AI can currently do and bring elements that are, for now at least, uniquely human. “We need to push ourselves and recognise what makes us different – from each other but also from the machines,” he says. “Work that has your own unique signature and your own ideas is what will stand out, and what makes every one of us different. As people we bring different perspectives, and I think that’s still, for now at least, a human-only strength.”
The entry-level threat
But while this human enhancement is an exciting prospect, there are likely to be significant challenges if and as we move towards what McKinsey researchers recently described as “the largest organisational paradigm shift since the industrial and digital revolutions.” Anthropic Chief Executive Dario Amodei told Axios last year that while AI currently is indeed augmenting human roles, the balance could soon tip into automation – with entry-level roles among the hardest likely to be hit (with many entry-level admin tasks such as legal admin, data entry and simple coding already automatable with AI).
“AI could wipe out half of all entry-level white collar jobs and cause an unemployment surge of 10 to 20 percent in the next one to five years,” he said in the report, adding that AI firms and governments need to stop “sugar-coating” the impending era and take action to raise public awareness of the very real threat. “Most people are unaware that this is about to happen,” he said. “It sounds crazy, and people just don’t believe it. We, as the producers of this technology, have a duty and an obligation to be honest about what is coming.”
A McKinsey survey in 2025 on the new era of work found that 51 percent of organisations reported generative AI was reducing their need for entry-level positions. Shi says he has witnessed a similar trend in the computer programming field. “I am seeing that the big computer companies are only hiring people with 10–15 years’ experience and PhD graduates,” he says. “On the other hand, graduates with bachelor’s and master’s degrees are facing tremendous employment challenges that were unimaginable only three years ago. So even highly skilled and highly valued positions are now under threat.”
Widening the social gap
This narrowing of the workforce is perhaps inevitable if the AI agent vision becomes reality. “In the agentic organisation, structure will pivot to small, outcome-focused agentic teams,” researchers wrote in the McKinsey report on the ‘agentic organisation.’ The analysis found a human team of just two to five people could oversee an agent factory of between 50 and 100 “specialised agents running an end-to-end process such as onboarding a customer, launching a product or closing the books.”
Roles and sectors especially at risk have been widely publicised. A Microsoft report last year ranked the 40 jobs with the highest ‘applicability score’ (or crossover) with generative AI, putting historians, writers, customer service reps, telephone operators, sales reps, journalists, mathematicians, data scientists and personal financial advisors high on the list.
As the list suggests, it’s not just ‘knowledge’ sectors likely to feel the impact; lower-wage workers are especially at risk, according to a 2024 McKinsey analysis, A new future of work: the race to deploy AI and raise skills in Europe and beyond. Researchers found that in Europe, workers in the two lowest-wage-bracket jobs were three to five times more likely to have to change occupations than higher earners; in the US, this figure surged to 10 to 14 times.
Many believe this could cause a concerning gap between those possessing AI and other in-demand skills in the ‘agentic’ era, and those who don’t – and are underlining the importance of educating every area of society in order to democratise the market. “Workers will need to acquire new skills to transition to better-paying work,” wrote the McKinsey researchers in the Race to deploy AI paper. “If that doesn’t happen, there is a risk of a more polarised labour market, with more higher-wage jobs than workers and too many workers for existing lower-wage jobs.”
A new economic model?
While some have argued AI could actually shrink the disparity by making skills more accessible to everyone who has access to the intelligence, Shi believes the social and cultural upheaval is one of the biggest risks we face as AI grows in its dominance – and that wealth distribution will be key if the more negative forecasts become reality. “I think if we don’t change the way we distribute wealth, the social bifurcation in the AI era could get drastically worse,” he says. “I think societies need to have a safety net in place to support those who may pay a bigger price.”
Anthropic’s Amodei likewise believes these discussions are of the essence. “It is going to involve taxes on people like me, and maybe specifically on the AI companies,” he told Axios, floating the idea of a ‘token tax’ on AI revenues. Earlier this year, Elon Musk proposed the idea of a ‘Universal High Income’ to counteract any potential large-scale, AI-induced unemployment. A development on the earlier economic idea of ‘Universal Basic Income’ (UBI), this assumes that AI would lead to a significant increase in productivity, boosting the economy and creating enough wealth to go around.
“Universal High Income via checks issued by the federal government is the best way to deal with unemployment caused by AI,” he tweeted. “AI/robotics will produce goods and services far in excess of the increase in the money supply, so there will not be inflation.” While it’s an interesting, potentially utopian idea, some have already posited issues around its predecessor, UBI; this would likely bring further challenges still.
Psychological threats
In any case, it’s not just the financial impact that needs to be considered if this fundamental transformation comes about in the way many are predicting it will. McKinsey’s A new future of work paper claimed that up to 12 million occupational transitions could be required in Europe alone as a result of AI – double the pre-pandemic rate. That is likely to bring a significant psychological toll, especially among societies where identity and self-worth are often synonymous with work, career status and progression.
In a study, Gen AI and the psychology of work, researchers Erik Hermann, Stefano Puntoni and Carey Morewedge looked at how generative AI in the workplace can “frustrate basic psychological needs for competence, autonomy and relatedness.”
“Unlike previous technologies, GenAI can demonstrate cognitive, creative and interpersonal capabilities that challenge traditional human–machine boundaries and redefine the knowledge, task, and social characteristics of work,” they wrote. They outlined the need for employers and employees to address these threats, fostering “human-centred” workplaces to balance the benefits and risks of AI.
The potential psychological impact isn’t just limited to identity, of course; with ever-greater reliance on LLMs and other forms of generative AI, are we dumbing down our skills, especially as younger generations grow up around it?
“I believe this is the most challenging question,” says Shi. “I think if Google Maps or GPS is an example, I think we could lose those skills or cognitive thinking ability. I think this is something we have to find ways to address.”
In a recent study published in The Lancet, a group of experienced doctors were given AI tools to help with diagnosis in colonoscopies. After they returned to human-only diagnosis following the three-month trial, their performance had dropped from before they had started using it. “This was a period of just a few months,” says Shi. “So if you scale that up, it is a little alarming.”
An unpredictable future
All of this is based on the assumption that AI continues to advance at its current rate, of course – and largely depends on the development of agentic AI. How advanced agentic AI really gets is a key question that even experts right now can’t fully be sure of. “So far, AI agents are still put together by humans for intended purposes,” says Shi. “It is hard to predict if they could become fully autonomous – including in mission critical situations such as military operations.”
If such agents were to become more advanced, to what level they would be trusted to make high-level decisions is another key obstacle. “AI could make decisions in critical situations such as military operations in milliseconds or nanoseconds,” says Shi. “They could arguably make fewer, less severe mistakes than humans. But what is the right way? These are fascinating ethical, legal and moral questions.
“I think there needs to be a global consensus to ensure AI isn’t used for bad outcomes,” he says. “AI could be more dangerous than nuclear weapons if there aren’t guardrails in place to ensure it is used for positive, not negative, outcomes.”
There are multiple other risks and roadblocks to overcome, of course – from bias in AI algorithms to how much trust consumers will have in AI, alongside ongoing inaccuracies and ‘hallucinations’ (or false information, disguised as accuracy) still sometimes seen in LLMs. And as with every prediction, where any of this really goes is anyone’s guess until it actually happens. “The story of humanity is one of evolution and replacement,” wrote the researchers in Gen AI and the psychology of work. “Many skills and jobs will be replaced by GenAI. Organisations and workers will adapt, as they have in response to earlier technological innovations, creating demand for and developing new skills and roles.”
While this may well be true – and while the raft of newly created jobs could go at least some way in mitigating the impact – re-training displaced workers to take them up will be a race against time if AI continues apace (and it is growing rapidly; McKinsey analysis found that AI systems could potentially complete four days of work without supervision by 2027).
AI may well bring net good and provide a major boost to economies. It could save some of the greatest challenges of our time, from climate change to cancer. But as with any major economic revolution, governments, businesses and individuals need to prepare themselves for an unpredictable ride that could lead in many different directions and to many variable outcomes.
Now is the time for tech companies, business leaders and policymakers to take control of which way it goes – for the future of not only the workforce but the very identity and humanity by which we live our day-to-day.
