Why we believe AI reshapes work more so than it reduces overall payrolls

Hello, and welcome to Global Research Unlocked, where we discuss what's rising from both industries to rising risks and opportunities in global markets. I'm TJ Thornton, Head of Product Marketing at B of A Global Research, and we're recording this episode on Wednesday, May 6th, 2026. In the U.S., it is also true that the unemployment rate among the 20 to 24-year-old has been rosing roughly a decade high in the second half of last year, but recent events suggesting that the pressures on that cohort may have also kicked.

So if I were to say that a more accurate interpretation is that AI is interacting with the demographics, it is weaker hiring environment, as well as changing value of the entry-level jobs. It's perhaps a fairly obvious conclusion that AI is likely to displace a significant amount of jobs, thus leaving a lot of people out of work. But the obvious conclusion is often the wrong one, and so many times in history, from technology revolutions to pandemics, the dire prognostications have been incorrect.

B of A Global Research's global economics team view on this topic is somewhat more encouraging. As part of their AI series, the team recently published a note on implications for labor from AI, and today we'll talk to the authors of that piece. We're joined by Benson Wu, part of the Asia economics team and the head of Korea economic research, and Nick Stenner, head of economic research for Australia and New Zealand.

Thank you both for joining us today. So in the note, you make the point that the apocalyptic views on job destruction from AI are unlikely to materialize. So is the view that we'll see many jobs destroyed, but that new jobs will take their place?

Thanks TJ for the question. So I would frame it slightly differently. So our view is not simply that AI will destroy many of the jobs, and then new job will mechanically replace them.

The more precise point is that AI is likely to destroy or transform the tasks, rather than eliminate the entire occupation at scale. So most jobs are a bundle of the tasks. Some tasks are routine, codifiable, and information heavy, and those are clearly exposed to AI.

But many jobs also involve judgments, accountability, client interaction, coordination, creativity, or contextual understanding. Such human elements are much harder to automate fully, even with agentic AI. So in many occupations, the jobs may survive, but the task mix will change quite meaningfully.

We have cited the number calculated by the ILO, the International Labour Organization, of which about one quarter of the jobs globally are exposed to generative AI. But exposure does not mean disappearance. It means AI can enter the workflow.

So in fact, our report is showing that the share of jobs with augmentation potential is about 13%, and that is significantly larger than the share with the high automation potential of only 2.3%. So yes, there will be disruption. Some roles will shrink, especially in the collaborative support, customer service, admin work, data entry, and also some of the ICT functions.

But for many of the roles, the bigger story is a redesign. So workers may spend less time drafting, summarizing, searching, processing data, or doing repeated analysis, and more time they will have on interpretation, on decision-making, on communication, and also strategy. We like to call this interesting number, and that is also on the did-you-know part of our report as well, that 60% of the U.S. jobs today are in occupations that did not exist in the 1940s.

Of course, there will be new jobs to be created and also transformed. Okay, and I wanted to follow up on that last point. I get, as you're saying, it's not just a function of a bunch of jobs being destroyed and then being replaced by other jobs.

There will be transformation in the existing jobs, but there also will be new jobs, and you made the point throughout that piece that we've seen that sort of thing in past transformations. What sort of new jobs are you expecting to be created? What jobs that may exist currently are likely to expand?

Yeah, I think there are three or four categories to think about. The first is certainly very obvious. The new AI specifies the roles.

Some of these are already visible. So do we see the new hiring on the prompt engineers, AI trainers, or model evaluators? So these are the roles that emerge because of AI systems themselves need human oversight, need the quality control, or the domain exercise, and also the ethical judgment.

The second category is much larger. We see the hybrid roles inside existing occupations. For example, as a research analyst like us, one may become less focused on mechanical data collection going forward, but more focused on integrating AI-generated analysis, challenging assumptions, communication views to clients, and also taking responsibility for conclusions.

A lawyer may also use AI for document review, but spend more time on legal strategies and also negotiation. This is also what we believe the augmentation is happening. The third category is jobs that expanded because of the AI rises, the overall income, and also productivity elsewhere in the economy.

So let's say that if AI makes part of the economy more productive, the real income overall will be increasing, demand will shift towards services that are still human intensive. So in a report that we highlight in areas such as healthcare, nursing, elderly care, education, personal services that are other jobs centered on trust and on experience, on care, and also human interaction. So these sectors are not necessarily the most directly exposed to AI, but they can benefit indirectly from a higher income and also changing the consumption patterns.

There are also the rise of the OPC, the one-person company. So the agentic AI could allow individuals to perform functions that previously required a small team that could lower the barrier to entry for the entrepreneurship and also the small business formation. So even if there's some corporate roles facing pressures, AI may also create a new self-employment and also the business formation opportunities.

So the new job story is not only about the AI engineer roles, it's broader and also consists of the AI specialists, the hybrid professionals, the human-centered service workers, as well as the AI-enabled entrepreneurs. The difficult part is certainly many of these jobs are hard to predict in advance. Okay, thanks, Benson.

So this next one is for Nick. As you say, it's hard to see into the future. Sometimes it's helpful to look at the past.

Is there a particular historical parallel that you think is most appropriate to look at for cues as to what may happen with the labor market in light of AI? So there's obviously no perfect historical parallel because AI directly affects cognitive and white-collar occupations in a way that earlier technologies didn't. But there are several useful comparisons we touch on in the report, and the broadest is probably looking to earlier general-purpose technologies, whether we think about the Industrial Revolution, computers, or the rise of the internet.

Each of these triggered fears of mass technological unemployment, and they certainly displaced tasks and some occupations. But each also ultimately raised productivity, they lowered costs, they created new sources of demand and generated jobs that were very difficult to imagine beforehand. So when we think of the broad sectoral shifts, one that stands out is agriculture.

In the early 1900s, agriculture accounted for roughly 40% of total US employment, whereas today it's only around 1%. So that's a massive sectoral employment collapse, but it did not mean collapse of total employment in the US labor market. Labor moved into manufacturing and services sectors.

So there's lots of examples, and another one we touch on is the IT revolution. So the IT revolution increased demand for information analysts, as the cost of data gathering declined. Excel is another interesting case, because it automated parts of spreadsheet work, but it didn't simply eliminate accountants, it instead raised productivity and expanded the range of analysis that firms could do.

E-commerce didn't eliminate total US retail employment, the US still has roughly 15 to 16 million retail workers, which is around the same level as in the 1990s before e-commerce became mainstream. So the most appropriate lesson that we take from history is probably not that any single technology is going to displace labor, but from the pattern across technologies, we find that automation displaces tasks, productivity expands demand, and these new roles emerge endogenously. I would say, though, that AI may move faster and affect a broader range of cognitive tasks.

So history gives us confidence that work will not disappear, but it shouldn't necessarily make us complacent. The adjustment could be faster, more uneven, and more concentrated among white-collar work and early career roles. Got it, okay.

And I can vouch for the fact that even 11-year-olds, because I have one at home, who have kind of grown up with tech, love to go to the mall, which is something you might have not expected 20 years ago at the dawn of e-commerce. But a question about advanced economies, so you talked about that in the note, made the point that advanced economies, including the US and Europe, tend to have the largest percentage of jobs that could be disrupted by AI, but that these economies are some of the better positioned when it comes to the sort of threat or opportunity from AI. And so why is that, you know, despite the fact that it would seem that they have a lot of roles that could be negatively impacted?

Yeah, this is a really important point, and it goes to a key theme of our report, that labor market adaptability is the new job security. So we find that advanced economies are more exposed because they have more white-collar cognitive and digitally-enabled jobs. So we cite from the report ILO work that shows high-income economies have around a third of jobs exposed to generative AI, and that compares with only around 11% in low-income economies.

Europe and Central Asia have the highest regional exposure rate, followed by the Americas. But it's really important to note that exposure is only one side of the story, and the other side of the story is adaption capacity. So advanced economies generally have higher levels of human capital, more advanced education systems, deeper digital infrastructure, and more flexible labor markets.

That ultimately means these advanced economies are better positioned to convert exposure to AI and to productivity gains, rather than only suffer the disruption and the kind of negative side effects. We also touch on an important demographic point, which is to say that aging economies, so when we think of, say, Europe, North Asia, and parts of North America, these economies face shrinking workforces and rising aged care needs. So for them, AI can help offset labor shortages by boosting output per worker and automating some routine tasks.

So for investors, the point is that high AI exposure is not automatically negative. It can mean high disruption risk, but also high productivity upside. The key differentiator is whether a country has the skills, infrastructure, and the policy frameworks to absorb the shock and adapt appropriately.

Okay, and this next one is for Benson. So this note came out about a week ago. What's been the feedback from clients?

Is there agreement with the points that you have made, or do you find yourselves are generally in the minority? Sorry, let me pass that one again. So this next one is for Benson.

You published a note about a week ago. What's been the feedback from clients? Is there agreement with you, or do you find yourselves in the minority when it comes to the views on AI?

The feedback has been quite engaged because AI in the labor markets are clearly top of mind of the investors. I would say that clients generally agree with the starting point that the AI exposures is large and also the white-collar jobs are more directly affected than in previous automation waves. Where the debate becomes more interesting is around the interpretation of that exposure.

So some clients are more concerned about the automation risk is underestimated or under-predicted, especially with the agentic AI. So their point is that early generative AI was more of a tour, but agentic AI that are happening in starting from 2024, 2005, can increasingly take actions, coordinate workflows, and also potentially automate larger parts of the production chains. I'm seeing that some clients are much more worried than us, especially when they saw their investor companies starting to lay off employees due to AI impacts, and also AI continues to evolve in a faster speed.

So that is, I think, one of the debates that we have. The other sort of debate or pushback we have is on the job creation estimation. Some investors are thinking that if capital owners are getting more of the wealth going forward, only a minority of groups of people can be able to pay for the extra services.

So hence, the new job creation would be less than we expected. That could be a fair concern as well, given government is doing nothing and the wealth are unevenly distributed to the capital owners. But we do believe in the government well stepping up early and to be more active.

So that is exactly what we saw from what we see in Europe, in China, and the US as well. So if government can play their role and to allow more even distribution of the productivity gains of the economy, we are in high belief that the job market can come through with the transitions. Yeah.

Understood. OK. And this next question is somewhat related.

I mean, we've mostly focused on just the sort of sheer number of jobs. You know, some of them are going to transition. Some of them will be replaced by other jobs.

But there could be many other effects from AI on economies and workers. Wages, hours worked, could fall, that would be nice. Perhaps capital will benefit more than labor interest rates.

So what are your take on some of the secondary effects? And this question is for Nick. Yeah, so this is where the second round effects and spillovers become really important.

So on wages, evidence so far appears to be mixed, but not yet really dramatic at the aggregate level. So we cite work from the Dallas Fed showing no meaningful relationship between occupations, AI exposure, and wage growth in the post-2022 period across just over 200 US occupations. And a range of other papers that generally show we're not yet seeing broad-based wage declines in AI-exposed occupations.

But within occupations, the story could be different. I mean, AI can increase the productivity of experienced workers who know how to use it effectively while reducing demand for entry-level or routine tasks. So that may widen wage dispersion within the same occupation.

And in the report, we note that wages have risen faster in AI-exposed industries, even as employment growth has lagged, particularly for younger workers. So this suggests a combination of AI augmentation for experienced workers and substitution pressure at the job entry level. On hours worked, we don't make a precise forecast, but conceptually AI could reduce the time needed for routine tasks.

And the key question, though, is whether that time saving translates into fewer hours, more output, or redeployment into higher value tasks. So historically, productivity gains often lead firms to expand output rather than simply reduce time working. The bigger issue, in my view, which you touch on, TJ, is the distribution between capital and labour.

So we highlight that early evidence points to productivity gains flowing disproportionately to capital owners and firms with strong complementary assets, be that data, infrastructure, and broader organisational capability. And if AI mainly automates existing tasks without creating enough new labour-intensive tasks, productivity gains may ultimately raise the capital share of income while leaving labour's share flat or lower over time. So a key risk from AI is not just unemployment, it's that AI raises productivity, but the gains accrue unevenly.

They accrue to leading firms, capital owners, and only highly skilled workers, which is why policy around training, labour mobility, wage insurance, and tax design all becomes really central to the policy debate around AI. Okay, thanks, Nick. Another topic I'm sure you heard feedback on this one, too, and I know you considered it when you put the piece together, is youth unemployment.

It's been on the rise, at least over the past year or so. It's also viewed to be a difficult job market for new graduates. What do you make of those dynamics?

Some people have suggested that that's because of AI, but do you see this broadly, this weakness in youth employment, and do you think it has anything to do with the technology? Yeah, so I think it's an important signal, but we should be careful not to attribute everything to AI. It is very true that in the high exposure sectors, AI has reduced demand for routine entry-level work that traditionally helps young workers building their experiences.

We have cited our research as well, showing that the employment around the early career US workers aged 22 to 25 in the AI-exposed occupations has declined significantly. However, we do not think youth unemployment should be interpreted as the pure AI displacement. I cover China as well, so taking China as one example.

The higher youth unemployment in that region was due to several structural forces, including the demographics that we see almost doubling the new graduates nowadays versus 10 years ago, as well as the subtle labor demand due to the property or some of the other sectors slowing down, rather than the AI displacement alone. Okay, last question. This one is for Nick.

Interest rates. We've been in this period now, at least the US, where rates have been stuck in a range for about four years. There's hope that they may come down, and then often those hopes are dashed by something or other.

Obviously, more recently, it's been what's going on in the Middle East. But there's a view that maybe productivity benefits, some of which could come from AI, might allow the Fed and other central banks to cut rates and look through these temporary inflation spikes. Do you agree with this, and what sort of impact do you think AI could have on long-term rates in the US and elsewhere?

So we don't make an explicit Fed call or rates forecast in the report, but we do discuss the macro channels. So in principle, AI-driven productivity growth should be disinflationary over time, but I think this is a longer-run story. Ultimately, if firms can produce more output with the same labour input, unit labour costs could fall, margins improve, and supply capacity in the economy can expand.

So that could give central banks a bit more room to look through temporary inflation shocks, and that may lower the inflation component embedded in longer-term rates, but also going in the other direction, boosting productivity may push up the neutral rate. So the policy trade-off is not immediately clear to me. There are also several caveats that I'd note.

First is that productivity gains from AI may take time to show up in the macro data. So in the report, we argue that firms initially layer AI tools onto existing workflows, and the larger productivity gains come when firms redesign roles, teams, and broader processes around AI. So the near-term inflation and rates impact may be very limited.

Secondly, the gains are likely to be uneven. If productivity gains accrue mainly to a small group of leading firms with strong data and infrastructure, the aggregate macro effect may be smaller than what we're seeing at the firm level. We also warn that gains may flow more to capital than labour, which matters for demand inequality and also policy.

Third is that AI can raise some costs in the transition. So AI-driven labour market pressures could lead to a rise in fiscal spending over the medium term, which would obviously matter for long-term rates if it adds to fiscal commitments and bond supply. So AI is potentially a positive supply shock that could be disinflationary over time, but this isn't automatic, and ultimately, if AI does raise productivity, while it could help central banks tolerate some temporary inflation spikes, it also would likely push up mutual rates, leading to a difficult policy trade-off for central banks.

Okay, Benson and Nick, thanks again for an excellent discussion on the back of what's been a great piece, and I'm sure there's going to be a lot more to talk about on this topic over the coming year and beyond. So we'll look forward to having you back. Yep, thank you.

Thanks for joining. Our global economic team's fairly sanguine view on AI disruption is driven by a few things. Yes, some jobs will be made obsolete, and many of these workers will probably find work doing different jobs.

And many jobs that can currently put AI to use will change such that people spend less time on certain tasks that AI can do well, and more on other, perhaps higher-value-added tasks. Also, better productivity and income should increase demand for jobs in fields like healthcare and education. And if global investment shifts to certain economies because of AI, that could benefit jobs in those economies.

Nick had numerous examples of past technology shifts that may have seemed to automate jobs, but actually increased demand for the people who use these tools. Excel spreadsheets, for example. Speaking of which, it's interesting that while youth unemployment continues to get a lot of focus, unemployment levels for that cohort are actually off the highs seen last year.

Thanks for joining. Bank of America and B of A Securities are the marketing names for the global banking businesses and global markets businesses, which includes B of A Global Research of Bank of America Corporation, Lending Derivatives and other commercial banking activities are performed globally by banking affiliates of Bank of America Corporation, including Bank of America N.A., member FDIC, Securities Trading Research Strategic Advisory and other investment banking and markets activities are performed globally by affiliates of Bank of America Corporation, including in the United States, B of A Securities Inc., a registered broker dealer and member of FINRA and SIPC and in other jurisdictions by locally registered entities. Copyright 2026 Bank of America Corporation, all rights reserved.