Ep 730: Is AI creating a great recession for white collar workers? Inside Anthropic’s labor report

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips.

Listen daily for practical advice to boost your career, business, and everyday life.

Buying AI tools is the easy part.

Getting your employees to actually use them for anything meaningful, that's where most companies fall apart.

Sections own research shows over 90% of employees only use AI for basic tasks.

That's not ROI.

Section coaches every employee on real, role-specific use cases, tracks who's driving AI impact,

and gives you the data to prove it's working.

For more, check out section at sectionai.com.

Inthropic just published, one of the most detailed roadmaps on AI's impacts on jobs that we've ever seen.

And the results are kind of misleading, especially if you take them topically.

I think that most people are getting hung up on the biggest takeaway, that as of today, AI hasn't yet caused huge unemployment.

But if you read beneath the surface, there's a much larger and more impactful finding from anthropics recently released AI Labor Report.

It's that mass AI-induced unemployment hasn't happened yet, mainly because the majority of companies don't understand AI's capabilities.

It's not because AI isn't capable yet of automating jobs, because it is.

And there's one more finding that is going to hit the younger generation right now, and might spell trouble for companies in the future.

All right, let's get into it.

I'm excited to jump into today's topic, and here's what you're going to learn on today's show.

Well, we're going to talk about why anthropics says the AI job apocalypse isn't happening, but something worse might be happening.

Why AI is actually a threat to more highly educated white collar workers and not blue collar ones that we always thought automation and AI would come for first.

We're going to learn which five white collar jobs are already being automated the most right now.

And we're going to dig into a little bit more on the massive capability gap in AI that almost no one is talking about and how you can actually use that to your advantage.

All right, let's get into it.

If you're new here, welcome, my name's Jordan, and this is every day AI.

If you're new here, this is an unedited, unscripted daily live stream podcast and free daily newsletter helping everyday business leaders like you and me make sense of everything that's happening in the world of AI because it is non stop.

I tell you what's important.

What's not you take that information to grow your companies and your careers sounds like a good trade off, right?

It starts here, but make sure to go to the next level, go to our website at your everyday AI.com.

We break down each day's podcast as well as giving you all the other AI news that you need to know to stay ahead, right?

Like as an example, Microsoft just released their their new co work, right? Yeah, work is changing.

All right, but let's get straight into something that's been dominating the headlines a lot, right?

I actually, you know, at first saw this study and I'm like, okay, this is important.

We shared about it in our newsletter.

And my wife was actually like, I'm seeing this all on my Apple news, right?

So she's seeing it everywhere.

So like, there's a certain point when AI news starts to hit the mainstream outside of our little, you know, AI, you know, closed circle here.

I'm like, okay, this is worth diving into a little bit more.

But let's first kind of zoom out and talk about this new study and what it found from in profit.

So researchers tracked US employment data from 2016 through well, the post chat, you can see era.

And the biggest finding was there's no mass AI unemployment, right?

So AI isn't taking millions or tens of millions of jobs yet.

That's because while their findings over the last 10-ish years show that that hasn't happened.

And the unemployment estimate for highly AI exposed workers was statistically nothing, right?

So even in areas where AI has been shown capable to automate a lot of jobs.

We're not seeing, right, millions of people being laid off, right?

I think early, you know, very early on in the early generative AI, large language, mild days.

You know, there's a lot of these predictions that, you know, there's going to be tens of millions of fewer jobs fairly soon.

And well, we haven't seen that happen yet.

And inthropics study, I think was one of the best that really dug into this.

But here is one thing that this study found that is kind of concerning, especially if you are a younger person.

And I think it has larger implications, even if you are mid-career running a company.

And this is actually something I talked about last year on the show because it's something that I spotted a long time ago.

And I've been talking about it quite a bit.

But the study found that hiring of workers age 22 to 25 into AI exposed fields quietly dropped by roughly 14%.

Let's think about that percentage point right now.

Because I think when people are looking at AI and its impact on jobs and unemployment,

the thing that most people look at is the unemployment rate, which seems to be smart, right?

Because if AI is impacting millions of workers, well, then that means the net number of Americans with jobs is going to go down, correct?

Well, yes and no, right?

Because what we've seen happening and this study from anthropic showed as much, a 14% drop, right?

For the most part, if you don't follow unemployment numbers like I do, it varies by a little bit.

But for the most part, the unemployment rate in the US is between 4% to 5%,

aside from anomalies like the pandemic, the financial crisis of 2008, 2009, right?

But for the most part, for the last 25 years, the unemployment rate in the US has teetered 4% give or take, right?

So when people still see that 4%, they're like, OK, AI is not causing mass unemployment.

But what I do think it has already started to cause, especially in the younger generation, is mass under employment.

Because if we had an unemployment rate at 14%, that would be a global disaster, right?

People would be in the streets, probably rioting, right?

Hey, big AI, you took away my job, right?

But that's essentially kind of what's happening for the younger generation.

They in those highly exposed areas, at least, companies are just not hiring anymore.

I've been talking about this since I believe 2024, this process of quiet hiring,

how there's been this kind of thing, you know, as post pandemic, right?

More and more people are remote and hybrid.

There's this thing called quiet quitting, right?

Where employees, whether they're using AI, augmenting with AI or not, they're kind of doing the bare minimum and just scathing by.

I think companies have already started in the anthropic study shows this, they've kind of done this quiet hiring.

They're just no longer hiring for junior people or when people leave, right, to avoid having these.

All right, we got to cut 10,000 jobs, 20,000 jobs, right?

We heard reports Oracle might go up to 30,000.

We've seen tens of thousands from Amazon.

So for bigger companies to avoid this, well, they're just not hiring, right?

So when the silver tsunami hits and we have millions of seniors, you know, people in their 60s retiring,

well, they're just not going to refill those roles and those younger people that are in these highly exposed areas are just not going to be able to get jobs.

Also, some key findings and we're going to dig into all of these a little bit more.

But there's a massive gap, the capability gap between what AI could do versus what it's actually doing and what it's actually being used for.

And the most exposed workers right now are female, educated and higher paid.

So the great recession for white collar workers, well, that scenario, even though it hasn't happened yet, it is still very much so on the table because that is where the exposure and the risk is.

And I think with this report, Anthropic actually built an early warning system to track where disruption may show up first before any unemployment numbers can diagnose it.

So let's talk about the biggest finding.

And that is the massive gap, right?

And this kind of chart obviously went, you know, very, very viral online, right?

Whether you're reading Twitter, LinkedIn or the news, you probably saw this chart.

So for a podcast audience, you can always, you know, check this out, obviously in the newsletter, but you can go watch the video version of this on at your everyday AI.com.

And nothing we can't describe here, but essentially this chart showed the different occupational categories, right?

So everything from management, business and finance, legal, health care, food and services, personal care, office admin, right?

All these different sectors of work.

And then you had a theoretical AI coverage line, right, which is in blue.

And then kind of the spider graph that shoots out.

And if the AI could theoretically do a hundred percent of the job, then it goes all the way to, well, the exterior of this circle.

So in the blue, you have your theoretical AI coverage, right?

And then you have your observed AI coverage in the red.

All right, we're going to talk a little bit more how infropic got to that.

Essentially, it was a combination of U.S. employment data and millions of anonymized chats with infropic's clawed chatbot, right?

So what you essentially see is, well, AI is theoretically extremely capable in many areas.

And right now, just not have a lot of capabilities in others, right?

So some of the biggest areas where in theory, right, AI could do, you know, 80 to 90% of the work comes in fields like management, business and finance, computer and math, legal arts and media, right?

Those are areas where there's at least 80% coverage up to the mid 90s.

And then there's areas, at least right now, that large language model in their current capabilities, well, they don't really touch, right?

Like sectors like production, installation of repair, construction, agriculture, right?

Those jobs that, for the most part, require you to use your hands away from a computer for the majority of the time that you're working.

And that's kind of how we got to this capabilities gap.

But when we talk about the theoretical AI coverage and the observed AI coverage throughout the rest of the show, this is essentially what we are talking about.

It is what AI can actually do, right?

According to benchmarks and AI's actual capabilities.

And then the observed AI coverage, which is what inthropic found through millions of anonymized chats, well, what people are actually using it for.

And even in those areas, right, management, business and finance, computer and math, nothing, nothing hit the 40%.

Right? In the majority of those, even with high theoretical AI coverage, right?

Where, hey, technically AI could do 90% of this right now out of the box with nothing else happening, right?

A lot of those areas, right? Like management, legal, right?

The actual observed AI coverage was low.

It was less than 20% in many instances.

So let's talk about one of the most obvious categories and that's in computer and math roles.

So the observed capability was 90, or sorry, the theoretical capability was 94%.

So inthropic found by matching it with US jobs data that AI's capabilities right now can do 94% of tasks.

But it's only being 33% observed.

And that is of all the different columns that is actually the highest observed AI coverage.

So maybe you're thinking, oh, well, yeah, people are just using AI for different things that maybe aren't falling on this map.

Absolutely not.

That is the highest observed AI coverage.

And the gap there is still enormous, right?

I've been talking about this capability gap.

I talked about it a lot on our 2026 AI and roadmap series about this huge gap in that people, especially I think since quarter four of 2025,

they still are looking at AI like a fun little chatbot, not realizing it's a genetic nature, the improved scaffolding and harnessing.

Well, I can probably do the majority of your work and you just don't know it.

So not only is there capability gap that comes from training.

Well, there's also just the education side.

People don't even know.

I think a lot of people understand that they maybe don't, you know, can't keep behind a computer and, you know, fully use a chatbot like Claude or chatTPT

or Gemini or co-pilot to its fullest capabilities.

But I think the majority of people, even those that follow the technology fairly closely, don't even understand what those capabilities actually are, right?

So the 61 point divide in the computer and math roles as an example, that just defines the current state of AI at work.

Here's kind of my hot take. I kind of already talked about it a little bit.

See, it's Tuesday, right? We're not doing as many hot take Tuesdays, but I'm going to go ahead and throw my hot take opinion in here.

This is going to happen. We are going to see the great white collar work recession.

It is a lagging factor, right?

That 30, what was it? The 33% observed coverage in math and computer right now, that's going to go up.

It's going to go up. It's going to go to 40, 50, 60 in the coming months and quarters.

The same thing with these other areas, right? I talked about this on the prediction of roadmap series, but I think especially in anything that cost people a lot of money, right?

Legal management, business and finance, that gap is going to shrink.

I think we're going to start to see measurable shrink in 2026, but I think by 2027 that gap is going to close very quickly.

I think that's how long it takes for the average, you know, US company.

It takes nine to 18 months for companies to truly number one, realize that there's a gap.

I think studies like this one from infropic that put it into concrete terms, obviously help executives in boardrooms understand that, oh, wait, there is a gap, right?

And you can see the methodology, which we're going to talk about here in a second, but then it takes them time to start to learn to close the gap.

That is, unless you and your company listen to the show every single day, because I've been talking about this gap before infropic or anyone else, you know, officially identified it.

Granted, I didn't have millions of anonymized, anthropic chats to really bring teeth into it, but this is something I've been observing.

Well, for the last three years since I've been doing this show.

Here's a harsh truth. Your company is probably spending thousands or millions of dollars on AI tools, they're being massively underutilized.

Half of companies have AI tools, but only 12% use them for business value. Most employees are still just using AI to summarize meeting notes.

If you're the one responsible for AI adoption at your company, you need section. Section is a platform that helps you manage AI transformation across your entire organization.

It coaches employees on real use cases, tracks who's using AI for business impact and shows you exactly where AI is and isn't creating value.

The result you go from rolling out tools to driving measurable AI value. Your employees move from meeting summaries to solving actual business problems and you can prove the ROI.

Stop guessing if your AI investment is working. Check out section at section AI dot com. That's S E C T I O N AI dot com.

College graduates. Here's the problem. They are getting squeezed out. Right. If you graduated between 20 late 2024 and 2026 and you have a full time job in your area of study, consider yourself very lucky because you are in the minority.

Right. And I'm talking about it on a couple of their shows. Right. The majority. And I think one of the reasons why we haven't even seen a huge technical unemployment spike in the younger generation is well because they're having to take jobs outside of their college major.

We talked about that in one of our, you know, AI is impact on college shows in 2025 that a highest, the highest number ever of graduates are having to take jobs full time rolls outside of their field of study.

Well, because no one's hiring, right. There is this quiet hiring that's going on. And then not only that when we do see the silver tsunami, the baby boomers that are going to be retiring in mass, right.

One of the biggest problems is, well, what's happening with this junior generation, right. If, if companies are hiring fewer and fewer junior researchers, junior analysts, because all that's happening is the senior people who are sticking around and maybe aren't getting laid off in mass.

Well, they're augmenting their job with AI. So they don't need as many junior people. So well, where are the future middle management. Where are they now. And I think that's actually pretty problematic.

All right. So let's quickly talk about how anthropic got to these conclusions, because again, I think this is one of the best studies looking at the kind of capabilities gap that we've seen.

So they use the federal, the US federal own database. And that breaks the essentially 800 US occupations into 20,000 plus specific tasks. Right. So this isn't, you know, looking at broadly, right.

We're guessing and throwing things at the dartboard. This is 20,000 specific day to day tasks that the average American performs in those different sectors. Right.

So this is the US census employment survey that tracks who is currently working unemployed or entering the workforce was also used by anthropic. And then the big piece here, which I already said, is millions of conversations anonymized.

They're able to then tie up or map to those 20,000 specific tests. And that's what kind of gave them the ability to look at AI's current theoretical capabilities and then the observed right.

Right. This isn't guessing. This isn't a vibe. Right. MIT. Right. This isn't actual legitimate study that shows this huge capability gap when it comes to real world day to day tasks.

And a little bit more on how they actually did it. So they assigned each kind of task a score. So a score of one meant that AI alone could double a worker speed without any extra tools, a score of 0.5 meant that AI could double the speed but required outside tools like browsers or databases.

So more augmented and then a score of zero meant that AI could not meaningfully help. Right. So through this scoring system and matching it up.

Researcher scored every individual task across those 800 US occupations using that 1.5 or zero scale. Then they check those theoretical scores against real caught usage data to see which tasks workers are actually delegating to AI. Right.

Here's the interesting part. They found that 68% of real usage landed on task scored one. So the overwhelming majority of people that were using Claude were using it for task that AI was actually really good at and was able to fully automate.

And only 29% of usage was on task that scored 0.5 and 3% on task that scored a zero, which is interesting that 3% of people were still trying to get a chat box to do something that it theoretically did not have the capabilities to do.

So the scoring acted as a baseline prediction in the real usage data acted as that reality check. And that's where you got the huge gap between the two and how that has turned into the core of the study.

So let's look at the most exposed roles. So computer pro and this is those roles where well AI has the highest capability computer programmers led with 74.5%.

And then routing out the top five right the top five most exposed current roles. Well, one was computer programmers. Then you had customer service data entry medical records and marketing analysts. So essentially any those roles are ones where you're constantly in front of a computer.

You're doing things that require a screen. They're text heavy and repeatable.

So this is why it's hitting that group of people who are higher paid and more educated sitting in front of a desk. That's because these jobs are automated through company API systems, not just individuals chatting with a chatbot. Right. So they can be done and automated with AI at scale.

And the other thing right and you have to look at the whole corporate greed thing right the most exposed workers earn 47% more per hour than those with zero AI exposure. So in theory, the jobs right now by today's technology that are most replaceable and automatable by AI and haven't been discovered yet are those that well are expensive.

And but physical jobs obviously like mechanics lifeguards cooks showed zero exposure.

The other thing it is that young generation right because the study found that senior workers can stay productive with AI well companies well they just have stopped backfilling junior positions almost entirely right and that 14% hiring drop not only appears for workers.

Age 22 to 25 but it's just for them right so it's not like there's a 14% hiring gap for any person or any group of people that had that high exposure.

It was just the younger generation right so what that meant is people with experience senior people who had been at companies for a while right they're not getting they're not going through that 14% hiring drop

because companies still want people with experience so instead they're just not hiring any entry level people.

So that doesn't change the fact that AI induced layoffs are still happening in mass.

So in 2026 already we've seen thousands of AI linked cuts hit big companies like block Amazon meta auto desk in sales force and then in late 2025.

Same thing huge I mean we're talking in the tens of thousands for each of these companies in late 2025 attributed to AI Accenture city group Dell IBM Microsoft right reportedly Oracle looking at up to 30,000 jobs also Intel.

UPS IBM right the AI induced mass layoffs are still happening in they will I believe continue to ramp up probably toward the latter end of 2026 and early 2027 as companies understand that gap.

Okay and that gap right now it is an unclaimed territory it's not a looming threat right so I think that workers and company might might might my big takeaway here right.

AI is still going to come for jobs I've been saying since the literal very first episode of every day AI that yes AI will create millions of jobs that we don't even know exists yet but I think ultimately.

AI is well it's going to change the the face of traditional full time employment I think it will ultimately you know take away more full time roles than it will ultimately create.

But right now there is still an opportunity right so I don't want you know my message might take away my hot take on this to be one of doom and gloom because I understand that many of.

Our listeners and viewers of every day AI right myself included I have a graduate degree right technically highly educated sitting in front of a computer doing a lot of these tasks that are highly automated by AI right this capability is gap.

You need to attack it right because.

Yes you could say this is a looming threat but right now I do think companies departments individuals have a nine to 18 month.

Window where you can do something about it right if you can close that gap first in your organization in your industry if you can be first to market that truly.

Right unlearns and rebuilds being AI native in some of those areas and you can exploit it with revenue tied services you will own the advantage in the advantage is out.

We saw this with the entropic study tying real world jobs data real world 20,000 different tasks all mapped out through millions of anonymized claw chats the opportunity is there if you go season.

That's it for this episode I hope it was helpful but big takeaway here is AI creating a great recession for white colored workers well even though entropic study says no today it is definitely coming but if you are listening to this show every day if you're putting what we teach into practice you can stay ahead and not be impacted and well hopefully you can actually capitalize and take advantage.

So if this show was helpful make sure you also go check out episodes 712 and 713 are 2026 AI prediction and roadmap series and then go to our website at your every day AI dot com sign up for the free day newsletter thank you for tuning in hope to see you back tomorrow and every day for more every day AI thanks y'all.

If you're leading AI at your company and your employees are barely scratching the surface of what AI can actually do you need a better plan section coaches employees on real role specific use cases tracks adoption across your entire organization and helps you prove the ROI to your CEO your board or whoever's asking that's the job section helps you do it.

Check out more on section at section AI dot com and that's a wrap for today's edition of every day AI thanks for joining us if you enjoyed this episode please subscribe and leave us a rating it helps keep us going for a little more AI magic visit your every day AI dot com and sign up to our daily newsletter so you don't get left behind go break some barriers and we'll see you next time.