AI Slop, Supply & the Search Process: Ben Zweig Breaks Down the Data

Is AI coming for everyone’s jobs, or are we all just wildly overreacting while spamming hiring managers with bot-generated resumes? On this episode of The Chad & Cheese Podcast, hosts Chad Sowash and Joel Cheeseman sit down with Ben Zweig, CEO of Revelio Labs, NYU Stern professor, and sleepless father of three to separate panic from reality.  

Ben breaks down Revelio’s AI Labor Market Tracker, revealing why computer science enrollments are dropping, why AI adopters are actually increasing headcount, and how automated "job-slop" applications are breaking the hiring process for real humans.  

In this episode, they cover:

• The CS Panic: Why college kids are abandoning computer science majors out of premature automation fear and why they might be wrong.  

  
  

• The "Slop-pocalypse": How candidates using AI auto-apply tools are signal-jamming recruiters and ruining job boards.  

  
  

• The Hiring Paradox: Why intensive AI adoption is driving a 27% increase in hiring rather than mass layoffs.  

  
  

• The Zweig Paradox: Why 93% of labor market shifts aren't about disappearing roles, but how jobs quietly transform from within.  

  
  

Are we heading toward a post-work utopia of government-funded chocolates, or are we just going to build more data centers?

ENJOY! 

PODCAST TRANSCRIPTION

Joel Cheesman (00:24.078)

yeah. It's the podcast your starting quarterback warns you about. It's the Chad and Cheese podcast, everybody. I'm your co-host, Joel Cheeseman. Joined as always, Chad Soash is in the house as we welcome Ben Zwig to the show. He's CEO at Ravelio Labs. He's also an economist and adjunct professor at NYU's Stern School of Business. He's an IBM alum and frankly putting all that other shit to shame.

Chad Sowash (00:27.517)

Yeah.

Chad Sowash (00:35.399)

Hello.

Joel Cheesman (00:53.826)

He's the host of the podcast, The Economics of Work Podcast. Ben, welcome to HR's Most Dangerous Podcast.

Chad Sowash (00:57.757)

What?

Nice.

Ben Zweig (01:00.929)

Thanks guys. Happy to be here.

Joel Cheesman (01:02.808)

Good to have you. Good to have you. So that's that's some impressive stuff, but we always want to know who the person is that we're talking to. So let's give the listeners and viewers a little bit of Twitter bio about what makes what what makes Ben tick.

Ben Zweig (01:18.101)

Okay. let's see. I mean a lot l lot of things a lot of things powering this ticker. I think the economics stuff is is pretty central. So I still kind of think like an economist, feel like an economist. So that really like, you know, was is a core part of how I think and feel. lifelong New Yorker, just moved out to the burbs and it still pains me every day. got three, three little girls. so you know that's that's keeping things exciting.

Joel Cheesman (01:24.398)

So many layers.

Joel Cheesman (01:47.662)

wow, what are the ages?

Chad Sowash (01:49.535)

girl dad.

Ben Zweig (01:51.179)

Yeah, five, two and a half, and seven weeks.

Joel Cheesman (01:54.351)

you're b so you're sleepless in the suburbs of New York?

Chad Sowash (01:56.06)

Wow, dude.

Ben Zweig (01:59.409)

I am in the thick of it. So if I start drooling, you know, don't take it personally. Yeah.

Chad Sowash (02:01.607)

Please.

Joel Cheesman (02:02.494)

man. Man, we're gonna take it easy on you, my friend, cause we both know two seven. Woo. man. That's brutal.

Chad Sowash (02:07.239)

But think about think think about that spread. Think about that spread when they start moving into adolescence. my God. Just it'd be yeah, that's gonna be fun. That's gonna be fun, Ben. You're gonna like that.

Ben Zweig (02:17.788)

Yeah, people have been warning me about that. I mean, I I feel like in some way I'm like, how hard could it be? Like plenty of bozos do it.

Joel Cheesman (02:24.054)

Chad Sowash (02:25.597)

Yeah. Yeah.

Ben Zweig (02:26.112)

But I I also thought that about, you know, being a parent to like infants and toddlers and it's it's really hard. Yeah.

Joel Cheesman (02:32.802)

So suburbs of New York, what are we talking? We talking deep sub we're talking long li is it a borough? Is it deep suburbs? Like Connecticut. okay. I love that a whole different state is the suburbs of a different state of a different city. I love how New York thinks of that.

Ben Zweig (02:38.354)

It's it's Stanford. Stanford, Connecticut. Yeah. So

Chad Sowash (02:41.954)

okay. The

Ben Zweig (02:47.396)

yeah. I mean, you know, New Jersey is you know, North Jersey is definite suburbs. And Connecticut kinda kinda makes it if you're like commutable. So it's like a forty-five minute train. So we we can count it.

Chad Sowash (02:47.49)

yeah.

Chad Sowash (02:51.729)

Yeah.

Joel Cheesman (02:58.688)

Okay. Okay.

Chad Sowash (02:59.641)

So question, i y you were a quantitative strategist at an emerging markets hedge fund hedge fund. You are a quant?

Ben Zweig (03:09.642)

I was a quant. Yeah. Yeah. Now now people don't use the term quant anymore. Like everyone just calls themselves a data scientist.

Chad Sowash (03:14.139)

Really?

Joel Cheesman (03:15.68)

Is that discriminatory? Is that sort of racist now? I don't

Ben Zweig (03:20.39)

No, it's funny. I mean Yeah.

Chad Sowash (03:20.699)

No it has anything to do with race.

Joel Cheesman (03:23.36)

I was thinking the movie The Big Short, where Ryan Gosling's like, My quant, look at him. He's Chinese or whatever it is in the movie. So I didn't know if Quant was, you know billions, yeah.

Chad Sowash (03:30.727)

The thinking of billions, but yeah, go ahead.

Ben Zweig (03:33.181)

Yeah, no, I mean I think I think quant for for a while really meant like someone cre you know creating options strategy like like someone deep into like the structure of like derivatives and now it really means someone who analyzes a lot of data for trading strategy. So I think you know quantitative finance has become much more data oriented. So now everyone just calls themselves a data scientist. And you know, they they get a lot of people from big tech to to work at these funds now.

Joel Cheesman (03:42.403)

Mm.

Chad Sowash (03:43.378)

Mm-hmm.

Chad Sowash (04:00.016)

Gotcha, gotcha. I'm a quant. I'm a quant.

Joel Cheesman (04:00.3)

And it's a really cool pickup line for chicks in Manhattan when you say I'm a Yeah, guant or an economist. Yeah.

Ben Zweig (04:06.14)

Yeah, yeah. Well, d didn't work for me at the time, so you know

Joel Cheesman (04:11.054)

It's working now where it worked out. Three kids.

Ben Zweig (04:13.046)

Yeah.

Chad Sowash (04:13.501)

Well, I yeah, no kidding. I I tell you what's working is this AI labor market tracker thingy. Is it's that that's the technical term, that you guys came out with and and we're looking at the the July numbers. Can you give us kind of like an overview of why why'd you come out with this in the first place? What is it? And what happened in in the your your latest report?

Ben Zweig (04:37.192)

Yeah, okay, so a lot to unpack there. I mean, so first first thing to to mention, I mean, we're we're a labor market data company. So we're we're tracking all this employment data, and you know, a big question that everyone's asking is how is AI affecting the labor market? And we felt like, okay, we have a lot of data to to weigh in on that question, which is an important question today. And, you know, we wanted to create something that was specifically not speculative on what might happen in the future.

Chad Sowash (04:42.973)

Mm-hmm.

Ben Zweig (05:04.618)

We wanted to track what was happening and identifiable today, like right now, and then continue to track that and you know weigh in on like, you know, use that as evidence for what might happen in the future. But this is specifically like not speculative. It is, it is descriptive in in what's happening today. So the way we the way we thought about thought about it, and the way I generally like as an economist, a lot of people ask, you know, ask ask economists, what is happening with this market?

And they're like, yeah, right, right. Most of the time it's I have no clue. But if I feel pressed for an answer, I I try to break it down into like four questions. I try to say like, all right, what's happening to supply? What's happening to demand? And of course, through supply and demand, you get equilibrium quantities and prices. So that's kind of derivative from supply and demand. So those are, you know, you have supply and you have demand. And then you have what's happening to the nature of what's exchanged? So is the actual good changing? And also what's happening to the method of that exchange.

Chad Sowash (05:34.479)

Usually say, I have no clue.

Ben Zweig (06:03.06)

So, what is happening to the search and match process that that enables that exchange? So, those are like the the parts of a market that I think are worth analyzing. And so, you know, we we follow that same structure in this report. So we have a section on supply factors, demand, quantities, prices, the content of jobs, and the the search and match process of finding jobs, you know, employer and employee matching.

Joel Cheesman (06:28.024)

So a customer hires you for what? Like the data to do what? Like forecast jobs and t explain to me sort of like why someone would would hire you or why the why the information would be of advantage to a customer.

Ben Zweig (06:40.318)

Yeah, I mean so so most of our customers are are in HR. They're they're in corporate HR. So we so you know, a lot of sometimes in TA, sometimes in like talent intelligence, workforce planning, people analytics, compensation benchmarking, employee experience, like anything where anything that we would consider like analytical HR. So it so parts of HR where it's useful to kind of like have a handle on what's going on and be able to benchmark versus peers. So we get a lot of yeah, research.

Joel Cheesman (07:07.352)

So so give us give us the state of of that HR customer. I I'm I would guess they're really confused about what's going on and you have answers. Yeah.

Chad Sowash (07:07.473)

Market research.

Ben Zweig (07:16.412)

Definitely confused. They don't have a lot of data, or or even worse, they do have a lot of data and don't know how to use it. so that's that's fairly common. and you know, I mean I I used to I used to run a people analytics team at IBM, and you know, we would try to answer important questions, and it was worthwhile, but it took a long time. Like every every question we'd want to answer was a six month project. We had to like dig into the data, get to know things, and even at that point, we only had IBM data.

Chad Sowash (07:22.485)

God, yes. Yeah.

Ben Zweig (07:44.683)

We never in our wildest dreams imagined that we could like benchmark against Oracle. Like they weren't going to give us their HR data. so so we we have so we're we're basically using a lot of public statistics, public data, so LinkedIn profiles, job postings, et cetera. And and we're we're trying to essentially reconstruct what the HR database looks like for a company that we have no affiliation to. So then, you know, we we can we can allow you know someone to analyze a competitor.

Chad Sowash (07:48.519)

Yeah. Yeah.

Chad Sowash (08:06.663)

Okay.

Ben Zweig (08:13.812)

And so, you know, through that, like it's useful for benchmarking, where you're not just benchmarking against some like ambiguous industry. You're benchmarking to like this company who you want to emulate and excluding that company which you think is like not worth emulating, to say it nicely. and we also have a lot of like hedge funds that use the data for, you know, to to track like workforce trends within companies they're tracking.

Chad Sowash (08:30.311)

Gotcha.

Joel Cheesman (08:37.484)

Yeah you do. Yeah you do.

Chad Sowash (08:38.905)

I bet you do. So it siloed data is is really the big problem because companies only see their own data and they really need to take a look at market research, not to mention competitive data, which you you do all of that. as as we start to dig into the report, one one of the one of the biggest pieces that really caught my eye was you you talked about supply. Supply being down twenty eight percent, the d decline in computer science and the IT side of the house since twenty twenty two.

Ben Zweig (09:08.351)

Yeah.

Chad Sowash (09:08.481)

now talk about that because I mean th there's still a lot of it's a lot of computer science, you know, jobs that are out there. Why why are these rolling down? Are they just are they just transforming into AI positions or what what's happening? Or are they plumbers? Who knows?

Ben Zweig (09:23.284)

Yeah, so one is like it's a huge increase. So 28% in in like a few years is is really nuts. so so students are not majoring in computer science to the same degree. I I think, and this is this is conjecture a little bit, but but I think it's fair to say that this is in response to AI, that there is a sense from students that computer science is ripe for automation.

And they are considering that major to be like potentially displaced. So they're putting l a lower premium on that training. So on the one hand, like it's very nice to see that supply is responsive to expectations of a certain technology. So, you know, one way to think about like, will there be job displacement? The way I kind of think about that is there's like three determinants to that. Like one is is supply reactive? If if workers can basically like

Chad Sowash (09:59.218)

Mm-hmm.

Ben Zweig (10:19.636)

go wherever they want very flexibly, then we have nothing to worry about. So so if if supply is like, you know, following market needs, that's great. That should make us feel comfortable. On the demand side, like if firms are adopting very quickly, that could potentially be disruptive. If firms are basically ignoring the technology, then nothing's gonna change and we have, you know, you know, then we don't have to worry about displacement, we have to worry about productivity. It's a different story. but the third, and I think the most important is like

Joel Cheesman (10:38.956)

Uh-huh.

Ben Zweig (10:47.114)

How quickly can the work actually change within a job? So, how quickly can the tasks of a given job kind of reconfigure to adapt to the certain technology? so I think the the supply being responsive is encouraging. I happen to think it's a little premature. so I I I feel like this might be an over.

Joel Cheesman (11:05.868)

Is that a data driven opinion or is that sort of your own nuanced opinion?

Ben Zweig (11:10.526)

Okay, so a little bit of both. So I think I think the the data that we're seeing from the AI lab. So so we so we had, you know, we we collaborate a little bit with OpenAI, with Anthropic, and now most recently with with the DeepMind team at Google. and and they are very quickly analyzing their usage data. And at least you know, OpenAI just came out with a paper yesterday, I think. So hot off the press.

Chad Sowash (11:15.398)

Yeah.

Ben Zweig (11:36.457)

About like the transition from sort of chatbots to agentic AI. And so it so that that is that is something that is definitely happening. And I think the the thing that's interesting about chatbots is that unlike other technologies in the past, they have had such low barriers to entry. Like anyone's grandma can use ChatGPT. It's so easy. so it so it doesn't it doesn't favor the digitally native. And and so so you know, you can sort of like outsource.

Chad Sowash (11:41.426)

Mm-hmm.

Chad Sowash (11:56.861)

Yeah.

Ben Zweig (12:05.844)

you know, question answering. Like that's very easy. But with with more agentic systems, it's pretty complicated. Like, you know, I I consider myself fairly technical and like, you know, I try messing around with OpenClaw and I'm just like intimidated. Like you have to you use the terminal. You have to, you know, it's it's like not so easy. You need to like really sit down and know it and learn it. and I kind of, you know, suspect that

Chad Sowash (12:25.745)

Right. Yeah.

Ben Zweig (12:31.732)

That the people who are more capable of using these these systems are the ones who are better trained in computer science.

Chad Sowash (12:38.993)

Well let's go over, I mean, just a little historical facts that we overindex on this a lot. we we push kids into college. Now we need carpenters, we need plumbers. we we push kids, our our our kids wanna start using some of the the the better, newer developer languages and whatnot. And we're getting, you know, dudes that are 70 years old to be able to go back and do COBOL programming because nobody's doing COBOL. So it's like it feels like

Ben Zweig (13:03.563)

Yeah.

Chad Sowash (13:07.889)

We're we're going in this cycle of, here we go again. Here we go again. We're over indexing on something. This is stupid. What what do you feel about that?

Ben Zweig (13:18.036)

Yeah, no, I I think it's a real thing. I mean, now we have a shortage of radiologists, right after Jeff Hinton, the godfather of AI, predicted there would be no more radiologists. So yeah, I think it's definitely a real phenomenon. I mean, one one thing that gives me a little bit of comfort is that when we talk about like, you know, college education specifically, like this is this is a a a change in the flow of new of new people, not not the stock of candidates. Also, I mean computer science generally, you know.

Chad Sowash (13:22.321)

Yes.

Joel Cheesman (13:25.026)

Mm-hmm.

Ben Zweig (13:46.785)

Has has the following trait where like people go in as like a software engineer and then they're a software engineer, and then there's a lot of exits from that occupation where they end up being product managers or whatever. So so there's there's quite a bit of like outflow rates. So if there were a shortage of you know software engineers, I I I'm not terribly concerned because the stock is pretty high and the outflow rate is pretty high. So like if if it were more

You know, if market needed more software engineers, like it would, you know, it could be just more attractive to stay in those roles relative to, you know, more of the exit options. But I do think people are overreacting a bit. yeah.

Joel Cheesman (14:29.206)

Yeah. And I I think, you know, when when I when I whenever we get an economist, I love it because they're data driven and they're not just sort of this is my opinion and we'll see what happens in the future. But but to ask the question of like, are are all the jobs going to go away? Are we all going to get UBI and chocolates from the government every month? And and do or or what I what I'm hearing I think is more of like Javon's paradox, which I assume you're familiar with, where

Chad Sowash (14:54.311)

Chocolates.

Joel Cheesman (14:58.242)

The efficiencies of coal meant we just had more trains and more energy and power. Where do you I that's my guess, but I want to hear it from you. If someone says, What the hell is the future gonna look like, your answer is optimistic, negative, and why?

Ben Zweig (15:13.982)

Okay, so definitely can't say what the future is gonna look like, but z that that's that's very tough. But in terms of job displacement, I'm I'm more optimistic.

Joel Cheesman (15:22.25)

If you were in Vegas and placing some bets, where would you place your bets? Let's let's let's question that one.

Ben Zweig (15:26.078)

Okay, so let me let me understand the question. Is this like are we going to see technological displacement? Like are going to see job displacement or are we or do we have less to worry about there?

Joel Cheesman (15:33.761)

I I think we all agreed there will be jobs that go away. But will there be more jobs created as a result of the new technology where more different skills are required? Which I think part of what your research does, we'll get to that of of people learning new skills. But so I I think you're optimistic on where we're going. Like you have faith that humanity will be okay with AI, or is Terminator coming and we're all doomed?

Chad Sowash (15:38.129)

With a beep replacement.

Ben Zweig (15:46.773)

Yeah.

Ben Zweig (15:56.021)

Yeah.

Chad Sowash (15:58.098)

He has three little girls. He is hoping.

Ben Zweig (15:59.863)

Yeah, yeah, I I'm praying. So I I mean

Joel Cheesman (16:03.214)

Two two were born pre-Chat GPT. Just you know, just

Chad Sowash (16:05.808)

Yeah.

Ben Zweig (16:08.028)

Is that right? No one. yeah, that's right. Right at the cutoff. I got one right at the cutoff. She's the yeah, yeah. She's the control group. Yeah. Yeah. okay, so so I'm I'm more optimistic that we won't see widespread technological displacement and that we won't see big spikes in you know unemployment rates due to technology. yeah, for a few reasons. I mean one is that

Chad Sowash (16:11.813)

I

Joel Cheesman (16:13.728)

I know the I know the numbers. I could be an economist. I could be at I could be at NYU starting school of business.

Chad Sowash (16:15.761)

Yeah. Conception different than birth, cheesemen.

Joel Cheesman (16:33.463)

Okay.

Ben Zweig (16:37.844)

I think if we if we take kind I mean, yeah, let me let me let me just bring some historical context. So in in wait, what was the year? In 1930, John Maynard Keynes wrote this essay, Economic Possibilities for Our Grandchildren. And he, you know, predicted, if we keep growing by 2% a year, you know, we are gonna have such enormous productivity that there will just not be as much to do, and we'll all be working 15-hour weeks.

And we'll just live a life of abundance and luxury and leisure.

Joel Cheesman (17:10.85)

He got everything right except the hours that we work, right? Yeah.

Ben Zweig (17:12.946)

Exactly. Yeah. So so I think he got he got everything. He actually like underestimated the growth rate a little bit. So so, you know, i but I mean he was like extraordinarily right on like the productivity effect, but was very wrong on what we would do to that. So I think there's yeah, first of all, like there's the vast majority of the world does not live in abundance. So so I think, you know, we we are very, very far from having like

nothing to do and nothing to build. I I think, you know, really anyone you ask is gonna think, yeah, we haven't run out of stuff to do in the world. Like there's plenty of useful, you know, we have we have much bigger appetites than like the the economy that we have today. And that's especially true on a global scale. And this is a global phenomenon. So so I think, you know, just like general, you know, growth, it just doesn't I I don't think we're running out of stuff to stuff to do, stuff to build. And you know,

There's this law in economics called SAIS Law, which is basically supply creates its own demand. So so basically what that really means is that the economy is constrained by its productive capacity. So if you have more productive capacity, if you have this big boost in in productivity through AI, then that loosens the constraints on what we can build. And maybe in a long, long time and you know, a hundred years down the road, you know, maybe

maybe we'll we'll, you know, have such abundance that we'll actually run out of stuff to do. But I think in the short term

Joel Cheesman (18:40.632)

More data centers, Chad. We need more data centers. More supply. Build them.

Chad Sowash (18:42.685)

god.

Ben Zweig (18:45.621)

Yeah.

Chad Sowash (18:46.889)

that's gonna that's gonna create so many jobs, Cheeseman. So with that being said, it it's actually taking more jobs today, more job postings today to get to the hire. Why is that?

Ben Zweig (18:58.772)

Yeah, so so that that I think really really gets at what I think is the the worst thing about AI today. I think I think you know there there's one area where where AI is actually harming the labor market. And I think that's really in the search and match process. And this is this is a market that you guys know really well. But you know, right now, you know, any job candidate can use these these new automated tools to kind of auto-apply for a hundred jobs a minute, whatever it is.

And there is a a race for speed. There's so many like, you know, hiring cafe and whatever, you they're they're all kind of popping up to like get you to be the first in line. Yeah, and and and they're, you know, creating these like sort of, you know, AI applications that are totally signal jamming the employers. And there's some asymmetry in in ha in who can use AI. You know, in in a lot of states in the US and and outside the US, it's actually illegal.

Joel Cheesman (19:35.47)

Slop it up.

Ben Zweig (19:55.627)

To use AI to evaluate candidates. Now, I don't think that's actually enforced. Like, you know, a lot of lot of companies are. Yeah. But, you know, there is some liability for for employers using AI. So I think they've been more reticent to to kind of deploy these tools, but but candidates are using lots of AI. So it's very hard to determine like who's a good applicant. And I think employers are having a very tough time.

Chad Sowash (20:02.365)

Not in this administration.

Joel Cheesman (20:02.648)

Not today. Yeah.

Ben Zweig (20:24.052)

Which means that the the actual like you know utility of posting a job is declining quite a lot. So it so now like the actual like numbers of hires that you can get from a given job posting is pretty low. So you know, in the economy at large, you know, in the US economy, job posting volume is like kind of flat, but hiring rates are very, very low. Hiring rates and attrition rates are very low.

So we have this environment that's kind of like the opposite of what we saw in the Great Resignation where people are just staying in their jobs. And I think it's it's due to th this like added friction in the ability for people to match with employers. It's just gotten a lot harder.

Chad Sowash (21:06.439)

Well, I think there's there are a hell of a lot less open opportunities as well. I mean, you've got the high high churn rate for like hospitality, you know, obviously healthcare, so on and so forth. But for the most part, I mean, if you've got a job and you've got a paycheck, it this is it this is the time just to sit still, be quiet, and wait for the market to to to open up. Yeah. Yeah. to your point, to your point, real quick though, it LinkedIn is we talk about

Joel Cheesman (21:25.324)

Yeah. Keep your head down. Do the job.

Ben Zweig (21:28.96)

So Yeah.

Chad Sowash (21:33.414)

all this slot that's being out there that's being used by you know by by job seekers. I mean LinkedIn, who was crowned the king of slop, they were pushing this on their users. and now they're asking their users to police it. the slop that they actually created. So again, it another point of over-indexing where we have these platforms that are like jamming AI in for their users to use.

It is backfiring on them and they're like, Holy shit, now we have to have our users police the mess that we told them to make.

Joel Cheesman (22:06.082)

Yeah. And you have indeed indeed pausing their automated, you know, apply process as well. So they're people are putting the br people are pushing the brakes on some of these some of these technologies.

Chad Sowash (22:09.691)

Yeah. Yes. They're bottiply. Yeah. Jesus.

Ben Zweig (22:11.807)

Yeah.

Ben Zweig (22:16.362)

I think that's great. I mean, as they should. I think I think, you know, for such a long time, and I mean, you know, you guys know more more about this than I do, but I mean, job boards, you know, have been basically ad businesses. And and this is and you know, I think there's an opportunity for them to kind of rebrand themselves as more like clearinghouses that that facilitate the match. And that that requires some costly signal for someone to apply to a job. So if I had my way, you know, if I were running Indeed or LinkedIn.

I would limit the number of applications that someone could submit per day, would maybe do some rank ordering or do you do something and and and you know try to figure out ways to facilitate more matches.

Joel Cheesman (22:56.024)

Yeah. And I I think to your point about the sort of the tsunami of auto automated applies and lazy apply and these technologies, I think a lot of real humans have been swept up in this process where I don't know what's a real human, what isn't. And I think real humans are really struggling with applying to jobs because they're competing with robots that are applying to hundreds and thousands of jobs. One of the the numbers in your research showed that companies that have successfully adapted

adopted AI show a 27% increase in headcount. so when you is that using AI to the hiring process where they're better able to filter out the real humans from the the bots? Explain sort of that number and where you guys came to AI good, headcount goes up.

Ben Zweig (23:45.397)

Yeah, yeah. Okay, so so it's a very tough so I think in this in this like how is AI affecting the labor market question, the the hardest part of that is to measure AI adoption. So I think we you know we have these measures of AI exposure, what are the jobs that that AI touches, like that that's fairly straightforward and very useful. But it's hard to measure adoption. So we had had a paper that came out with with RAMP. so Ramp you know is a payment company and they can track, you know, companies that that are spending more on AI.

Chad Sowash (23:52.754)

Yeah.

Ben Zweig (24:14.688)

Tokens. And the most intensive adopters are hiring a lot faster than the other adopters. So are hiring a lot faster than the not yet adopters. So so importantly, you know, companies that use AI don't look like companies that don't use AI. They're different, there's selection bias. But in this RAM study, we we're able to control for that by using, you know, firms that will adopt but haven't yet. So it's a kind of

For those statistic nerds out there, it's a treatment on the treated. So it's a causal effect for those who eventually get treated. And we do see an increase in hiring, even among young people, which which which is interesting because there's some other research that young people are being especially hit hard. And now I think we can conclusively say, maybe not conclusively, but we can for now say that that's not the result of you know displacement because of adoption. Now, the problem with the RAMP study.

which, you know, is is that, you know, that's an anonymized set of firms. So there there's, you know, they don't release their their company data. So we have a a different approach where we can't do this kind of, you know, isolated treatment on the treated effect, but we do look at you know, companies that we determine are intensely using AI through through, you know, basically onboarding like AI integration teams.

This is a this is a method that was established in in a different academic paper that works pretty well, tracks pretty nicely. And there we see an actual bigger effect of the adopters on hiring. So I think what what to make of this is that like, you know, when companies are investing in AI, they're not investing this, you know, they're not investing in like displacing workers, at least not yet. They they're they're trying to build more, they're trying to change the way things, the the way they operate.

it's something in economics called the the productivity J curve. You know, you have to take this big upfront investment in order to get the payout later in adopting these technologies. So when we hear of companies yeah. No, go on. Yeah.

Joel Cheesman (26:15.95)

So why do you why do you why do you think why do you why do you think the layoffs are happening if they're not because of AI? Are people we overhired in COVID? So we're sort of right sizing. we're getting rid of jobs that we know don't need ever, or getting rid of high, high paid people and replacing them with lower pro lower paid people that can use AI. Like what is your opinion on why we're seeing so many layoffs, particularly in tech?

Ben Zweig (26:40.918)

I think we're actually not seeing so many layoffs. So I'll I'll push back on the premise a little bit. I mean, layoffs in tech have been actually lower than than they than the historical averages. And most of the layoffs that we see are in manufacturing. So so which are kind of you know not touching AI at all. Obviously, the layoffs in tech get a lot of publicity. And they and I think for a while it was very fashionable to attribute those layoffs to AI because it makes companies look efficient.

Chad Sowash (27:07.57)

Mm-hmm.

Ben Zweig (27:10.612)

But you know, I think I think when I think about like block laying off 40% of its workers, I think number one, it's not consistent with with what we see in the aggregate data. So it's like an isolated case. Number two, you know, for anyone who works in a company, you know that like the worst environment to implement a change is an environment where everyone's overworked and like doesn't have the bandwidth to to change the way they do things. So like just a bad strategy.

But also number three is like even if it even if it were to take advantage of productivity improvements, it's such a depressing state of like the prospects for for block. It's basically saying like we don't have anything else to do. You know, we're we're just gonna like do the same amount we've always done with fewer resources and like do less RD or something. It's basically saying we have we've capped out our growth. So I think it's a bad strategy.

kind of inconsistent with like the nature of technology, and I think Wall Street's not really buying it anymore.

Chad Sowash (28:16.317)

Well, that's then that's who you hedge on. Then that's who you hedge on. so okay. I we're seeing a lot of optics flipping around. We've heard so many companies, dumbass CEOs say, we're not gonna have any, we're not gonna be hiring, we're gonna be using more AI, we're gonna be doing these things. Yeah, Clarona to only, you know, have to outsource and or bring, you know, humans back. This is this is something I got from a press release from I

Ben Zweig (28:19.198)

Yeah, yeah.

Joel Cheesman (28:35.032)

Klarna. Clarna.

Chad Sowash (28:44.997)

It it might be one of your competitors, Global Data. And here's a quote: In the automotive industry, Canadian Tire Corporation has a posting for the Director of Engineering Performance and AI metrics, which points out establishing an a agile and AI performance measurement practice. The initiative aims to augment traditional agile metrics with new KPIs that reflect total system.

Health, including this is the important part, including human AI collaboration and automation effectiveness. End quote. Is this a true transformation or is this just trying to s be the the the virtue signaling of you're gonna be okay and not scare the shit out of their employees?

Ben Zweig (29:37.405)

It's a good question. So first of all, love global data and and this is I think coming from Link Up who they acquired, which you know we we partner with and we love them. Okay, is that right? my god, I didn't realize that. Wow. my god. Okay. Wow. End of an era. Okay. okay, so so wow, no, no, I'm just like

Chad Sowash (29:42.897)

Yeah. Toby just left. Yeah, yeah, yeah. Yeah. Toby out.

Joel Cheesman (29:47.35)

Yeah. Toby out. Yep.

Toby out, man. Yeah.

Chad Sowash (29:54.812)

Yeah. Yes, very much.

Ben Zweig (30:02.248)

letting that sink in. But anyway, back back to business. Yeah, yeah. Wow, I need to, you know, shed a tear. okay, so so yeah, is that is that signaling, is it real? I mean, you know, job postings in general are like, you know, they're they're expectations about the future. They they're anticipatory. So who knows whether this will like actually pan out. And, you know, they're

Chad Sowash (30:03.943)

Twenty five years, dude. Twenty five years. Yeah, I know.

Joel Cheesman (30:04.042)

If you need a minute, if you need a minute.

Ben Zweig (30:26.822)

They're they're useful for signaling, you know, what what your intent is to the market, but they're also economic artifacts. Like this is this is something that they are putting out there to get candidates who are drawn to that to some degree. So I think I I don't think the content of job postings is is signaling, is virtue signaling. I I think it really is the expectation that they're gonna find someone who is drawn to this type of thing and it will like be more productive. Now

That's not to say it's not also like very optimistic that they can like start changing things and do a bunch of human AI collaboration and like change their, you know, the way they operate, especially in like some big, you know, procedural company. So I think it's gonna be hard for them, but I think it's at least a slightly optimistic signal that they are at least trying.

Joel Cheesman (31:21.55)

Sort of the big question and what your data shows that most of the change in work is happening within occupations. Say more about that. Explain what that means.

Ben Zweig (31:33.887)

Okay, so so yeah, I I think this is one of this is like my my my pet obsession because I think public labor statistics really do not address what happens within occupations. So basically like sometimes when people think about, you know, the effect of AI, the effect of trade, the effect of anything, you know, they think, what are the jobs that will increase, what are the jobs that will decrease? You know, will we see more of this occupation, less of that occupation? And I think what what gets missed is that the occupations themselves change a lot.

So, you know, I think it's it's especially kind of obvious if we think about like the white-collar workers that we know. You know, the stuff we do today is different than the stuff we did five years ago. You know, work changes. And, you know, that that is just the nature of change. You know, what when we think about this example of you know, bank tellers, which is like the the canonical example of like, you know, what happens with AI. You know, I think the story, which I don't really buy, is this Jevons Paradox story.

that that you know we we had you know this this automation this ATM which automated bank tellers and therefore it create it got so much cheaper to like do you know banking and it now we see a chase on every block and you know bada bing bada boom we see more demand because like it got cheaper to deliver these services. That's the Jevons paradox story. I think I think the the story that I prefer is that

You know, the bank tellers are no longer bank tellers. It's kind of an artifact of history that they still get called bank tellers, but they're really just doing like customer relationship management stuff. And if like, you know, Sandy Weil, you know, the chairman of Citigroup at the time said, Hey, we're gonna call these customer relationship managers, we'd be having a very different conversation about you know the effect of AI. And that's just the arbitrariness of language. So, you know, it's

Chad Sowash (33:17.425)

Yeah. Yeah, yeah.

Joel Cheesman (33:21.73)

Mm-hmm.

Ben Zweig (33:23.146)

You know, the the the job itself transformed into something totally different. And sometimes we say that's part of the same occupation, sometimes we don't. But I think the reality is that like jobs themselves change pretty quickly. I mean, I mentioned, you know, I was like, you know, a quant and then data scientist. And you know, what what was a quant, you know, in twenty fifteen is different than what a quant is today. What a what was a data scientist in twenty fifteen is very different than what a data scientist t is today. Same with sales, same with marketing, same with

Chad Sowash (33:50.141)

Mm-hmm.

Ben Zweig (33:52.126)

So many occupations. So, you know, if we think about the total work done in the economy, if we take all the tasks, you know, that are just just sum it all up, like the total stock of work in the world, and we think of that, you know, relative to a year ago, there's different things. Sometimes, you know, industries, you know, grow and shrink. Sometimes jobs transform. Sometimes there's more occupations of this, lower less occupations of that.

So there's this kind of like between occupation change and this with within occupation change occupation change. And if we kind of decompose that change, we see that actually 93% of the change in the the change in the work content is attributable to changes that happen within occupations, not by some occupations growing and some shrinking.

Chad Sowash (34:37.275)

And that is the Zweig Paradox. Ben Zweig, economist and CEO at Rivelio Labs. Ben, if somebody wants to connect with you, I don't know, maybe check out the tracker. where would you send them, sir?

Ben Zweig (34:40.444)

Yeah, you heard it here.

Joel Cheesman (34:43.458)

Vag.

Ben Zweig (34:53.334)

LinkedIn, you know, for to to follow me, I'm always like constantly posting this stuff. it's a real profile, so don't worry. and and revelabs dot com, you can see the AI labor market tracker, public labor statistics. We put out a lot of newsletters, free content, just yeah.

Chad Sowash (35:01.872)

Yeah.

Chad Sowash (35:10.247)

Great stuff. Yeah.

Joel Cheesman (35:12.334)

Slop free bin, everybody. Chad, that is another one in the can. We out.

Chad Sowash (35:14.141)

That's what I'm talking about.

We out.