Citrini co-author Alap Shah: White-collar jobs are down 8% from 2022 peak — and corporate AI adoption hasn't even started

show. A lot. How are you doing?

What's going on?

Oh, great. How are you guys? doing great. Is uh is this your first time triggering a a global sell-off?

The first time so far, but you know, I'm just the messenger is the way I look at it.

We've got a lot of opportunities and a lot of scary things coming down the pike.

Okay. So, yeah, take us through the the the thought process like how long had this been simmering? What was the actual process of putting together this report? And then what do you want people to take away from it? and then maybe we can go into some of the reactions and your reactions to those reactions.

Absolutely. The the process ultimately is that you know I've been building in AI for 15 years and I've been uh an investor for 20 and so especially the last six months as I've just been using aic coding myself and my teams have adopted it. uh it's just been a step change function and how much we can get done

and just thinking through hey how is this gonna you know we're we're early we're a startup you know we're going to be at the leading edge of how people are adopting things you know assuming the corporate world is a year or two years away uh it it's going to be pretty profound and I think the underlying thing you know as you know sort of an amateur macroeconomist is we're just not producing white collar jobs to begin with uh I hadn't actually seen the extent of that until I kind of looked at you know specifically what we call like the information sector so different parts kind of technology. Those jobs are down 8% from the peak in 2022 already. And so those are the places where people are adopting the most aggressively already. And we know, you know, every week there's firings out of like big tech.

And so in that world, what happens when the technology that big tech's been using for a while has gotten a lot better and now, you know, your average corporate starts uh using it as well. Uh it can get quite scary. And so, you know, we wanted to kind of think through the implications of that. and you know the piece

but how much of those how much of those layoffs do you think are are

you know we've talked about a bunch of those layoffs on the show

they're usually attributed to AI but if you dig under the hood it's like they just wanted to kind of resize or get more efficient or they're rep re rep prioritizing

resources

and not actually because they just launched some new agent and suddenly everything's changed hey we don't need these thousand engineers anymore So, I think, you know, those are all great corporate euphemisms and of course that's how they're going to say it. Uh, but I think the the way I would think about this is it's not necessarily like agentic powers happened and now everyone's going to get fired. You know, agents and LLMs broadly are are just sort of on the tech tree as a continuum from software. And so software has been making companies more efficient for decades. And you know, that has caused a lot of downstream effects. And now that software has just become much more intelligent.

And so in that sense I think you know companies that are efficient uh have been doing a form of this for a really long time and we think about you know the the the age starting now in 26 as just something that's going to accelerate that.

Okay. Um so yeah what what else was uh like key in the thesis or maybe potentially overlooked that you think people uh should be really focusing on? Uh I think the problem a the first thing the most important thing is just the labor market dynamics.

Uh we've just been in a really weak labor market for a while. Uh and that's before these things roll out. But then you put that together uh with the fact that you know we just have a very structural

uh environment where what what is the thing that drives our entire economy? It's wages. Uh most of those wages that are ultimately driving all the discretionary spending is coming from the white collar worker. And the problem with that is that we're now entering this place where you made all these assumptions on like loaning money to all these companies uh you know to mortgages and everything else like white collar economy is our economy. if you all of a sudden just take a leg out of that economy, it has a contagion effect into basically every asset uh in the world. And so that I think is the part that people haven't thought about because when you know people were making these loans, no one ever consumed of a world in which wow okay now like white collar jobs are in sort of permanent decline right if that's at 2% a year then I think we can skate through but if it's at four or 5% a year then you know we have to we need a action a lot more quickly. Is the is the white collar economy actually the full economy or is it more just like the stock market because it feels like white collar workers are disproportionately allocated to assets versus consumption and you see things like you know the like there's a lot of health in more bluecollar sectors healthc care is growing and then you also see dynamics like um just uh uh you know like we've seen like like jitters in the in the consumer market for a long time and then we just see the health of the American consumer just continue and continue and continue and it feels like it's maybe driven by something like lower level and there's always this disconnect in my mind between like the economy and the market.

It's a great question. Uh I think the issue here is that it's all just one labor market and right now blue collar is doing better because there are not firings there. Yeah.

Uh I don't think you know I think robots are probably 18 like 24 to 36 months behind. Yeah. uh other forms of LMS that are you know just diffusing through society. But the problem is, let's just say it's one labor market ultimately. And if there's no more white, if the white collar jobs are going away, let's say, you know, in our scenario, we talk about 5% of folks might get fired in a couple years.

Those 5% if there aren't white collar jobs for them to relocate into, then they're going to have to move into the gig economy and the blue collar labor force. And so that puts pressure on the entire labor market, uh, not just, uh, the white collar one. And to answer your other question, healthcare is growing, uh, education is growing. Yep. uh the reason those things are growing ultimately and we did uh some work in our piece to try and isolate white collar that is not government driven

and so the government continues to spend more that's why healthcare is growing they're the biggest payer in uh in healthcare they're they're guaranteeing all the loans in the uh the education industry and so those those sectors continue to grow because government spending grows but that's again it gets very circular if government spending is coming primarily from taxes

uh and primarily payroll taxes because the average worker pays a lot more in taxes you know per taller than the average corporate does. Uh and so some corporates make a lot more money, workers, payroll taxes go down more, then there is a bit of a a contagion effect into bonds as well there too.

Mhm.

Uh on Saturday, John and I were going back and forth about uh some of the really wild predictions around the impact of the internet that were being made in the '9s. There was clicks replace bricks. People were predicting total die off.

Clicks did well. I mean, to be fair,

no, I mean to be like I'll just finish. They were expecting a total die off of all brickandmortar stores in 5 to 10 years, which was like widely widely discussed uh prediction. It was like why would you ever go to a store to buy something if you could just get it online sent to you directly?

Yeah.

A couple a couple a couple others. So like uh not as relevant to your piece but uh people were predicting like permanent high growth, the end of business cycles. Uh there was the like media disintermediation narrative which was like the the Napster era. Everyone was going to get all media for free forever. Newspapers would would die off. Record labels would die.

Aren't you guys the aren't you guys the media dis mediation narrative at large?

Yeah, we are. 20 years later and CNBC is still a much much bigger business than all

all business media uh at least in our world but news but newspapers like magazines completely gone right all of that has moved to the internet

totally totally it's just like 5% employment shock in a quarter is way different I mean like like a 5% unemployment shock is completely different if it happens over a quarter than if it happens over two decades right like these are just fundamentally way different things.

So the other the other thing

the last thing I would say is like there was like this concept of like frictionless capitalism meaning that like middlemen would be eliminated because you could just go directly to the source and that would push pricing pressure down. My question and I know you guys are not writing your piece saying like you know this we we believe we will stake our entire reputation on on this sort of uh narrative but what do you think what what how much did you pay attention to like the '9s early 2000s internet predictions? What do you think they got wrong? Why is this time different in terms of how a new technology will diffuse the economy? Uh I think the difference is if you just plot uh what's happening to technology, it's all just going exponential. These are all just continuous timelines of like we have microcomputers, we have the internet, we have mobile phones, and today, you know, we have very powerful AI.

And so I think most of the predictions that you you ticked off there, it's kind of interesting. I would, you know, just looking at them today, you know, I would say they couldn't really happen until you had proper AI because like if you have the ability to just freely, you know, convert like have commerce the way you do today doesn't work uh if you still have to do all the work ultimately like you have to go and you have to log in and think about the frict amount of friction there is in buying a product for most people today right you still have to go to the website you have to put your credit card in it's all work we only have gotten to kind of the tech required for those predictions I think this year and and that's why this is the year that I think it really begins because now it is completely seamless. You just and no one's really doing this yet but it's going to happen I think you know in the next 6 months is just tell your agent you know tell Gemini tell chat GBT go buy these things it has your credit card and now that world that they were describing is is truly going to come to pass.

Yeah. What about uh the canary in the coal mine analogy? Uh I was looking at uh unemployment statistics in India and the Philippines and it doesn't seem to be doom and gloom over there. I don't know. I didn't dig in super far, but would you at least expect that the unemployment rate would spike overseas before it spikes in America or do you think this all happens simultaneously?

It's a tricky question. And I think ultimately white collar work is a lot more of our economy than it is the economy of India and the Philippines and they are much sort of like more immature economies that are growing through investment and through things like that.

Uh but certainly I think we called it out the the uh the consulting sectors in India are certainly going to be challenged uh in other places as well. But the reality is like the the timing timing is everything in the markets clearly. The the trick here is if you're a corporate and you are hardressed to get AI uh into your organization today,

you know, chatbt and open AI will send you a forward deployed engineer if you have billion dollars in budget, right? But if you have a $10 million budget, they're not going to uh and so who are those folks turning to? They can't usually do it themselves. And so they are going to uh the outsource providers, the Accentur of the world. And so I think those businesses are are are likely going to be in a lot of trouble over the medium term, but they probably will have a big bump uh from people really uh putting that AI into their organizations first. And so it's it's a bit of a tricky timeline there. What moes do you think hold beyond this? Because I think a lot of people latched on to like the Door Dash example as something that they thought had a moat and that in the post you sort of underline like how that could maybe not be as durable as a moat as people thought. Um but in the long case like what what modes do exist? Like do network effects stay? do complex coordination, intellectual property, like what what doesn't break down

uh you know real brand value uh where people are choosing you over other things because of the brand and the status signaling across brands matters a ton.

Sure.

Uh network effects are are more powerful than ever I think in this world. Uh so things like meta really have a lot uh to to sort of gain in that sense. But I think things that look like they're network effect businesses but in fact are just the ones that are doing the hard work of aggregating demand and supply I think will be more challenged. And so Door Dash is a good example there. It's not necessarily the the biggest risk versus some of the other things, but I just was in a thread with Gavin Baker talking about this, but the problem for Door Dash and Uber and folks like that is right now they're doing two jobs. They're doing the job of aggregating demand and the job of aggregating supply. Uh they're both hard jobs. uh but the demand side is the harder side and in we think the world of the future uh there are lots of folks in let's say food delivery uh you know Instacart wants to get a bunch of market share uh and you know GrubHub wants to get a bunch of market share and so let's say the agents are the ones doing the buying it's 2028 and 40% of the the sales are through agents you just tell Gemini hey order me some noodles uh in that world instead of it's going to go to each and every provider and right now there are four providers that do that but now it's very easy as if I'm building a startup in the space previously I had to get all the drivers on board get all the restaurants on board and acquire customers now Gemini and chatbt are acquiring the customers for me and all I have to do is get this get the supply side going so it makes it much easier for new entrance to come in and for existing you know sec third fourth tier players can really sort of say I'm going to like relax my margins try to get more topline and so you think that you know whatever the 15% vig is that Door Dash gets today maybe it's more than uh you know some of that I would think Gemini and CHBT are going to ask for themselves. Wherever I send the traffic, I'm going to get a piece of that. Uh and then some of that's going to go back to the consumer.

Yeah, it feels like Was this the most like stretched or controversial prediction?

It seems like it was certainly the one that got, you know, getting the most chatter. Uh and I think we did it for a reason. We wanted to be a little provocative in in thinking think it through because you know it's an amazing business and they're gaining a bunch of market share but the fundamental idea that

you're cuz what the lock in right like the drivers have lock in uh on Door Dash or on Uber

not really right they're you know most most drivers are doing and Uber so they're they're not locked in the real lock in the real business value the franchise value of an Uber or Door Dash is the customer lock in because the customer gets comfortable they've got everything saved they want to hit a couple buttons they don't they don't price shop agents are happy to price shop as much as possible. And so if you take that away, then it's a real problem for businesses that, you know, are ultimately built on customer lock in.

Yeah.

Yeah. I don't know. I think the interviews that we've had with the lift, I mean, you know, again, take take it with a grain of salt. They have a narrative that uh is important to their business, but like if you ask these people what is the greatest challenge, it is managing managing the supply side. It is not the demand side is not where they're saying like hey like this is really what we need to solve. It's like hey as we get more drivers on the platform revenue naturally naturally goes up. And so I just I'm I'm just hardressed to imagine a world in which uh you know somebody think about if somebody uh in my town which is like 15,000 people like vibe codes a delivery a delivery app and I go into chatbt or with another agent and I say like I want food. It's like the agent wants to get the best possible service. I would imagine the agent to route to the platform with the supply that is going to be able to deliver in the shortest possible time horizon and imagining a world where there's like this you know vibecoded small team operating that just happens to aggregate as much supply which is just as increases the likelihood that my order will be delivered on the best possible timeline which is going to be the number one factor for customer satisfaction. I just don't see how solving the front-end kind of demand piece actually makes a better consumer experience, which I assume the agent would optimize for on behalf of the user.

So, let's let's consider what actually happens here, right? You make the the the order on Door Dash. Yeah.

Door Dash sends it to the restaurant. Y

uh the restaurant essentially, you know, sometimes they use their own driver, sometimes they send the drivers from Door Dash. Uh but now imagine the agent can take you directly to the restaurant site and place the order directly with the restaurant. Uh and you can keep half the the savings and uh the agent can keep half the savings, right? The

where's the driver where's the driver coming from? Because I I feel like I understand I understand the the the customer demand side like you start with an LLM or an agent who shops around for you. So maybe that's solved. Maybe it'll find you just via SEO and you can just put out like we only take a 5% cut instead of 15% and the agent picks you. I understand getting all the restaurants on board because you email them and say hey it's 5% instead of 15%. They're sure we'll turn it on. But for the drivers, how do you actually reach out to them and get them on the platform? And how how does AI like lower that cost? Because right now I I think about like what was the driver marketing budget over the last decade at Uber or at Door Dash and it's probably like in the billions of dollars. And so I feel like to to to generate that much liquidity I have to invest that much to onboard all those drivers build awareness. Maybe it just goes viral because they're like hey I can make more money here but that feels hard. I think it's going to take time, but I think there are a bunch of smaller sort of driver aggregation networks that exist today that are not the ones that we know about. For instance, uh I started a business called Thistle and we do delivery of healthy foods to your door. Uh we we split it between half of them our own employee drivers and the other half, you know, I think we have like like 500 or 700 drivers that we just use a third party service to provide. So I think there are a lot more of these businesses. All those businesses now will also just have huge opportunities to kind of take market share. Ultimately, what we're saying is the friction in doing commerce is going way down. Places where there are rents, the prices can go down, but ultimately this is just an opportunity for more entrepreneurs to kind of build businesses uh for the new world.

Yeah, I think the the the it's interesting because we're here like debating like this this like somewhat temporary thing because like self-driving cars, robotics like changes all of that like in a huge way. Um but uh but but we we we use the term sloppable for companies that are uh that can be vibe coded away and and clankable for companies that can be disrupted by robotics. And I've always put the delivery services more in the clankable category than the sloppable category. So I was I was shocked to see

what what are the spend more time on if you knew you were going to get uh 50 million views and uh the the markets would react in the way that they have. Uh I would have finished writing the third piece where I talk about solutions which I have not gotten.

A lot of people are demanding solutions. They're like you just gota you just hit me with a ton of problems. That's funny. Uh do you think that there's any uh there's this question about like in my mind like yes Google and Nvidia are public but uh anthropic openai and uh XAI through SpaceX are not public. uh there's sort of like this massive uh you know multiple hundred billion dollar selloff in the public markets that sort of should if you believe your thesis that should sort of funnel to the labs I would imagine if when I read it like there's a lot of doom and gloom about companies that are out there but it's a lot of bull it's a lot of bullcase for AI labs and but that can't happen in one day because like rounds happen every once in a while they're private there's all these different things but uh do you think that the world would change when the big labs get out in the public markets.

Uh I think it's absolutely going to change. I have a strong suspicion that Anthropic is going to go, you know, in the next 3 to 6 months. They just have so much momentum and there's a lot of value being cursed.

Uh P&L also just looks a lot better than anyone else. So, uh I would think that gets public and it's going to be pretty interesting if it happens. Certainly labs are are ultimately they seem like they're they're we're very well positioned to win. Uh I would wonder over the medium term like you know what happens with some of the uh the Chinese models and whatnot if people actually want just something that's more local and something that they own. But it does seem like the most likely outcome is going to be that the existing incumbents are going to get the most share and you I think Google is particularly well positioned since they already own all those customers today and they can finance uh losses uh from inference a lot longer than everyone else. But I think ultimately like there's a world in which the labs are the biggest winners here. Uh there's also a world in which like you end up with just a lot more competition and people trade and and change. But the thing that seems very clear to me that the absolute like there's no way they won't be the hugest winners here is going to be the underlying tech meaning the semiconductors.

So you know everything

you go even deeper you could go into like you know commodities and like copper and energy and oil and natural gas and stuff and people have

Yes.

Did you see the

Yeah. Did you see the the some of the criticism was that the the essay was very Marxist?

Heat said Mark Mark's writing during the industrial revolution predicted capitalism would periodically devour itself. Firms replace labor with machinery to boost profits, but competition diffuses the technology drives prices to marginal costs and the gains get competed away. Meanwhile, displaced workers lose purchasing power, hollowing out the demand the whole system depends on. Production rises, but no one can afford to buy what's produced. the contradiction between production and realization. Satrini's piece describes this exact dynamic, then declares there's no natural break. It's the most Marxist piece of financial analysis. Not my words.

I don't think you were expecting that critique.

And makes the same errors Markx did. Uh yeah. Uh creative destruction doesn't just destroy, it creates industries we can't yet conceive of.

Yeah, that's an interesting. I mean, maybe that's going in the solutions.

Is that going in the solutions?

So, let me address it a few ways. Bars was a really smart dude. uh he got a lot of things right very early. Marxist can mean communist. Marxist can also mean just understanding how capital and labor interact. And in that sense, yes, it was Marxist. He had he was very insightful.

Um but I I think the thing that uh that we're missing here is that

it's there there's there's the economic layer, but ultimately it's the political layer that matters. And you know, we're in a world where we've we've had two parties

and both parties uh you know, economically have a little bit of difference, but not a huge amount of difference. And so we kind of bicker. But in a world in which jobs are going away really fast, uh I think there's going to be a much stronger alignment uh for, you know, just the laboring class overall to say, "Hey, we need to fix this problem." It's a very fixable problem. What I'm what we're actually expounding here is that GDP if done properly will absolutely explode. Right? We're getting way more efficient. We have, you know, we've built a machine God. We build machine intelligence, but we have to structure our society such that as those things happen very, you know, hopefully very slowly, you know, we we do the right thing from a taxation perspective to say the winners should win, but you know, if that's what's causing the displacement, let's sort of make the pie a little bit bigger for everyone. And that I think ultimately should be something that appeals to a lot of folks in the AI complex. Y because if we don't then something like this is likely to happen and you know AI progress will slow down because we'll have an economic crisis and we're not going to be able to finance nearly as much of it as we otherwise would.

So do you think the future is uh what anthropic head of sales position in France? The company uh will be spending 530,000 per year. the government will get €340,000 and the employee will get €190,000. Is that is that the level of taxation you think we're we're headed for?

Uh I think when we're at you know France's level of government spending then you know the math probably means roughly that. Uh I I would say that you know government spending would be at France's level I'm guessing like you know five seven years from now if this if this scenario kind of comes to pass. Uh and so I think we'll head there over time, but I think it's less a question of the percent of spending and how much goes to the employee versus goes to the the government and ultimately what is the size of the total pie.

So the bet here is that the pie if done properly can in just increase multiples of what it is today and and thus you know there it's just a win-win.

One question I mean it sounds like you're working on potential solutions post which I'm very excited to read. Thank you. Um I'm interested to know your reflection on the messaging that's coming from the leaders of the AI labs because they've outlined many sort of low probability but you know potentially negative scenarios. You know we have the white collar work number. We've had many of these comments from lab leaders, but I rarely hear them follow it up with and the answer is print print or interest rates will be will save us or unemployment insurance or UBI like all of those like solutions that I think people it's funny because people are quote quoting your post being like this is easily solved with this solution. It's like okay well that's great if we all agree. Um and I think you might with some of the some of the quotes people are all over the place. Um, but I'm wondering about your reflection on like the the like like the messaging from the labs around solutions versus pure focus on problems.

I think it's a really interesting question and very interesting setup in that the labs are on the one hand, you know, want to get the word out there and so, you know, Daario especially has been the loudest here. There's a really good Axios article from last May where he's he kind of sounded the alarm bells. Uh, people aren't really he's like saying people are not listening. Obviously, a lot's changed since then. Uh, but they can't go so far as to say like, hey, if you put the pieces together, then this is how it's going to play out. I think it's too sort of damaging to sort of their reputations and like, you know, their ability to fund raise and things like that. And so, I think it's other folks like ourselves that kind of have that

uh duty to go and really start thinking that through. I I I sort it seems like Anthropic is pretty engaged. Yeah.

Uh, you know, should that conversation really start happening and I think this is the year it needs to really start happening. Uh and so I think they all kind of get it and so it's just a question of like how do we as a society start moving in that direction.

Yeah. I I I think uh you know obviously there's there I'm still processing part of the piece. I agree with some of it. I disagree with some of it. But uh what's really underrated is just like how useful this process of writing an article for a particular audience is. like I I I disagreed with a lot of you know something big is happening but it hit with a very different audience than machines of loving grace or the adolescence of AI or of machine intelligence. Um, and and there's there's pieces that are written for like, you know, AI insiders, leaders, researchers, then there's like the broader tech community. Then there's like everyday people. And you clearly hit the nail on the head with like speaking to the financial community. And we see that in the markets, not amazing results, but maybe it's maybe it's worthwhile because we will get really great solutions and a better conversation around it. So uh I I I I I I think uh I think in in due time uh this discussion needed to be had. So thank you.

What uh what what's it what's an industry or job of the future that you could see emerging?

I think again if we solve this like everything related to sort of leisure is going to absolutely zoom and that those are going to be the biggest growth industries of the future right like what do humans want to do? Total ski victory.

Watch polo.

Watch a cloned horse play polo for sure.

Yeah.

So, you know, imagine humans have like the entire day to just enjoy themselves uh instead of having to work.

Now, that is something I've been promised for a hundred years. So, I'm I'm deeply skeptical, but this time is different. I want it to be different. Let's bring on the leisure boom. I'm I'm I'm here for it. I'm here for it. uh on anything in your solutions doc around reindustrialization. I mean the the frustration that so many people in tech that have that have been building in in kind of hardware in the real world uh or trying to recruit people that uh that are getting offers from social media companies or now labs or SAS companies. You know, one of the the problems of America, you know, for America in the last 20 years was that if you just wanted to make a hund00 million, you probably were much more likely to do that building enterprise software than uh building critical infrastructure or or anything in the real world. So, is that is is uh kind of new new infrastructure and re-industrialization like a a a potential landing point for people that had the 180k a year PM job that might be going away.

It's a great question. Uh I think there's there's certainly going to be a lot more opportunity in those sectors and I think we we've done some pretty smart policy things that are moving us in that direction. uh but we're also you know just in a lot of ways so far behind China there and doesn't AI affect kind of those jobs both uh for you know on the reindustrialization side just like it does for for writing code

and so that's where I think it will get trickier I think over as a as a country we're going to spend an awful lot more on that and I think we're going to we're going to catch up but we're not uh it's not clear that's going to be through just creating a bunch of additional jobs versus you know the ultimate thing we're seeing with AI period is just high agency people uh who really know how to use the tools can just do the work of many many people.

Yeah.

And and I think that trend applies in every industry to some extent.

Yeah. Uh what an exciting time. Thank you so much for taking the time to

when's the next piece dropping?

Uh hopefully by the end of the week, but don't don't don't hold me to that.

Now when you when you when you know that it then it could uh be hard for the followup to get as much reach as as this one. That's kind of the way these things go. But now now the pressure is on to really pay attention to Don't have any SQL anxiety anxiety. You'll be fine.

We're excited to read it and we'll talk to you soon.

Yeah, great to meet you.

Have a great rest of your day. Thanks so much. Let me tell you about Turbo Puffer serverless vector and full tech search built from first principles and object storage. Fast 10x cheaper and extremely scalable. And I'm also going to tell you about Gusto, the unified platform for payroll benefits and HR built to evolve with small and medium-siz businesses. And without further ado, we have Will Brown from Prime Intellect in the TV.