Monthly Macro: The Work AI Cannot Reach

Richard de Chazal 00:06:18

Good morning.

So, it's the 15th of July, and welcome to the July edition of Monthly Macro. I'm Richard de Chazal, William Blair macro analyst.

Today, we're doing something a little different. Instead of me talking about the Fed, markets, and the latest economic data this month, we're going to be discussing one of the biggest macro questions out there at the moment. And that is: what is the likely impact of AI on employment? Are we starting to see an unfolding job apocalypse, as some doom mongers in Silicon Valley are suggesting? Or, are there reasons to be a little more optimistic about what's likely to take place?

And I'm very pleased to say, here to discuss this question, is my special guest, Professor Luis Garicano.

Dr. Garicano is a professor of public policy at the London School of Economics. He has a Ph.D. in economics from the University of Chicago, where he was also a professor there for a decade. And Luis has spent much of his career studying how technology changes firms, productivity, and the way work is organized. He was also a former member of the European Parliament, so he brings both an academic and policy perspective to this debate.

And he also has a new book out, along with his co-authors, Jin Li and Yanhui Wu, titled Messy Jobs: The Work That AI Cannot Reach. It's an easy and entertaining read, and helpfully full of real-life examples.

So, Luis, congratulations on the new book, and thank you very much for coming on to the podcast to discuss it.

Luis Garicano, Ph.D. 01:41:27

It's my pleasure to be here, Richard. Thanks for having me.

Richard 01:46:00

Pleasure.

So, perhaps, let's start at the beginning. The book is really about the interplay between jobs and technology, and specifically how we should be thinking about what types of jobs are likely to be impacted by a AI. And the way you go about doing this is really to distinguish between a job and a task and how those change with the introduction of new technology.

So, maybe you could discuss this? and exactly what is a messy job and why, maybe, going forward, we're all going to want one?

Luis 02:23:00

Thank you. Yes. The starting point of the book is, indeed, the task is not the job.

We really were motivated. We started writing a year ago. We were motivated by the stories that Dario Amodei, Sam Altman and Mustafa Suleyman the CEO of Microsoft AI, all of this, all of these gentlemen like to tell of this story that you name this "job apocalypse."

The idea that all white-collar jobs will be gone. Let me quote from Amodei. "Whether you're sitting down at a computer, either being a lawyer, or an accountant, or a project manager, or a marketing person, most of these tasks will be fully automated by an AI within 12 to 18 months." This is what he said in April 2026.

The book is motivated by these types of statements. And the starting point in it is that a task is not a job.

Since the start of this AI revolution, since maybe the first papers on these in 2022, 2023, it's all been about exposure, which jobs are exposed to AI. And, you measure exposure, and then you say, okay, this job's gone.

And there is really no such connection. The exposure could be helpful to some people, it could be helpful to others. And the key idea here is to think of the bundle of tasks that compose the job.

There are some clean, simple, single-task jobs where you have reinforcement learning. You get computer feedback on how it's doing, and it just climbs and gets better.

But those are the minority of the jobs, those clean, single-task jobs. Most jobs are messy. Most jobs are packages of tasks that are strongly interrelated with each other. That's what we called a "strong bundle," which is one of the concepts in the book.

When the tasks are strongly bundled together, then you might have help in doing some specific task, but the job persists, the bundle persists.

The automation is not able to replace you, but, in fact, it's going to help you, complement you.

So, of our listeners, of the people in the investment, what they want to look at is to kind of go beyond the task models and try to think of, okay, which sets of tasks are bundled together in this particular job.

We talk about Geoffrey Hinton, for example, making a prediction in 2016 that radiologists were out of luck. You shouldn't train to be a radiologist.

As you know, the demand for radiologists has never been higher. Both prices are the highest, the salaries highest ever, and quantities are the highest. There's never been as many radiologists doing training in the US.

So, there was a task that the computers can do very well. Hinton was right about this, which is scans, it can read scans very well, but obviously a radiologist does it. And the job is much, much more than that.

And that's a messy job. A messy job is a strong bundle of tasks that includes relational tasks and cognitive tasks interrelated.

And we will talk more about that interrelation, for sure.

Richard 05:41:18

Great. So, messiness is effectively a bundle of tasks. And what's, kind of, left of the messiness is what AI can't tackle. So, things like authorization, responsibility. For a doctor, it's not AI, you know, the radiologist has to take the responsibility for what he's doing. It's, sort of, the friction amongst all of those things.

But, what about, in the book, you talk about Jevons' Paradox and, actually, how AI and technology could actually increase the number of jobs or employment out there. Is that that part of the discussion?

Luis 06:25:04

So, I would nuance a bit what you said in terms of, it is not just that some tasks are left over, like, indeed, as you said, the liability or responsibility or all that, but, the bundle itself might persist when the bundle is sufficiently strong, when, kind of, peeling away the automatable task is actually more costly, in terms of the lost context, the lost synergies, the lost spillovers between tasks, than actually leaving them altogether.

So, the messy job really is not just a messy task, in that sense. It really is the bundle.

And yes, you're completely right. We're talking in the book about Jevons' Paradox. It's really a very important aspect of the analysis. We take a very price-theoretic perspective in the book. And Jevons' Paradox says, look, even if something, one particular input, becomes much cheaper, it could lead to a big jump in final demand if it is sufficiently elastic.

So, Jevons' was talking about coal, which, of course, many people know. He was saying, look, I mean, coal got much, much more efficient. We've gotten better at using coal. So, we developed new steam engines, etc., and, as a result, our demand for coal got much better, even though each usage of coal was more efficient. We were using more oil than coal at the end.

And, so, we could be using more cognition if the cost of cognition becomes cheaper. But, we learned all these other cognitive tasks that we were not thinking about.

So, let's think of the radiologist example. So, radiologists, clearly, scans are probably very elastic. We could probably have a full body scan every month, if this was a cheap thing. If this became really something that could be relatively easy to automate, then it would be one radiologist on top of a pyramid of agents.

We would, with this agent and these robots, would be able to do the scans, and we would be able to have a scan every month, potentially to the point of actually increasing the demand for radiologists, even though each scan is taking much, much less time to diagnose, etc., from the radiologist.

So, the two things interacting together, which is, obviously, if the task gets automated itself, you need these Jevons' products to be really gigantic because then you have a task that requires almost no human component, and then the little humanness is going to be, kind of, for it to really increase, you're going to have to have a very large amount.

But, if you have a messy job where the task of radiology is linked to these other aspects, then it's very plausible that jobs in the price of this particular component of the entire messy job are actually not going to lead to a job, but to an increase in the utilization of this and increasing the overall demand. And, as a result, indeed, Jevons' Paradox could be very relevant here.

So, the book very much goes through different jobs, and different sectors, and things about elasticities, elasticity demand, and tries to figure out under what conditions, indeed, we will see increases in the number of jobs.

And we do expect, I mean, we are very skeptical of job apocalypse. We do expect increases in many of these professions, indeed.

Richard 10:09:23

Good. And that's encouraging, the lump of labor fallacy is there.

Luis 10:13:12

It's difficult to feel like this.

Richard 10:16:10

Yeah. When I'm thinking about, you know, the types of jobs and who's a winner and who's a loser here, who's going to be the winner here? Is it going to be, sort of, lower-skilled workers or higher-skilled workers? Maybe, suddenly, the lower-skilled workers become much more productive, so their pay increases, they get paid more?

Or do you see, you know, the grunt workers, you know, more of that grunt work can be done by machines, so managements hire fewer of these lower-skilled workers? So, you know, labor costs come down, and then the value drives up to, you know, upper management, and they're, sort of, the profit winners here?

Luis 10:52:25

Yeah. The concept we want people to use to think about this is the... go back to the strong bundle idea that we had, and think about the strong bundle, and weak bundle, and the re-bundling and unbundling of tasks.

So, you have a job. The job is a bundle of tasks. And what you want to do is, you want to, in order to understand this returns-to-skill question you were posing, you want to understand how the bundle changes.

We basically have two configurations that we focus on. One is the robots-below configuration. So, the robots-below configuration has a star trader, for example, a star financial consultant, etc., who replaces workers below by multiple AI agents. And, here, what you get is an increase in leverage. You can leverage your talent much more, and you have a superstar effect where one person, plus many units of labor, means very big returns to talent and big increases in rents of this margin of people.

There is another re-bundling that we think is very important, which is when you can re-bundle the non-cognitive tasks of nurses, or plumbers, or other people in the economy with this newly cheap cognition.

So, here, what you have are robots-above configuration, in which the nurse, who before was not able to diagnose a lot of illnesses, is able now to, with this AI, imagine it's nurses somewhere in Africa, for example, and it's facing some complicated illness and before they couldn't do anything about it.

And, now, they can use the AI to provide these emotional, physical, and all the other tasks, together with more cognition. And, here, what you have is convergence. We have the lower, the least-skilled workers getting closer to the more-skilled ones.

So, what you have to do is, you have to try to understand how the bundle changes. You have to try to understand when the bundle is strong, which aspects of the bundle are remaining, and which aspects of the bundle get complemented by AI. And, in different cases, you're going to have these robots-below or robots-above. And then you have to think, what are the remaining barriers to entry? And what are the elasticities?

As you said, with Jevons' story, because, of course, as nursing becomes easier because the cognitive task gets done by the AI, it doesn't necessarily mean nurses are better off. You will still need more demand for nursing, which, I think, in this case, is probably legit. And we also need some barriers to entry, because if everybody can be a nurse now because there's no cognitive demands, then, of course, we will have a drop in nursing.

So, the professions that will do better will have this strong bundle where the bundle, kind of, AI is going to complement these cognitive, etc., non-cognitive tasks, plus remaining barriers to entry and elastic demand.

Richard 14:13:03

Thinking about, I mean, one of the things I was thinking about is, and in the book you talk about it, these, sort of, helicopter drops. You know, the analogy that suddenly you can take a completely unskilled worker, put them in front of AI, and it's as if they're, sort of, dropped, you know, three quarters of the way up a mountain, in terms of what they're capable of doing.

Whereas previously, that unskilled worker, you know, might have started in the mailroom, work their way up through the firm, gather a lot of knowledge and experience along the way. And, you know, that's extremely helpful that they know the firm when they finally get to the top. And is this new, kind of, helicopter drop of information and getting you there much faster, is that is that a bad thing?

I mean, are you suddenly going to find workers that are near the top that just don't have the experience? And then when the older workers, sort of, move out, retire out, that company is left much weaker because, you know, those existing workers, remaining workers, never really went through the hard climb to get to the top, as it were.

How do you think about that?

Luis 15:33:02

This is divided into two moments.

So, let's do, first, the particular project. The helicopter drop in a particular project can be problematic.

So, think of some complicated analysis. Before, you had to spend 20 hours on it. Now, you spend half an hour on it. Something pops out. You have no idea what popped out.

So, in many cases, the temptation is, okay, I'm 80% or 90% of the way. I'm going to take another 19 hours, almost as much as I took in the past, to really figure out what has popped out, why this answer is the way it is, really checking and verifying everything. So, I'm going to submit this 80% work thing.

So, in many situations, what that's going to lead you to, project by project, is moral hazard and a slope into things that are not quite finished because people are just like, "The return is really very low on the margin. Like, okay, now I have to spend hours and hours getting another 1%. Another 2%. Well, let me just deliver it."

Whereas, before, you know, if you delivered the thing that you did in half an hour, you would get fired instantaneously.

So, project by project, there is this risk, indeed. This risk of increased moral hazard, of increased temptation to just leave things halfway and not know what you're valuing. And, indeed, as you were pointing out, there is the career side.

And the career side is, we're not going to hire the person who can give you 60% maximum because they are untrained and they are just coming out of MBA, if the AI already can give me 80%.

So, the people who are starting in the company, who are paying for their training by doing menial tasks, and you, Richard, and everybody in finance knows what I mean. You often start by doing spreadsheets and doing all sorts of menial tasks. You are paying with these menial tasks that now have zero value.

So, the currency that you were paying for your training with has become completely depreciated.

Plus, there's a machine that can do the spreadsheets, that can do the PowerPoints that you were doing before. You're not going to be hired.

And that means that the company is, indeed, potentially not going to have, eight, ten years down the road, the experienced people that could have potentially gotten this training and become the stars.

So, I think that that is something that is calling universities, and you guys in the private sector, to really rethink how we do this.

We probably cannot hire as many juniors. I mean, the private sector can hire many of the juniors. They will have to, just if they are going to use many AI agents, they will have to hire just a few guys that can do the work. But they, particularly, are going to be on the pipeline to seniority, and they have to be exposed to other types of tasks. And we can talk about what that could be.

And, on the university side, we will have to deliver workers who are not at 60% and are closer to the 80%, to where the AI is actually getting.

So, we will have to teach people not just the basics, which you always need in order to learn to think, but we will have to do part of the apprenticeship ourselves.

Richard 18:58:13

So, I mean, if I was an investor, which a lot of our listeners are, and, you know, I was thinking about, you know, this, you know, we're long-term investors. So, short-term, this could be quite beneficial and profit margins go up. But, over the long term, there's a question then if that talent is not going to be there.

So, maybe it's a little unfair, it's not really your area, but what would be some of the questions that I would be looking to ask management about what they're doing for the longer term, in terms of getting around this this?

Luis 19:35:10

To answer this, I would go back to the first question: the messiness. The book is really about the messiness of implementation and the messiness of real-life work bundles. Jobs are not these clean, simple tasks that you just do repeatedly. And when, you know, when you hear the lab bosses talking about, "Oh, we automate office work. We're training people for computer use." I mean, anybody who's had 20 minutes at this job knows that there's no such thing as computer use. What these people are doing every ten minutes is different. And the training that you would require to understand each individual one of those of those jobs when they are messy would be enormous. And you're not going to get there.

So, to me, the key aspect you want to be looking at, as an investor, is implementation. You want to be looking at, in this messy context, in which companies have complicated processes with jobs that have multiple components at multiple tasks. You want to look at how are they confronting the messy reality of everyday work? How are they implementing?

My sense is that the bottleneck down the road is going to be in implementation. I don't think the bottleneck will be on the large language model. So, if I was an investor, I wouldn't be putting my chips on the big labs. Let me not name any. Because I think that, at the end of the day, that does tend to commoditize as people, kind of, train and get the data.

I mean, getting data is very hard. And getting data from those who already have the data, who are the more advanced labs, if you are the less advanced one, is not hard. You can do distillation. Whereas on the other hand, the companies that are able to use the AI properly, and all of us are using AI and all of us know how hard it is to actually get the right outcome.

We talked about 80%, getting the 80%, how easy it is, how hard is to get to the top. You have to decide which output you're going to stop early and which outcomes you're going to really go all in for 100%.

All of those decisions are complicated. And all of that messiness of bundling, and re-bundling with the jobs, and taking care of the careers, all of those are going to be where the competitive advantages will be and where the differentiation will be.

I don't think the differentiation will be on the LLMs themselves.

Richard 22:23:25

And, how do you think this, sort of, compares to past, you know, the industrial revolution or the internet boom? I mean, one was about muscle, one was about, I guess, intelligence as well, or information in the internet boom. What's particularly different from this AI?

Luis 22:40:16

So, the way to think about all of these things is to try to think of what remains as scarce after each round of evolution.

So, when you have the AI revolution, you increase the amount of machines that are available. At the start, the machines are scarce, and there is a big increase in the return to those machines. But, the machines' supply is pretty elastic. So, people can expand, and expand, and expand, the implementation of those machines. And the thing that is not plentiful, and that you can't expand, and expand, and expand, are the workers.

So, eventually, the return to the machines goes back to the normal return. And the returns go to the labor.

So, in the Industrial Revolution, it took a while, but eventually the returns went to the labor. I mean, we know that Engels supposed that, in the '10s and '20s of the 19th century, workers were not better off. But, eventually they were. Their grandchildren or children were doubling their living standards within the following 30 or 40 years after 1840.

The second episode you're asking about: internet.

Okay, so the internet, all the software revolution did, and the internet, etc., was it developed a set of technologies that were complementary to highly skilled workers who had cognitive jobs, that could actually profit from these, and that could use all of these machines.

So, what we saw was a very different result from the Industrial Revolution.

The Industrial Revolution implemented basically all workers because you needed a worker to work with the machine. The internet revolution, what it did was, it complemented only the highly skilled workers.

So, we had a big increase in the demand for highly skilled workers. A job in their routine tasks, both cognitive and non-cognitive, manual and ongoing tasks.

And we saw that the basic, the lowest-skilled people, were unaffected because the service jobs, etc., were neither increased nor decreased, in terms of their productivity.

So, two very different waves.

How do those two compare to AI today?

So, AI has a replay of affecting broadly cognitive tasks. It's going to make cognition more plentiful. And it's going to complement both the cognitively more advanced people, as we said in the robot-below configuration. And it's going to complement, as we said in the robots-below and robots-above configuration, it's going to complement all of those middle-class skills, such as emotions, and physical strength, and all of these other things that AI is not doing.

So, we are going to see a sharply different pattern, it seems to me, than the previous two revolutions, partly because most of these skills are very broadly distributed in the population.

And, we could have a situation where we develop what David Autor has called this "new middle class."

Richard 26:00:27

Maybe we switch gears a bit and pick your brain about regulation. So, you were a member of the European Parliament. You think deeply about this stuff, and presumably you're still advising some members of Parliament. What do you think about the regulatory approach to all this? What is the right way for politicians to think about that?

Are we seeing something quite different from the US approach versus the EU approach? The EU typically tends to be more cautious, I suppose. How do you think about that?

Luis 26:36:12

Let's put on our investor hat, since that's the audience for this podcast mostly.

It seems to me that a huge part of the growth heterogeneity we're going to be seeing in the future years is going to come down to the political economy heterogeneity, how do different places face this and regulate this.

Nuclear energy is cheap and it's excellent. It's environmentally friendly, climate friendly. And yet, we all know that Germany has decided they don't want to do anything with it, and they've decided to just close all the nuclear plants.

A similar situation could be happening here. Technology, people tend to think technology will rule, and if technology is really advanced, then technology will create productivity.

In fact, that's not the case. We see quite a bit of heterogeneity, first with data centers. And this now concerns the US. We see several states which are really fighting data centers, which are introducing data center moratoriums. We see very important politicians, very smart politicians in the US, who are really seeing the political gain in fighting data centers, partly because of the really clumsy, catastrophic handling by the laps of the public communication concerning all of this.

On the other side, on the European side, we have, in particular, yes, on the European Parliament, but also on the Council, we have this European precautionary principle, which is basically Europe. You know, you have cars which are self-driving and which reduce, by multiple orders of magnitude, the risk on the roads. And yet, as long as they kill one person per millions, and millions, and millions, of drivers, it's going to be considered risky and is going to be regulated away.

And we saw in New York, which has a more European approach, they said they're not going to allow Waymo.

So, Europe is adopting, as it has in the past with genetic and with many environmental and AI technologies, etc., this precautionary principle, which is very problematic for growth and which is also kind of very premature.

The truth of the matter is, we don't know. And when you don't know, you have to be cautious. You cannot just go and start talking about teraflops and, you know, that's what we did in the AI Act, and say, like, "Oh, this technology with so many teraflops has to be regulated right away," when, in fact, well, it turns out that all technologies are going to be way above that.

So, it's all kind of a little bit, I mean, I would say arrogant or premature, if you wanted to use some more friendly words, to rush with all of this regulation.

Richard 29:33:01

Good. Maybe on the monetary policy side, I think you do speak to some central bankers. But if you were sort of thinking about the impact here on inflation and interest rates, and Kevin Warsh was talking yesterday in his testimony to Congress about AI and how he's quite bullish on the longer-term impact, on how disinflationary it's going to be.

What do you think about that? Is this going to be quite disinflationary? Or it depends? Or how are you thinking about that?

Luis 30:08:21

So yes, indeed. I have been thinking about monetary policy and talking to people on that side of the debate. We did write a Princeton University book last year on the European Central Bank. I profit to plug with the audience on some of these issues.

So let me just talk in terms of r*, since this is something that you were announcing in your question. The asset supply has been relatively low relative to asset demand over the last period. And that's going to change.

So the asset supply is going to be... Clearly, there is more profit, there is more CapEx, there's more data centers, there's more GPUs, potentially market concentration. So we have all of this additional supply. And from the demand side, we're going to have higher productivity, and future workers are richer. And, you know, younger cohorts potentially can save less if this does what we expect.

So this would point to the opposite pattern that we've seen over the last 20 or 30 years. And instead of a secular drop in r*, to an increase in r*. If you look at the recent Auclert, Malmberg, Rognlie, and Straub, and I've just told your listeners about this paper, but I think it is an interesting one.

They talked about how the deterioration of productivity, etc., over the last years has dropped r* by 118 basis points in their estimation. It could be higher, it could be lower, whatever. And these are very rough numbers. But a reversal here could significantly increase r*. My fear, of course, is: think of countries that don't see changes in G but see these changes in r*. This is my big fear.

So you could see, in Europe, the R-minus-G gap very much expanding. The scenario I'm worried about, or in other places that are not the US. My fear would be: look, you're in the US. You see an artificial intelligence explosion, increases in R, increases in G, but G grows way more than R. That is sustainable, even though none of us are very pleased with this 6% deficit in peacetime that the US is running.

But now think of Europe and think of a world where we regulate this technology away. We don't work on the implementation, we don't work on the data center side, etc. And then we import from the global markets a high r*.

Okay, if the returns are very high in the US because there are all these data centers and all these big, you know, possibilities in individual stocks that I won't mention, but we import that. But you're not going to import the G. And G is the growth of the European economy.

And if G is stagnating, for the same reasons as it was when the information technology revolution happened, and which the Draghi report has discussed, you need to reassign workers and employment and capital, and Europe is not great at reassigning because we have this rigid protective framework. And then you could have this world where we face higher r*s with not higher Gs.

And you can immediately imagine what the fiscal consequence of this could be.

Richard 33:48:01

Yeah, difficult. So, US gets it both, the higher G and higher r*. And Europe gets the higher r* and not the higher G.

Luis 33:56:22

Yeah. In my Silicon Continent blog, I call it R without G, instead of referring to the R minus G, as the syndrome that Europe could be facing.

Richard 34:09:09

So, we need we need the Draghi plan to be accelerated, I think.

Luis 34:12:27

We need the G.

Richard 34:13:22

We need the G. Yeah. Last question.

Luis, you know, you teach students, and earlier in the year you wrote a letter to youngsters, sort of, giving them some advice. You know, if I'm a student today, I'm probably worried that I don't have, you know, the skills, contacts, or confidence out there.

And, you know, machines are now taking the PowerPoint jobs that I probably would have been doing had I first started in my career. So, what are you telling students and youngsters when they come to you and ask about, you know, what kind of roles they should be doing, or what they should be studying, that kind of thing?

Luis 34:53:06

Yeah. No, I mean, I think the summary of that letter was in the subtitle, which was "take the message up."

So, take the message up. That means, okay, you're looking at a job. Someone is offering you a job, and you're trying to figure out, is this a single-task, clean job that is pretty well subject to this reinforcement learning process we're seeing, where there's a right and a wrong answer? And this spreads over multiple different firms. They're doing, more or less, the same thing, like doing a PowerPoint or programming a spreadsheet. That's not a promising job.

The promising job is the one where there is a, which is the message, where you're doing different things, where you're working with clients, there's a big relational component, where there's a physical component, and there is a cognitive part, ideally.

But the cognitive part is strongly bundled with all these other parts. And as a result, there's this messiness.

So, it's not about complexity. Because we discovered chess is very hard. And chess was the first thing that these computers, these machine learning algorithms, did even before LLMs. They trained themselves to play chess, and they go and they beat any human.

It's not about complex. It's not about hard. It is about messy. That is my advice.

Richard 36:21:15

So, we should all try and make our jobs a little messier.

Luis 36:25:06

Or specialize in those.

Think of, if you are in a big company where there's a lot of programming. There are the programmers who, little by little, potentially get replaced if they are doing a clean programming task. There are the server engineers who are going to meetings and speaking about new products and so on. And that's very messy. And if you kind of look at the schedule of one of these people, they would be doing different things all day long.

And if you look at salespeople, you go to the other extreme. I mean, there's a cognitive task, yes, which is you have to learn about the client you're going to do the sales pitch to. But that doesn't get unbundled. You don't get the sales cognition unbundled from the sales. You need to be the person who understands this client. You need to go to the client and make the sale.

So that is a fundamentally strong bundle. The cognitive part and the emotional, relational task are going to remain together.

And each sale is different. And each relationship manager is doing slightly different things. And, it's not going to be just R&:Led, reinforced and learned, away by these systems.

Richard 37:40:23

Well, that's excellent advice and, I think, with that, our time is up. So, Luis, thank you very much for your time today. I think the topic is incredibly relevant to everyone. It certainly is to us in financial markets. And I'm sure our listeners would agree with that. So, thank you very much for your time today.

Luis 38:00:14

I hope they find it interesting. I hope you find it interesting. Thanks for your time, too.