Animal Spirits Podcast - Talk Your Book: The Two Biggest Stories of the Year: AI & Tariffs

Episode Date: November 8, 2025

On this episode of Animal Spirits: Talk Your Book, ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Michael Batnick⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠��⁠⁠⁠⁠⁠⁠⁠ and ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Ben Carlson⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ are joined by Ara Kharazian, Economist at Ramp to discuss: how companies run their finances, trends in artificial intelligence usage and why tariffs are so confusing. This episode is brought to you by VanEck. Learn more about the VanEck Rare Earth and Strategic Metals ETF: ⁠⁠http://vaneck.com/REMXCompound⁠⁠ Find complete show notes on our blogs... Ben Carlson’s ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠A Wealth of Common Sense⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Michael Batnick’s ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠The Irrelevant Investor⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Feel free to shoot us an email at ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠[e-mail]⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ with any feedback, questions, recommendations, or ideas for future topics of conversation. Check out the latest in financial blogger fashion at The Compound shop: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://idontshop.com⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Investing involves the risk of loss. This podcast is for informational purposes only and should not be or regarded as personalized investment advice or relied upon for investment decisions. Michael Batnick and Ben Carlson are employees of Ritholtz Wealth Management and may maintain positions in the securities discussed in this video. All opinions expressed by them are solely their own opinion and do not reflect the opinion of Ritholtz Wealth Management. See our disclosures here: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://ritholtzwealth.com/podcast-youtube-disclosures/⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ The Compound Media, Incorporated, an affiliate of ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Ritholtz Wealth Management⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, receives payment from various entities for advertisements in affiliated podcasts, blogs and emails. Inclusion of such advertisements does not constitute or imply endorsement, sponsorship or recommendation thereof, or any affiliation therewith, by the Content Creator or by Ritholtz Wealth Management or any of its employees. For additional advertisement disclaimers see here ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://ritholtzwealth.com/advertising-disclaimers⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠. Learn more about your ad choices. Visit megaphone.fm/adchoices

This episode is brought to you by Van Eck. Rare Earth metals are the hidden backbone of modern technology and defense, powering everything from smartphones and electric vehicles to fighter jets and wind turbines. Van Eck recognizes early, launching the rare earth and strategic metals ETF ticker, REMX, 15 years ago, well before supply chain security became a global priority. Today, China dominates the production and refining capacity of rare earths, creating real challenges for global supply security, as these materials are essential for technical. technological innovation, clean energy, and national security. That's why countries around the world are racing to build their own supply chains and

reduce reliance on China. As this global shift continues, investment in the rare earth ecosystem is growing rapidly from mining to advanced manufacturing. Investors can access this powerful trend through REMX. Visit vandek.com slash REMX compound to learn more. Welcome to Animal Spirits, a show about markets, life, and investing. Join Michael Batnik and Ben Carlson as they talk about what they're reading, writing, and watching. All opinions expressed by Michael and Ben are solely their own opinion and do not reflect the opinion of Ridholt's wealth management.

This podcast is for informational purposes only and should not be relied upon for any investment decisions. Clients of Ridholt's wealth management may maintain positions in the securities discussed in this podcast. Welcome to Animal Spirits with Michael and Ben. today we are joined by Ara Karazian, who is an economist at Ramp. He writes a subset called Ramp Economics Lab. Fascinating stuff. They sent us some of their work, some of their, some of their charts, and we said, okay, we got to talk to this guy. It's really interesting stuff. You might be familiar with Ramp, and you're not sure why the Sequin Barclay is doing commercials with them, unfortunately, which is neither here nor there. They are the fastest,

I think they're the fastest Fintech company to reach a bill. million dollars in annualized revenue. And they are doing so. How? Because essentially what their business is, is making expense reporting, automating business and all the annoying stuff that business don't want to deal with, receipt upload, credit card, whatever it is. They do that and obviously they do it really well. So what that gives them is the insight and the actual data to look through to their customers on where they're spending money. And today, we talk about the AI category. What are businesses spending on, which businesses, which sectors, the discrepancies

between government data, what they're looking at. So lots of fascinating stuff in here. Yep. So here's our talk with ARA from Ramp. Aura, welcome to the show. Thank you. Thank you for having me. All right, all that anybody is talking about these days, economists, individual investors,

everybody, is this AI explosion. And we want to know, they want to know, is it a bubble? Is it going to crash the stock market? Is it going to take down the economy with it? And it is a fair, legitimate question. It is a new technology, a new industrial revolution. The size and scale is enormous, the growth, the adoptions. Invidia just crossed $5 trillion in market cap today.

it's about to eclipse the entire Japanese stock market, like all of it. And so I'm super excited to talk with you today because you say, well, actually, if you really want to know what's going on, the best indicator to look at is how much businesses are spending. So you have something that you've built called the Ramp AI Index and you break down the different companies in terms of their size, what they're using, what they're spending. What does the data say today? Well, yeah, it's, it's, you're getting at something very important. about this conversation, which is that when we're talking about whether or not something is a bubble,

I feel like people are grasping for all the different data sets that they can possibly use to support their answer. But frankly, for Ayrite, there's just not a lot of data available. Most of these companies that develop the technology are private. So there's not a lot of data you can get from public markets in terms of what is otherwise shared publicly. It's typically shared by companies or individuals that have some stake in the game. So you're not really sure what you can trust out there. So I'm an economist that ramp to financial. operations platform that is used by everything from small startups to Fortune 500 companies to manage their operations. So I can see everything they're spending money on, everything except for

essentially payroll. So that includes AI spend. And what's really unique about this data set is that you can see not only the companies that they're spending on, so you can see if they're buying stuff from Open AI and Anthropic, but you can see whether or not those are actually large investments or if they're just small pilot programs here. I think that's the key thing about the moment that we're in. A lot of companies are trying out AI, but are they using AI in a long-term, concerted way? Because that's the real question about whether or not we're in a bubble. The real question is, is the technology that's being developed providing benefits to the potential buyers such that they'll be likely to buy more in the future? And they're actually going to see productivity gains from this. And for what it's

worth, when we look at where business are spending, that is exactly what we're seeing. Retention of AI products is higher. 80% now in 2024, about 50% of 2020. And AI contract sizes themselves are growing. We estimate that they'll hit about $1 million next year on average. So, I mean, are you seeing the growth exploding or is this more of just a slow burn over the years? Because it seems like this, I don't know, the enterprise stuff is just sort of being rolled out now. Well, it depends on which side of the market they're in. For enterprise, I think growth has been a little bit slower.

We talked to a lot of enterprise firms that say they're hesitant to adopt area across their firm because they're worried they won't be able to trust the results, especially if they scale it to across all their customers, things even like a customer service chat bot or someone you might talk to on the phone, there has intended to implement AI if it causes, there's a significant amount of liability at the enterprise level. That said in 2025 as well, we saw a significant run-up in the actual adoption of AI by companies. We're now seeing about 40% of companies on our platform have some significant AI contract on their books. Oftentimes those contracts aren't just for the main model companies. I'm talking about open-a-anthropic. They are for platform-level providers,

customer service agents, AI agents that are working across the tech sector or for tooling that's developed specifically for a team at the company. So it might be an AI tool for a finance team or an AI tool for a software engineer. Customers aren't necessarily going to see that, but they are producing productivity gains for the businesses that implement them. All right. So going back to that Ramp AI index that we're talking about, you break it down in a really neat way.

You look at the different sectors that are using it from technology to healthcare and everything in between. You look at the size of the businesses. You look at the different models that there. using. One of the things that I want to talk about is your estimate is at 44 percent. And we're going to talk about RAMP and the visibility that you have into the companies that you're serving. Why is there such a gigantic discrepancy between what you estimate in terms of overall adoption rate and what the government is estimating? They're at 9% and you're at 44%. What is happening?

Great question. So we are using actual business spent to track business adoption. Look, there's going to be differences between any different data set when you're doing these comparisons. The government estimate of AI adoption from U.S. businesses, I do think, has one significant flaw. First of all, it's based on a survey. And it's literally based on a survey that goes out every two weeks to businesses. And the question was written a couple years ago. And after I say the question, I think you'll kind of start to understand what's so weird about it. It's literally a question that asks businesses, in the last two weeks, have you used AI to produce goods and services?

Goods and Services is kind of like an econ speak term. And it was written a couple years ago. At the time, I think that question maybe made sense. We're getting at whether or not, or are using AI to produce things in your factory in a productive way, something like that. It hasn't really caught up to the way that most organizations are using AI to produce goods and services.

Does that include customer service automation by AI? Does that include a significant number of your software engineers using AI to automate their coding tasks? Because that's most of the way the AI product ecosystem has developed since 2023 when they started asking this question. And yet the question is written almost in a way that implies, hey, are you using AI to produce widgets in your factory? And when we've talked to businesses, they say that, yeah, this question is a little bit

confusing. So I think for that reason, the government is likely underreporting the level of AI adoption actually happening. The truth is, I think Ramps Index is also slightly underreporting AI adoption because RAMP's estimate is coming from just paid usage. And we know for a fact that there's a significant number of free options now available in the market. Google started integrating Gemini for free into all of their enterprise workspace plans. We similarly know chat GPT is used by a lot of employees who aren't necessarily using their employer plan.

So AI adoption is happening. You can draw lines about like what significant adoption is and where you want to say like, yeah, this business is fully like plugged into AI. But if you look at the government estimate, you know, they were at 3%. in 2023, it's not only about 9% in 2025. That seems really low. And I think you can point to the question being the reason. So for these reasons, I'm generally a fan of government data. I'm not one of those people who's weird to say you're a fan of it. I'm not one of those people who sort of criticizes it. But for this specific question, we have better data sets available and spend

datasets, for example. So I prefer to answer it with this kind of question. So one of our long-running bits here is that we're an anti-survey podcast for that very reason. You can you can kind of nudge people in different directions based on how you ask the question or who you ask. So that makes a lot of sense. So I'm curious, just to back up a little, you kind of talked about this a little bit, but your co-founder, Kareem, was on Patrick's podcast recently. That's like the best, explaining what Ramp does. But maybe you can just talk about what Ramp does just through the lens of the types of data sets that you're able to look at as an economist. And I'm sure for you, this has got to be great because you must have a ton of new and different data than you've

ever had access to before. Yeah. And short, the company's product is a financial operations platform. So it's automating work that an accounting team might have to do or finance you might have to do. But as a consumer, most of us interact with it with the expenses you have to file at the end of every month being the worst time of the month for you. Ramp's whole product is about automating that process. So instead of having to sit down and file your expenses and write it down where you spent money, you just take a picture of your receipt and the company software will analyze it for you, save it, file your expenses. You don't have to think about it. That's amazing for me as a reason.

researcher because I see line item receipts, which most transaction level data sets just don't have. So that's what allows me to do most of this work. I see everything that's on a receipt of as far as where a business spends money. So who, what, when, where, what did you spend on? Where did you spend it? And then what was the line item? For the AI work, that means that we're not just seeing whether or not a company has a chat GPT subscription. It means we can track API spend. We can track whether or not the actual dollar amount. We can track whether or not these contracts are renewed. We can track whether or not they're using multiple different AI companies at a time and track the rise of things like Elon Musk's XAI GROC, for example, whether or not that's

getting the business adoption that people are talking about. Now, the whole point of all this, people always ask why would ramp hire this kind of role? Because my entire job is to be public facing is that the whole company's ethos is around allowing you to have access to better data about where your business spends money. Because ideally, that would help you make better decisions as a CEO or CFO. I don't work on the product. But I do have some feeling that if I am able to put out important data about where businesses are spending money, that would help resolve some questions for businesses that

are on ramp and that are not on ramp who want to know, hey, what should I buy? If I'm looking to get into AI and I acquire software, like, where should I start? What are the leading companies? What are companies like me buying? How do I make these kinds of decisions? because they're genuinely very hard in a very fast-moving software marketplace, where typically a business owner, if they want to know what software they buy, they have to go to like a McKinsey or a Forrester and pay like a massive engagement fee,

and they still never get a data-driven take about what businesses like them are spending on. And so really when I think about my job, it's, you know, how can I produce work that helps make other companies more productive by informing the public about what the most productive companies in the world are spending on and investing in. That's really my research world. Do you think that, and this is not a data question per se, but do you think that these lines are going to continue to go up into the rights such that they reach like internet penetration

when we're talking about 90 plus percent of businesses using them? Because if that doesn't happen, then yeah, these valuations have a long way down. Short answer is yes. I do think AI adoption is going to increase to the point that we're at 100% adoption across businesses. I mean, there are businesses that will never be fully using AI like in any, you know, concerted way. Like I think most restaurants probably not going to be massive users of AI, at least in the day-to-day operations. But they will probably be using AI for, you know, back office tasks and marketing and finance operations. So even if you don't see it as a consumer, it's still going to be happening. Though growth has slowed, in adoption, the actual size of the contracts and the actual sort of

deepness of the adoption is increasing. So I think you're going to see as the sort of products mature, especially as we start getting more verticalized options, and right now, most companies' adoption of AI is pretty limited to just like chat GPT for all of your employees, but we're really early in the process. I was talking to a group of procurement professionals about what we're seeing in Ramp Data, for example. And I was talking about the main trends. Like, yeah, open eyes up. Like, anthropics growing really quickly.

And, like, the first question I got was, like, what's anthropic? And this is from, like, a group of, like, procurement professionals, like, people who are at companies in charge of buying software. And they didn't know what the second largest player was in this kind of market. And I think it's a very telling example of the fact that we're still very early. And, yeah, like, AI might be a bubble here and there, like, there might be too much KAPX, and we might be building too much data centers. But we're also in our own little bubble of people who are actively following the world of business and tech. And that means that I know what Anthropic is, and I've tried all these different AI tools, but a lot of people have had very minimal interaction with them because the products haven't reached them yet. Right.

I mean, what would you have to see for you to say, like, from a software adoption or your AI index, what would you have to see to be like, okay, this is a problem, this is not taking off? Like, what kind of numbers would you have to see to make you worried? If the size of contracts start to go down, and if retention rates start to go down, that means that companies are trying these AI products and services, and they're just not working for them. So that's not what we're seeing right now. What we're seeing is that when a company tries an AI product or service, even when they are otherwise, you know, mindful of, they don't know, no company wants to see like

software budgets explode for no reason. Like they are monitoring and evaluating these contracts over time. If those retention rates, we're going to start, we're to start declining. That tells us that the products that are making it to the market are not reaching customers in an effective way. They're not showing their value. They're not showing fit. They're either too expensive or they're not valuable at all.

Absent that, I can't say that we're on a bad trajectory. Because from the data right now, at least as far as where business they're spending, that's the big difference between this and the internet bubble. It's that the companies that are producing this technology are producing technology that people are buying. Oftentimes they're profitable. Yeah. And they're revenue generating.

That was not necessarily the case in the internet bubble. So I'm not, you're here to say, oh, no one should be thinking about like NVIDIA valuations. I'm not in that kind of world. But I do think this is a missing part of the discourse, specifically because no one has access to this data. All the companies are private, so they don't release it themselves. I was talking with Ben and Josh about this yesterday. For as much as finance people, just talk about the market cap or some sort of ratio, evaluation, whatever. And it's like, guys, we have to talk about what's actually happening. Like, what,

never is brought up. And it's brought up from, certainly from tech people, absolutely. And reporters talk about it, but the financial people often miss, it's like the four million square foot data center that meta is building in Louisiana, you think they're less like dumb? Like they're, they're just, they're just lighting money on fire and they have no idea what they're doing. And person on the, on the Twitterverse knows that Mark Zuckerberg's an idiot. Now, it's not said that they can be overspending. Of course there can be. But there is fundamentally things that are that are happening here. And one of the other sides of this conversation, that I'm sure you have insight to is the amount of money that's being spent on software

or the lack thereof. A lot of these names, at least in the stock market, are under pressure, probably rightfully so. I'm not going to put you on the hot seat and say, like, is this a value or a value trap? Like, what do you think about Adobe, for example? But is the, and we're going to hear from these companies very shortly, are you seeing material slowdowns there that would make you feel better about the fact that a lot of that spend is migrating to AI? Material slowdowns in software companies that are not fully AI companies. Just whether it's Salesforce, ServiceNow, Adobe, workforce, like whatever it is. Generally, no.

I mean, another way I'll say this is that the fastest growing segment and subsegment of RAN data is tech companies spending on other tech companies. It's actually not just tech companies spending on other tech companies. essentially all the, all companies spending a lot on software. In part, that's driven by the fact that there's now so many more AI solutions for companies, so it's oftentimes the software, the AI software themselves. But we're also seeing growth for these legacy software solutions. So there are going to be losers from the sort of software revolution that's happening.

But I think it's way too soon to say where they're going to be. So you obviously work with a wide range of businesses. and in your index model, you look at companies by size. Is it too easy to think that the smaller companies are actually going to be the biggest beneficiaries here because it'll allow them to be more efficient? Or do you think that the largest players are the strong just going to get stronger

and the big get bigger? It's a good question. It depends how they're integrating it. Like I think the large companies that are integrating it typically have a dedicated tech team, for example. So that means they probably have tech people, engineers who are able to take advantage of the AI that's developed for software engineering. And that's one of the greatest use cases of AI available today. It's just really well developed

for software engineering tasks. In addition to that, you can see, you can think about customer service operations. That's another place where AI has done very well, and we've seen enterprise adoption happen. That doesn't translate often to small businesses, but it's not to say small businesses aren't using AI and creative and productive, productivity enhancing ways. like the example I used earlier of like a restaurant probably not using AI serve people or make anything make orders but it's probably using AI to generate marketing copy or like make a quick social media post that they want a picture for that's engaging right heard of that happening often is that going to be a massive driver GDP of course not but it can help set a restaurant apart

in a way that it was not able to do before if they're able to adopt effectively the learn to code crowd has been, obviously, in a world of pain, there's charts that show the amount of software engineers that have been hired, and it's round-tripped, and it's ugly and painful. And do you have any visibility into the labor side of this, or are you just seeing the line item spending? There's really great research coming out. I mean, the question of how AI is going to affect the labor market is re, it's the question in AI and economic today.

That might be the biggest question that we see this century, right? Yeah. It's going to, and just have long-lasting impacts in so many unintended contracts. I think you're right. I think the optimistic view there is that, like, this is going to change the world. We're going to create new jobs. We're going to be better than we ever were before. That is the history of technological innovations. And it's also true that in between now and then, there can be a lot of people displaced and there can be a lot of societal change for the worst for the people that are impacted. And how do we bridge the gap between now and everything's going to be okay? It's, it's scale. Yeah. I mean, I think that you're on the right track. I'm not really an AI doomer. And usually what I think about this is it just seems like we go through this technological cycle every time we have a new technological cycle. And most things in the economy adjust. The economy is very large and dynamic. So I'll say even a lot of the automation happening today, a lot of the concerns about where automation happening is today, software engineering tasks, white collar tasks. Most jobs in the U.S. economy, to they are not exposed at all. Most jobs require some kind of physical component. One of the most common job that does in the U.S.

is just like nurse, waiter, restaurant worker, something like that, some health care worker. None of those jobs are going to be automated anytime soon, if ever. Now, that's not all you need for a recession. You don't need, like, 50% unemployment for a recession. You need, like, just a couple percentage points increase. That's a depression. Yeah. But the early research is that if there are,

is an impact on unemployment. It's happening in a specific few sectors right now on a sort of barely narrow segment of workers who are otherwise able to find other jobs because the unemployment rate has not increased in any significant way over the past two or three years. So so far, it's remained under control and has otherwise delivered some productivity benefits to businesses, but we're still tracking that. Last question from on the AI stuff. Just in terms of this index, it's publicly available. Or you can get it through Bloomberg, but people tweet this stuff. If there's a turn, we're all going to know about it, thanks to you guys. It's ramp.com slash data. It's all publicly available. Everything I do is, that's the whole point

of it. I think it's important for these conversations to have a little more data to them. And you have a great subset called Ramp Economics Lab. Yep. And I'm curious what your process is for trying to answer some of these questions. So you have one that Michael and I have been discussing. It's called why tariffs are so confusing. And I think a lot of people have been left scratching their head going, wait a minute, the economic theory would say that these tariffs are supposed to be inflationary. They should have impacted growth. We're not seeing it yet. And you dug into this and try to figure out, well, why is that the case? And it's one of those, I read your post and I kind of hit myself ahead, like, oh, duh, of course. And your point is that, listen,

all of the announcements and what people are saying, that's not actually what the businesses or consumers are paying just yet. Like, it still hasn't filtered through. So why don't you tell us what you found there? Yeah, so we can see in ramp data tariffs because we see that on invoices when a business is ordering manufacturing goods and they order it from abroad and they're importing. We see the actual tariff line at them or we see like a DHL receipt and it will show the import export duty. And what we found was that tariffs are increasing, like the share of transactions that include tariffs are certainly increasing, especially in manufacturing and retail. But it's much more slower and more granular than it's than you'd think.

Essentially, the average tariff rate was something like 1.4%. Now we're only about 3% as far as share of invoices with a tariff transaction on them. So we're seeing that... That's way lower than I ever would have thought. Right? Yeah, it doubled from last year, but it's tiny. The incidence of tariffs has doubled, but given that we have implemented tariffs on pretty much every trading partner, you'd expect something much more significant.

And the effect is very gradual. Like if you look at, you track the line over time over the past couple months, it's not even And there's no, like, drop line happening when the tariffs get announced. And I think what people miss about the tariff announcements is, I mean, it was unprecedented. Like, a lot of economists were very reasonably concerned when it was announced and remained concerned. But there are a lot of frictions to implementing these kinds of ideas. Like, you need to actually figure out, okay, wait, what is actually tariffed?

Because we have trade agreements that are still in place that tariff announcements for announced in this broad way, but they're not actually legal if they're against the existing trade agreements. So there's that. But there's also people at the ports who have to figure out how to assess the tariffs and be appropriately trained on this. Can the companies effectively hide them or reclassify so they're not paying them? I'm not saying they're cheating, but companies can probably skirt the issue because there's so many different variations or, hey, this product is, does have a tariff, but this one doesn't. Can't they, I mean, You can't have anyone who's a lot of companies pursuing every legal path.

Yeah. Yeah. And then there are also companies that are banking on the, so there's some, some suggestion that of maybe some of these tariffs will be rolled back, right? So there's not a lot of commerce change is actually happening. And then there are companies that we've talked to that say that they are moving their production in response to tariffs, but they're not necessarily moving it back to the U.S. They're moving to sort of the lower tariff nations.

Tariffs that were announced were broad, but they were not the same everywhere. And there are options available to companies. Now, it's not to say that there haven't been significant disruptions. There have been. Specifically, these frictions that I'm describing there are making it hard to collect tariffs are also making it hard to do really any commerce that involves imports exports. Because the companies that have to take things in from abroad are seeing delayed shipments. They're seeing delayed port operations.

All of this stuff takes a lot of time. And for that reason, again, there have been significant disruptions. But also for that reason, it means that the pace of the pace of the, tariffs making their way through the economy has been significantly slower than we think. Price increases haven't happened to nearly the same extent as we'd expect, specifically because a lot of companies haven't started paying these tariffs yet in a meaningful way. Again, it depends on the sector, and there are some significantly exposed ports, but it's been much slower than most of us thought.

All right, what are you working on next and how frequently do you plan to publish your work? I publish weekly on Ramp Economics Lab on Substack, and right now, look, I'm always looking for ideas. So people need to send me anything. Again, my business spend data set really allows us to work to identify what the latest trends are, what's just a vibe that people are talking about in the news versus what is an actual thing that's going on in our economy. Right now, the main focus has been AI and tariffs, of course. But, you know, we follow a lot of trends.

And a lot of the way I approach my work is just trying to figure out what can I confirm or correct. Hey, how about this? What about Chipotle prices in on New York City? Because anecdotally, I mean, I don't go to every Chipotle, but I feel like they peaked a while. I'd be curious to see the data. We asked Chat TBT and they couldn't give it to us. Well, I need the, I wish that the receipts also included the weight of what you bought.

Because that's right what the variability is. I could tell you the price in New York City right now because, like, it's the block. But I can't tell you the price per ounce unless we do like a crazy experience. That's what I said. We got shrinkflation. They're not flowing up anymore. But one of the interesting ones, so you said you've been focusing on AI and tariffs. You have this one that I think is interesting.

And Michael, someone, one of our listeners said that the 996 phenomenon is the new fire in terms of like, there's going to be way more reporting on that. So you had this data that looks at San Francisco workers are working more Saturday. So the 996 phenomenon is, what is it, you work six days a week for nine hours a day. Nine days a week, 9 a.m. to 9 p.m. 6 days a week. Yeah, that's this subculture that I think is working culture that was originally associated with Chinese working. This is virgin culture. let's be clear well actually no right because if you

if you talk to people who do 996 one of the main principles is get married early there's a lot that's right there's a lot of side components of the culture of like working late in the in the subculture where it's like not only do you have to work late and get married early for some reason there's a whole thing about eating a lot of meat

and and working out a lot so I don't know how you did actually look at like the takeout invoices, right? To show that people are, at businesses, are ordering more food on Saturdays, meaning that they're working more food. Oh, this explains DoorDash run. It's the 9-9-6ers. It's true. So when we, like the number of workers, at least in the SF area, working on weekends like Saturdays and Sundays, has increased, we estimate over the entire Bay Area, maybe about 40,000 workers across the metro area are now working Saturdays and Sundays in a way that they were not.

last year. Is that good or bad? Huh. Okay. All right. We'll leave that for the audience to discuss. All right. So, all right, I know you're not on the product side, but as we, as we exit the stage, what is the pitch for people that are listening and they're like, wait, what does RAMP do? And why should I care? Should I, am I a customer? Should I be a customer? What do you guys do? Get your finance team to use RAMP. I am the only spokes of the company doesn't have to sell it, but I use RAM, obviously because I work here. It makes it really easy to do a lot of these finance tasks, both for the finance operations in the back office and also for the actual user. So you don't have to submit a receipt anymore. You don't have to do all that

stuff. It's all automated. We often talk about the effect of AI and unemployment. And the way we see it here, the way I see it here as a researcher is, I would love for AI to take over a lot of these tasks that we just don't need to do anymore. Let AI read your receipts and file your finances for you. Let AI process through what you're allowed to buy and purchase on an online. offsite or what you're allowed to do for DoorDash. You don't think about that stuff. Auditation of all your finance tasks is really the shorthand for it. Okay.

All right. All right. This is great. Looking forward to seeing your work. Thank you very much for coming on today. Thank you guys for having me. It's so much fun.