The End of Tax Season Stress? AI’s Game-Changing Role in Taxes

Julien Redelsperger: "And I'm super happy to welcome Daniel Marcous. He is the former CTO of Waze and he founded April, the first AI powered tax system to reinvent how Americans approach their taxes. So today we will talk about taxes, AI and the most complicated aspect of personal finance, taxes and the ever changing tax code across the country. Thank you for joining me today. How are you Daniel?"

Daniel Marcous: "I'm great. Thank you for having me."

Julien Redelsperger: "Yeah, thanks. That's my pleasure. And actually my first question for you would be, how do you feel about filling your taxes?"

Daniel Marcous: "I'd say that it's mostly a love hate relationship. Mostly on the hate side, it's pretty stressful. Even for me as a person that knows his way around taxes, still super stressful. And the love side is mostly when it's done. So it's like the lifting a burden after you've done this, oh, thank God I have another year now until I have to get to it again."

Julien Redelsperger: "So you don't have a personal passion about taxes. Like it's not your hobby. It's not something you like to do on weekends or help friends about it."

Daniel Marcous: "Well, I said that the passion is more about solving something that's a pain and a personal pain and a pain that almost everyone that I talk to also feels. So that's where the passion is. And more for me, I really like complexity. So I love to solve complex problems and apparently taxes are just super complex and diving into that and starting to untangle it and saying, okay, this system is so messed up and I got to like reinvent it and find a way to make it more simple. That's really where my passion lies."

Julien Redelsperger: "Okay, good to know. So why do you think it's still so complicated for most people to fill their taxes, despite all the technological advancements that we've seen today? We have AI, we have tech, we have digital everywhere. Why is it still so complicated for most people?"

Daniel Marcous: "Yeah, it's a good question. So I'd say that in my opinion, taxes are still a labyrinth of exceptions, deductions, rules that keep changing every year. And what tech software has done up until now is mostly just digitizing the process. So instead of filling a form with pen and paper, you can do it online. Cool, that's a bit nice, but it's not extremely helpful. And what we've seen in the recent years was the progress in AI and things like Chet GPT is indeed helping to simplify parts of the process. And I sort of, and we can talk about my background a little later, but I think in navigation for my time in Waze. So I think of AI tools like Chet GPT of sort of upgrading the way that you navigate in tax from using a paper map to some sort of naive GPS that can route you through all of the complexities, but it's still not enough. You still need up-to-date maps and you can't really foresee every twist and turn. And the map, which is the tax law keeps changing year over year. You still need real-time updates, like traffic updates when you navigate. So real-time data that is changing about Julian's financial situation. And you most of all need to take all of this updated law and updated financial data and use that in order to personalize in real time the experience for Julian. So it's not a one size fits all anymore. And there's no friction for Julian if there doesn't need to be. And that's something that does not exist today. And basically I'm here because that's really what we're building in April, trying to solve all of these problems that can't really be solved with just throwing your last year's taxes into chat GPT."

Julien Redelsperger: "And so April, just for our non-US listeners, I assume you called your start of April because that's the month where the Americans are filling their taxes, right?"

Daniel Marcous: "Exactly, exactly."

Julien Redelsperger: "Okay, perfect. And so, yeah, I was saying you spend a lot of time working for Google and Waze. Why did you decide to leave these big tech companies to fund April, which leverage AI to help people with their taxes?"

Daniel Marcous: "So I did really learn a lot and have a really good time at both Google and Waze, great companies to work for, but I always had the passion to start something of my own from scratch, from like bring it in from zero to something that's really, really big. And I love doing something that's impactful for people in their day to day. And I really loved about Waze that I could save people time. That was the mission. And really in April, it's sort of like even better than that for me. I can save people both time and money. So that's a win in my head to try and achieve this mission and help the average person off the street and save them time and money."

Julien Redelsperger: "So how long did you stay for, how long did you work for Google and Waze and what did you do there?"

Daniel Marcous: "Around seven years. And I did a bunch of different roles, mostly around data science and AI. So I started as an IC, worked my way up to manager, then like founded and led the group that does data, data science, data engineering, product analytics for Waze. Then I started the CTO office in Waze and had this CTO role and worked with a lot of like cross-function projects across Google and doing data science best practices, data science processes, how you can get data science from research into production and what's the best way to make things not just stay in your notebook and actually reach Google's products."

Julien Redelsperger: "Okay, okay. So today, I'm sure you know that, of course, but we talk a lot about artificial intelligence. When you were working for Google and Waze at the time, did we talk a lot about artificial intelligence? Was it like a big topic when you were working for Waze as a CTO?"

Daniel Marcous: "Yeah, definitely. That was really my strong suit and where I started my career. So it's a lot of the things I did was to improve current products and like sort of upgrade them from traditional algorithms to use AI. And I got a lot of early exposure to all of the potential that AI and specifically language understanding can bring to different domains all across Google. And I really took that into heart was what we did here in April and really believing before, like before even the term LLM was coined, that we can use natural language understanding to disrupt the tax industry and do something that was not done before. And when we started April in 2021, we used classic NLP algorithms and the practices there, which started to work, showed some good promise. And as the domain progressed alongside April progressing, we ramped up our LLM game pretty seriously. And that's been amazing, like coincidental sort of, but amazing timing for us."

Julien Redelsperger: "So you've been in the field for a while. ChatGPT launched in November, 2022. It was a big shock for most of the world. Were you surprised to see the first version of ChatGPT? Were you, or did you know, did you see that coming? Like what was your mindset at the time?"

Daniel Marcous: "Yeah, I'd say the answer is sort of mixed. Like I was not surprised by the abilities and I were using language models before ChatGPT from, you know, open source, other companies that existed before open AI and things like that. So I was able to see the potential, but then the potential was mostly limited to specific use cases and more work on generating examples and so on. And I think the big leap that at least surprised me in ChatGPT is the way that it does a decent enough job for any random use case. So that was, you know, kind of surprising for me."

Julien Redelsperger: "So you founded April in 2021, is that correct?"

Daniel Marcous: "Yes, correct."

Julien Redelsperger: "Okay. At the time we did not really speak about generative AI and LLMs. I mean, of course in the tech industry, but not for the general public. When you founded April, did you have AI in mind? Did you have some use cases that were powered by AI or is the AI a component that you add later when you develop your company?"

Daniel Marcous: "It's a good question. So it's the former. When I started researching along with my partner, the domain of text and what's the problem there and what's the huge entry barrier in how you need to understand and sort of code the entire tax law in every different jurisdiction and all of the nuances, we figured out that in order for us to do something in this domain, we need like an army of 100 engineers and 100 CPAs to work for around two years in order to file taxes everywhere in the US. And that's not really reasonable. So I've bet from day one on AI and natural language understanding, which was the term that we used at the time on unlocking the power of understanding tax law and translating it into actual code that we can use and we can later run on Julien's data and using the tools that we've started to build in-house in 2021, we were able to increase the efficiency of the engineers and the domain experts, the CPAs that we have in-house by around 10X in like scanning and running over all of this tax law and putting it into our software. And luckily for us, this entire domain of natural language understanding, processing, and then LLMs moved so quickly. So we just improved our efficiency more and more as we went, but it was part of the building block that without it, we could not have reached the coverage and of the features that we have right now in our product. But it's just not feasible with so little working hands."

Julien Redelsperger: "Okay, so today, if you had to do it again, it would require, I mean, if you don't have AI into your process, it would require a lot of people, engineers, accountants, taxes specialists, which now AI can do it instead of hiring those people."

Daniel Marcous: "Yeah, and AI can help. It can make those people more efficient. It cannot, and I believe not in the next few years, at least, it cannot replace them completely, especially in something as sensitive as taxes, where you have to be 100% accurate in what you later file to the IRS and the jurisdictions. You still have to have humans that code your software, so your software ends up deterministic, but you can make them 10X more efficient with AI in the loop."

Julien Redelsperger: "Okay, so before the interview, you told me that for you, taxes have long been a zero-innovation industry. That's tough. How do you think AI will disrupt the way Americans approach tax filing?"

Daniel Marcous: "So I think in many ways it can disrupt, and we can talk about different things here. So there's the more superficial aspect of tax filing, and how we can make it easier, less friction, take minutes instead of hours, and so on. And that's a big unlock in its own, but I think the bigger unlock, which is exactly why we're gearing April towards that direction, is democratizing personalized tax planning, advice, and optimization. So we're not only reactive in filing your taxes, but we're actually helping you to optimize your financial situation by knowing you and knowing the tax law in extreme depth. And then we can be proactive and really render you a service that's sort of like a personal tax advisor that the very, very wealthy people already have these days, but it's all based on humans, and people like us can't really afford that right now. So we're trying to democratize this type of service for everyone, and I think that's the biggest potential in taxes."

Julien Redelsperger: "So do you think AI is more efficient at filling taxes when you compare it to human accountants? Can AI avoid mistakes, or miss opportunities to get some money back, for example? What's the relationship between how AI works on one end and how human accountants work with taxes?"

Daniel Marcous: "Yeah, so obviously there are pros and cons for both, but if you put AI in the mix, then I think that you have a lot of advantages that were just not feasible before. You have the advantage of knowing the entire breadth of the tax law and all of those different nuances and all of those different deductions and credits that you can find for someone. And even if you have like a super qualified accountant that knows almost everything, keeping up to date with the law that changes year over year is just impossible for a human to do. So that's, like knowing everything and having all of this context of updating tax law really enables you to unlock as much value for the user as possible. And then I think the second thing that's really something that couldn't have been done before is sort of opening this black box of taxes and doing something that's traditionally in machine learning and data science industries called explainability, and explaining how you got to a certain number. Like why this is my refund or why do I owe so much taxes? Why are you asking me this seemingly random question? How does it affect my tax situation? How does this decision affect my tax situation? And the funny thing is that if you take this kind of like anecdotal problem and you go to three different CPAs, you'll get three different answers. But the tax law is deterministic, right? It's law, it's deterministic. There shouldn't be three different answers. So really untangling that and understanding the full sort of like decision tree of taxes and explaining it in human language is something that's only been possible right now. And up until now, it's just a crazy black box that everyone interprets differently."

Julien Redelsperger: "And also I'm thinking about inclusivity and sometimes some people might have, might lack some knowledge about taxes. It's very complicated. And not everyone has the financial literacy to understand how it works. Do you think in that way, AI could be more inclusive to produce personalized answers, to guide people more effectively than a human would do?"

Daniel Marcous: "100%, yeah. Tax by nature is so personalized because as mentioned earlier, you can imagine tax law as sort of as a decision tree. So if Julien leaves here, then this applies to you. If you're married, this applies to you. If you have two kids and their ages are this and that, then this applies to you. So what we're used to seeing in tax software is a one size fits all solution. We're gonna ask you a thousand questions that are just not relevant for you instead of real time updates and personalization and making the experience contextual to your data, which is a thing that AI excels in, adding your like Julien specific data into the prompts. If we get a little more technical and merging that with the knowledge of tax law is able to render an experience that is seemingly like you talking to an accountant that has known you for years instead of just using like a one size fits all software."

Julien Redelsperger: "And a content that's not gonna charge you if you spend 10 minutes or three hours."

Daniel Marcous: "Yeah, exactly, exactly."

Julien Redelsperger: "Do you have a lot of accountants among your friends?"

Daniel Marcous: "Less and less so, but yeah, yeah, I do, I do."

Julien Redelsperger: "What do they think about your solution? Like, I mean, more broadly, my question is about the future of a accountant. Like, will they still have a job in the future?"

Daniel Marcous: "Yeah, they'll have a job in April." [both laughing] "You know, but more seriously, it's a good question. And a lot of people do ask me this question and there's been like really coincidentally last week, there has been a Wall Street Journal article that was talking about the mass decline in the number of accountants that exists in the US right now specifically talking about the US relative to the growing population and the growing demand for accounting services. So there's a real gap that's being formed here and is getting larger and larger. And April is trying to fill that. And I do believe that the accounting job will still remain in the future, but it will change. And what we're already seeing in the past few years is accounting is going more and more upmarket to where they can add value, do the more complicated use cases that are still extremely hard to imagine for AI and means a lot of creativity to serve people that really care about this human relationship. And there is a room and I believe that for the foreseeable future, there's going to still be room for that. So the job is still there, but it is rapidly changing."

Julien Redelsperger: "Okay, okay. And I think one of the challenges maybe for the accountants would be how to learn a new path toward AI-powered accountancy."

Daniel Marcous: "Right, right."

Julien Redelsperger: "Make sure they are guided in using AI to probably do a better job."

Daniel Marcous: "Yeah, that's exactly right. And we are seeing today some tools for accountants that are powered by AI. So when accountants can take, let's say, a hundred clients during tax season instead of just 10. So, you know, it's a win-win for everyone."

Julien Redelsperger: "Okay, okay. So you told me that filling taxes in the US takes about nine hours for most Americans."

Daniel Marcous: "Yeah, unfortunately."

Julien Redelsperger: "How does AI and, after all, your solution cut down on the time? And what part of the process is the most difficult to streamline?"

Daniel Marcous: "I'd say that the most effective way to cut down the time is personalization. So if you take down all of the non-relevant, one-size-fits-all pieces of information that tax software or accountants usually collect to get to know you, then you can dramatically cut down the time. And the way for you to do that is to use real-time personalization. So whenever April, for example, collects a new piece of data off Julien's, whenever you add something like a piece of income, a file that you upload, like a form that you got from your employer or so on, then we update our decision tree behind the scenes, we update the context that the AI sees around Julien, and we have our understanding of the tax law, and we match these two together in order to understand and rationalize what's the best next step specifically for Julien, for Julien's situation. And doing that dramatically cuts down the friction that you have with interrupting with something like taxes, because we just don't ask you what we do not need to know and what we do not need to know to file your taxes."

Julien Redelsperger: "And what does AI cannot resolve in terms of taxes? Is there stuff that it's still too complicated right now to work with AI or anything that you say that it would be impossible because of barriers or technological perspective or, I don't know, some form of resistance among the tax industry?"

Daniel Marcous: "On the good side, tax law and understanding tax law is a perfect use case for AI because it has a lot of patterns. Like law specifically has patterns, tax even more so. So understanding those repeating patterns is something that AI is really good at. On the other hand, though, tax has a lot of varied form structures and not just free text that's written as clear patterns, but you have all of those different forms and boxes and extracting data from the right place. And then different types of financial institutions that give Julien data that they need in order to file their taxes do that differently. So even every employer can issue a different type of W-2, which is the form that you get with your sum of wages. So understanding those varied format that are not rigid at all, it's something that AI still has a very hard time to do. And I think the worst problem that I'm seeing in the industry today, specifically applied to taxes, funny enough, is that people do really still have a misconception of how magical AI is. And you do often see that around Twitter where people just like post a random picture of something they asked and they seemingly got a good answer, but people sort of make that into a belief that they can just upload my files or something to ChattyGPT and it will just magically solve my taxes. But there's so much nuance to that. So many different laws, so many terminologies, compliance, working with the IRS and the state authorities in order to put the data exactly in the format that they expected in their backend systems. So there's really an immense amount of work to do something like that, that's nowhere near what AI is capable to do these days."

Julien Redelsperger: "Okay, so correct me if I'm wrong, but I believe that in the US, it's like in Canada, you have like two layers, like the federal layer and the state layer."

Daniel Marcous: "Exactly."

Julien Redelsperger: "That means your solution needs to know how each state actually works with taxes, is that correct?"

Daniel Marcous: "Yeah, yeah, which each state does their own thing and it's sort of like their own country. And in the US, you even have more layers sometimes. So let's say New York City, it's a big enough city that has their own layer. So if you live or work in New York City, you have New York City tax, New York State tax, and US federal tax."

Julien Redelsperger: "Oh, okay. And to navigate through this sometimes cumbersome situation, you emphasize that being proactive year round is a must to save time during tax season. So what does it mean? What kind of good practices or tips can you share to make sure that AI can help users be prepared through the year for filling the taxes?"

Daniel Marcous: "Yeah, so I'd say that the more the AI can get to know the person, learn about the person, learn about their financial situation, the more effective it can be because it can personalize what it does. So the more interaction points that you have with someone, the more you learn about them and the more value that you can generate for them. So if you find more interaction points and you need interaction points with users that bring them value, otherwise they won't just interact with your software. So we are finding all of these interaction points throughout the year on where we can add value to Julien in order to get to know you better, optimize a certain aspect of your taxes, and along the way, learn something about you that we can later use to file your taxes end of year. So you won't have to put that in the system again. So we'll already know you. We already can personalize things for you. And that saves you tons of time last year. So it doesn't feel so cumbersome. So even if it takes, let's say, 40 minutes in the end of year instead of nine hours, because you're using like a more advanced software, and we've already interacted with you in three occasions during the year, each time for like five minutes, we can save you half the time because we have a lot of the data that we already need to file your taxes. So this experience of accompanying you really generates a lot of value that's compounding as we get to know you more and more."

Julien Redelsperger: "How do you deal with more complex situations? How do you integrate with third-party software? I'm gonna give you an example. Like if I'm buying Bitcoins, for example, on a third-party platform, if I am the owner of a house in a foreign country, how do all these complicated situations work with AI to make sure the taxes are always compliant and contain no mistakes?"

Daniel Marcous: "Yeah, good question. So there's a few aspects to how we approach solving something like that. So one aspect is understanding tax law in general, and the way that we've built our AI system is that different tax use cases are just different use cases, and you're trying to find patterns that you've learned from other use cases that you've already tuned. So sort of going through more complicated use cases is just more work for the AI, and we really rapidly cover increasingly complex use cases that are most likely built on top of the simpler use cases that we already understand. So I'd say it's just like a linear process in understanding more and more tax law. And when you use something like AI that breaks down the law into this decision tree or graph sort of form, it's just a linear amount of effort, and it's not actually more complicated for a software like that. But then the second piece of how do I collect the right data and how do I integrate with, let's say, third-party platforms to collect this data and reduce the friction for Julien, that's another very interesting part that we try to architect it from day one to a way that we have our knowledge engine powered by AI, like the knowledge of tax law, completely separated and isolated from the knowledge of Julien's financial picture and financial data. So we keep them siloed from one another, and our way to create the applications is just like apply tax law, which is code, that's like the knowledge engine, apply the code onto your data in specific ways. So when we learn more tax law, we add this knowledge repository, and when we learn more data about you, we add that in a structured manner to this financial picture of you, and then we sort of tie the two together to apply it to different tax use cases where we unlock them in our software. Okay, now you can also do this. Now you can do that."

Julien Redelsperger: "Okay, so we talked a lot about AI, obviously, but my question for you would be, do you use third-party AI that you do not own, or did you create your own model, your own system that is proprietary for April? Like, how do you work with AI, and what's the backbone of the AI system in your company?"

Daniel Marcous: "Yeah, so we use a mix of solutions, and I think a mix is really, like when you're trying to innovate something and differentiate and not have the same answers that you get online everywhere, you really have to use a mix of things. So we're using platforms like OpenAI, and we use our own in-house models. And the most major aspect of it is that we're not just using models, but we're using a lot of logic that's led by engineers and domain experts to have domain experts crafted algorithms on top and around the AI. So the system that we end up developing is a system that has tens of moving pieces in it. You don't see it as a user, but behind the scenes, there are a myriad of models that are running, there are deterministic algorithms that we wrote that are running, and you actually get the crafted answers and the crafted experience without seeing all this complexity behind the scenes."

Julien Redelsperger: "Are you concerned about data breaches, privacy, data security? It's a big topic these days."

Daniel Marcous: "Yeah, I believe that it should be a concern. It always should be a concern. Specifically in fintech and financial data and taxes, privacy and security is always a really, really big concern. And AI is adding another surface area that you can get a breach from. So you really have to take that into account as a new piece of software that you need to protect properly and not really ignore its risks. It has risks. But luckily, as the AI industry is maturing and it's maturing extremely fast, then you see more and more built-in solutions in order to take care of privacy and security concerns. So you have things like OpenAI that have their own security certificates and have their own settings on how their models do not learn from your data and things like that. And those are critical for anyone that uses third-party AI in their tools to understand how to use and how to configure. But once you do, it's really solvable."

Julien Redelsperger: "Okay, okay. So you mentioned working with individuals for taxes. What about businesses? Do you work with corporate, SMBs? How do you plan to expand April’s reach?"

Daniel Marcous: "Yeah, we do, we do. We work with SMBs these days and we file their business taxes as well. And for us, we really look at it as just another tax use case. So our tax corpus that codes the knowledge also codes the knowledge of business taxes. And the financial data that we collect, instead of collecting Julien's data, we also collect Julien's business's data that we need to file their taxes. And then it's really just a simple extension for us in terms of software. And I’d say the heaviest piece of lift that we have to do in order to start supporting businesses is more the understanding that it's a different user persona and they have different needs, and we have to adopt the UX and the user journeys and the experience to something that's more fitting for a business owner, or let's say like a CFO of a business that wants to do that and not you as a person doing it for your household."

Julien Redelsperger: "And in the future, do you see April working with big corporations, big enterprises, or is it just too complicated for now?"

Daniel Marcous: "No, it's not too complicated. I'd say it's a matter of time and we have so many more ideas on what to focus on in the meantime. So that's really just a game of sequencing, I'd say. So many different things to disrupt in tax that we should get to all of them in the next 10 years, but I can't really say what's gonna be in the next three years."

Julien Redelsperger: "Okay, so that was my next question, actually. Like, how do you see yourself in like a year, three years, ten years? How do you see the tax industry evolving in the future, and specifically with the influence of AI, of course?"

Daniel Marcous: "Going back to the beginning of the interview, I think that our mission of disrupting the way that people interact with taxes is achievable these days and we're rapidly closing in on it to make sure that the way that you interact is personalized, is optimized for you, that we save you the most amount of time and money, and that you can get a service that you could not have gotten before without paying an immense amount of money for a personal financial advisor. And more than that, we're working on meeting you exactly where you're at, which is why April took this embedded FinTech approach. So we don't only want to file your taxes or go into the April app and optimize your paycheck, we actually wanna be there when you make a financial decision, like you sell a stock, you take a loan and so on, and show you the tax implications that it might have for you and recommend you, hey, you know what? You should actually tick this or that box because then your tax liability is half of what you would have done elsewhere. And I think that's like a big game changer because now taxes are just the black box that you're just gonna think of in April next year and that's by far too late to make an impact."

Julien Redelsperger: "Okay, okay, okay. So how many people work for April today?"

Daniel Marcous: "We're around 60 people, I believe."

Julien Redelsperger: "60 people. And do you know the proportion between like tech people and non-tech people, accountants, or tax specialists?"

Daniel Marcous: "Yeah, it's around half are engineers, like 50% of the company are engineers. And then I'd say around 10% are tax professionals, which are like 100% focused on building software. So it's not like the tax professional that will file someone's taxes. It's the tax professional that is paired with an engineer to help build the most correct, accurate, optimized tax software."

Julien Redelsperger: "Okay, so last question for you, Daniel. What's the biggest misconception people can have about AI in taxes? And how do you aim to change that narrative?"

Daniel Marcous: "I'd say that the misconception is around the maturity of AI for many different use cases, taxes included. And it's also kind of dangerous when this misconception happens, because sometimes people just go on things like ChatGPT and ask really complicated tax questions. And I have a few examples of these, and you see that the output that you're getting looks very professional. And the language it's written in is also very confident that it's correct. But sometimes it's not correct, and you might make bad financial decisions on top of this piece of advice, which is really dangerous for you. So I think that's what's starting to happen in the industry these days, which we're a part of, is more of this, what we call vertical AI. So you develop this type of AI for a specific vertical, and more than that, not just for FinTech, but for taxes, and more than that, for individual taxes or for SMB taxes, 'cause there's a lot of different nuances, and we do not wanna take risks with people's money."

Julien Redelsperger: "So bottom line, do not trust ChatGPT to fill your taxes."

Daniel Marcous: "Yeah, you can trust it for a lot of things, but not to fill your taxes."

Julien Redelsperger: "Okay, perfect. Well, thank you so much, Daniel. So at the end of each episode, the guest answers a question posed by the previous guest. After that, you'll have the

opportunity to ask a question for our next guest. Are you ready?"

Daniel Marcous: "Yes."

Julien Redelsperger: "Here's your question, courtesy of Mike Todasco, who is the former Senior Director of Innovation at PayPal. We can listen to his question right now."

Mike Todasco: "I wanna ask the next guest, what would it take in the evolution of AI to have their job eliminated?"

Daniel Marcous: "Well, let's start and break down what I do in my job. Then I guess I do a lot of decision-making, and decision-making around technology and products. And I do that by collecting data and talking to people. So you would need, like, not just a simple interface, but some sort of agent that can interact with people to collect relevant data, and then do some decision-making. I guess that we're getting closer and closer to that. But that agent also needs to identify, like, the next sets of problems that they need to solve, which I think is something we're still pretty far from. Like, what's the, for example, April, like, what's the next best thing for April to focus on, or what problems are we seeing right now that we need to solve? So I think that it will take a little more time for it to identify problems. But if I handed it, like, here’s the problem, go talk to these people and come up with a proper solution, I could pretty much automate a lot of my job. I would love for that to happen."

Julien Redelsperger: "So you feel safe?"

Daniel Marcous: "Yeah, yeah, I feel safe. As I said before, for accounting, I think jobs will change, not eliminate, but, you know, change and be powered by AI."

Julien Redelsperger: "Okay, sounds good. All right, so now what question would you like to pose for the next guest?"

Daniel Marcous: "So a very timely question. So given the recent and somewhat debatable uprise in the abilities of LLMs to really reason and truly start powering agents like GPT-01 model, what do you think is the next industry that can be disrupted now, given those improvements that couldn't have been disrupted a month ago?"

Julien Redelsperger: "Great question. Thank you so much, Daniel. It's been an absolute pleasure speaking with you today. Thank you for joining me."

Daniel Marcous: "Thank you, it's been my pleasure."

This transcription was generated by an artificial intelligence tool. It may not be 100% accurate and could contain errors and approximations.‍