Talking Expat Taxes #13

Of course. So Alex, thanks a lot for having me. I'm super excited. I think more and more I see tax professionals adopting AI and technology like contrary to what people would think. Like accountants are pretty like. Future forward thinking, and I like that. So a bit about my background. I was originally born in Bogota, Columbia in South America, but I, I became an expat really quick.

I started traveling abroad. I lived in Brazil, I lived in Mexico. Um, I lived in Europe for a bit. Um, and I, I, I'm not a CPA, my profession and I, I ended up in taxes and specifically in the international taxes by accident. I. I was working for a, for a venture capital firm in Brazil for almost three years. Um, and I was a product lead there.

I was doing company formation products and legal and compliance products for founders that would raise venture capital. So, I don't know if you know this, but in Latin America, funds don't invest locally. They invest in offshore vehicles. So this is. Us holding companies or what, what's so called the Cayman Sandwich, which is a Cayman holding company US pass through and then subsidiaries abroad, mainly in Latin America in this case.

So that's, that's how I came related to international tax.

Yeah, so I, I'm gonna tell you a bit about, that's a bit about my background. Then I'm building tax flow. Tax flow is AI for international tax preparing. So imagine a senior tax accountant with ai, right? It's like an, it's an artificial intelligence staff accountant that will help you do several of the workflows, either data intake, which is the information request.

All of the, uh, processing of the work papers, which is getting a bunch of data, unstructured data, and then the data entry itself. 'cause as you know, many of the tax softwares are very legacy, and then you have to input everything manually. There are no APIs for this. So we're, we're tackling those three things specifically.

I, I'm gonna give you a few opinions right there. Like, I think there has never been a better opportunity to be a small firm, because a small firm now with AI can do 10 times or a hundred times more than a mid-sized firm or a big firm, right? So I, I'm gonna take a step back and you had asked, Hey, where do you see the biggest opportunities for AI and tax?

Right? And. The biggest opportunities I see is starting with automating the grunt work, the data entry, the reconciliation, the document collection, collection form, preparation, right? Like those tasks that are very repetitive and high tasks that traditionally would be outsourced, like they can now be done by ai.

While you as an advi like top advisor, advise your clients. Um, so I think we're at a point where we're going from just using chat B to ask questions and getting answers like, as you say, a glorified Google to a place where the interfaces are getting so user friendly that you will be able to connect point A to point B.

For example, using a note ai, like a note, uh, taker on your Zoom calls to do something like, Hey, help me qualify the client. Help me take the notes, help me send a proposal, send me review proposals, help me chase the client, which is where accountants spend the most time, like chasing the client for information, missing information, reviewing documents, like all of that.

I think it's where we're, we're at. Like it's a phase where we're going from. Just getting answers to automating flows, like, and, and, and then eventually you'll be able to stitch them together until you have the full flow. Like, and, and, and I see an opportunity of accountants being in the middle, um, as a human in the loop.

Like, Hey, augmenting their powers, uh, to do the repetitive work. That's what I, I see the most. And how I see this translating into, into the, the taxpayer is like, eventually you as a solo firm will be able to provide. The same level of service that a Big four can offer, but at a 10th of the price, right?

I, I think I can give you my product vision on that. And after, like we're, we're currently working with several design partners and we've figured out a couple things. So. The first one is, yes, it can be overwhelming to say, yes, we wanna implement ai, but how, and in what parts of the process? Or even call it a process, right?

Like I think that if you take a step back and you choose a task that is a better place to stay, to, to start. Sorry. So imagine like saying, Hey, email drafting really takes a lot of time when I do, for example, a, a request, right? 'cause I have to write a hundred emails. Okay? Just use AI or an automation with AI to replace that task.

Then once you've mastered it, go to the other one. Like, is this document extraction? Like for example, I don't know, maybe you're using a desktop application instead of an online application and you have to go take all the client's bank statements and extract them like, okay, get just one tool or mini tool that will help you do that process.

And then when you add 'em up, and especially if those solutions talk between each other, then you'll be implementing AI much faster. Like I, I think it, it, it'll be faster to.

Yeah, that's, that's my take right now. 'cause what I've seen is, w when you try to have everything, then you don't excel at at anything, right? Like

Which, which I think it's the challenge that softwares like practice management softwares have, right? Like they try to do everything and then the customer needs to adapt things and they can't, or like, uh, and, and I mean, that's, that's my take today. And, and this AI thing is being built, you know, it's, it's been probably two, three or four years that this been, this has been happening.

And in the tax industry, I would say one, maximum, two, you know,

I think that's a fair concern, but I probably, I can share what I have seen both from the taxpayer perspective and from the firms. So the first thing I'd like to say is that there's a trusted relationship between the client and the CPA that cannot be replaced. It cannot be replaced by ai and it cannot be replaced by outsourcing.

Right? Like if I hire Alex, I'm trusting Alex. And ultimately when you sign my tax return, like you're liable for it. So that's why I trust you. So the first thing that I would say is like, even though there's technology, um, the taxpayer trusts the CPA, which is why I think the CPA should spend as much time as they can with the client.

'cause that's. Like, that's where trust is being built. Um, the second thing I think about AI as a taxpayer, like coming from the taxpayer perspective as well, is I think, um, it really depends what I perceive that the AI is doing, right? Like, do I think that there's a black box where you throw my documents and nobody's supervising that or viewing that and it's getting filed and you're not liable for it.

Like, I would be scared. But if I see a firm that is using AI to. Check calculations, make things faster. Do reviews, like answer complex questions based on their own firm's knowledge, right? Because if you have other 100 customers that have similar like backgrounds or like structures or like things like mine, I want you to be efficient and be able to, to respond faster or, or whatever.

So. I think that the key component here is maintaining the trust between the CPA and the person, and letting this person know that you're in charge, not the ai, like the AI is helping you out.

And, and, and I think there are three things that firms using AI could tell or could share with their clients that they're u that, that yeah, they're putting in place to keep that trust. So for example, the first one is, Hey, where's my data going? Is this going to, I don't know, an open AI or like anthropic?

Like, is it, is my data being used by a third party? And it's, no. There can be local models that are only run. By the firm. It cannot be accessed by anybody else. Those are encrypted. They have encrypted, uh, encryption protocols. You know, like you can assure the data is safe and, um, like not shared with anybody else.

Yeah. So when you use a chat, GPT, like the public one or like the. The $20 one, which is the pro, like they tell you, look, whatever you share, don't share confidential data. 'cause we might be using this to train our model to get better. But when you're in the enterprise mode that you're using their API, you can tell 'em, Hey, this data cannot be shared with anybody else.

It's only, it can only be used by me. And then you get a version of the large language model that only works for you. So the data doesn't go anywhere else like your, it's your data is secure and it's safe. You gotta tell that to the client, right? Like, it's not obvious. You, you gotta tell the client like, look, this, this is how we're handling our ai.

Right? So that's, that's the first.

Yeah. Imagine like your, your private environment. If, if we put it in like simpler terms. Yeah.

No, it can be either way. But what's not gonna happen is that private data that you have is not gonna share or travel anywhere else. It's just gonna stay with you. So imagine it's like Alex is an ai, he can come here, take a look at the firm's binder, but then he's not gonna take it anywhere else, you know?

So that's how it happens. So the second, like speaking about transparency, the second thing that I think firms can do, or softwares like ours that we're developing do is show an audit trail what was done by ai, what was reviewed by by humans. So for example, this data was extracted by AI on Monday, January the first, that this time, and it was approved by Alex.

Okay. And, and you see the whole audit trail. Just in case you need to go back and check and, and you can tell it to the client. If they ask for it, at least have it, you know. Um, and then the third one that I see super interesting or important is just you can assure the client that there's a human in the loop.

There's someone from your team reviewing each step before moving to the next one. And I think those three will, will build a confidence. Like what I've seen is I, I've heard taxpayers saying like, AI doesn't make mistakes. AI doesn't make that entry mistakes, right? Which can cause. Penalties, high penalties specifically in international tax, right?

Yeah.

I think if, if we can make a similar or a metaphor, like imagine we're at the same moment in history where we got from written ledgers to excel, right? Like Microsoft Excel, and nobody doubts now that Excel can calculate things correctly, right? I think the same is going to happen with ai. Like now we're in a spreadsheet.

And then we're gonna be doing things or calculations with AI and we're just gonna trust that it does the correct thing. I, I think we're right there as well.

Look, I don't think so. Like, and let me give another piece of information. In the previous startup that I worked at, like before founding my own company, we wanted to replace accountants. And for three years we tried. And the main challenge that we had was the trusted relationship between the accountant and the client.

Why? Because

Yeah, like people, people don't wanna self-service themselves in things that they think they will have huge liability first. And then second, humans or clients want a trusted advisor advising them. 'cause I was gonna say this, most of the times or many times your accountant knows more about you than your wife or your husband or your family.

'cause they know about your money movements, they know about your relationships, they know about you moving countries, companies. So at the end, I think. Even though AI could do it all and people could do it by themselves, they will still want somebody to take the liability or advise them in a better way.

'cause they, they got their own businesses, they've got, they got their own lives. I mean, their core business is not doing their own accounting for themselves. So especially when it comes to companies, I don't see AI taking over the profession at all.

Yeah. And, and taking a step back, let me tell you some things that I do see becoming a reality from a taxpayer's perspective. So first of all, they're going to expect smarter and self-serve experiences. Like, Hey, don't ask me via email, the same things that I've answered in the three previous years. Like, I expect you to.

Automatically check all of that and only ask me what you need. Like they, they will want, like, they will expect your firm to remember more about them, right? And have like a profile like, Hey, don't ask me the same things that I told you. Like the relationships, my structures, my everything. Like just, I want you to remember that the same way a chat gt remember versus things about me, right?

Like, I think that's, that's gonna happen. Like second. I think that they will ask better and smarter questions and probably even think that they know more than they actually do, right? Because they will be. Just the chat GPT away from trying to say, Hey, what's the best structure for me? If I'm in Costa Rica or I'm in the US or I'm Brazil, like chat, GP T's gonna say something

they know more.

yeah. And I also think just 'cause chat, GPD makes answers on demand so fast that they will start expecting that from their expert advisors as well. Like, Hey, I want an ans I want an answer and I want it now like in 30 minutes. Um, which is why I think if you're not ultra efficient in your own firm, you won't have the time to answer to that client, but somebody else will 'cause other CPA will be using AI like that for sure.

Yeah. I'm curious, in your view, what's the most overlooked trend in tax, like that firms need to prepare for?

I, I think, look what I've seen, and we're, we're focused on just small and mid-size practices, and what I've seen is that small firms can move like a rapid boat. While, while big firms like they need to, they, they, they, they have a long decision making process. A bunch of decision makers involved, like they just.

Need to move slower by nature. So I think there's a great opportunity. I see a lot of people trying to implement ai. They just don't know how or where to implement it, which is taking a step back, like I would say, just start with automating a simple task. Like, hey, every time I get a a, an email out of draft, out to reply with a draft or like prepare out, prepare draft for me to review.

And then every morning when you wake up, you just review drafts and then you edit it and send right. Until you feel in a place comfortable enough to like an AI taking over your email or, I don't know. I'm just thinking like, take one task.

I dunno if you've heard about this, the classical dilemma. Like will the challengers get the innovation first or will the, will they, sorry, will the incumbents, um, get the innovation first or will the challengers get the distribution first?

like you, the small firm needs to grow, like get the clients before the big ones, implement the ai like in, in, at scale.

Uh, and it's, it, it, that's the race and honestly. I envision a world where more people are, will come from like big firms and start their solo practices just 'cause AI will enable them to work faster than

So I, I am having a lot of thoughts right now. Like the first one is, I think banks are ahead of, in, in many ways. Like they have huge data sets. They can do a, a lot more things, but they're still very reliant in manual processes as well. And they're very like, siloed in, in between areas. But I do see some neobanks or like digital banks trying to go into accounting fields.

Like, I don't know if you've seen Mercury Bank or if you've seen, um, Brex in the us. They're starting to. Even do the automated bookkeeping from their own, like transaction ledgers, right? Like the first, the, the second they get a transaction, they boom, they provide an out category that the accountant can either review, change or um, delete, and they can push it into their QuickBooks or like the Xero, like whatever they use. So I think that the newer banks are thinking about this, like how can they embed themselves? Like how can they embed accounting processes within the, uh, the banks themselves, and especially in countries like the US where open banking is so good that you can connect multiple banks to a solution. Um, but I, I honestly, like, my perception is that they're doing this so that they can differentiate themselves from other banks.

Not, not because they wanna take over the accountants, right? Like I, I just see them as adding this. As adding a moat to their products. But I do see like the, the digital banks going into accounting, like accounting stuff, at least like account payable. It's receivables like, like some automations that they will still push into their, into your QuickBooks or your whatever your, um, yeah.

Bookkeeping software is.

no, I think they're very focused on their core business, like their core businesses lending money and receiving money, period. And then everything in between. I think there's a lot to do there. And, and I see use cases for AI there, like, uh, like, uh, identifying fraud or like analyzing transactions patterns.

Like I think there are more. Focused on this side of the business, like assessing risk and, and all of that, and not like it's a different business model,

I, I think it's different business models. I might be wrong. All of this is very new, but I, I will tell you what I'm seeing and what we're implementing. So at tax flow, especially in the US, that there's good open banking. We have been able to connect multiple banks to get the bank statements from different clients, extract the info, and give you a first pass of what the categorization is.

What we think is for you to review as an accountant. Um, so I think those kind of solutions that become kind of like an octopus that can like connect to different banks and give you the, like the, the, like the command center where you can see all of the financial data from your customer. I think those kind of solutions will start emerging.

But I don't see the banks themselves doing that. I see third parties like us doing that.

Yeah, of course. So we're focusing mainly on three things. Um, data intake. So preparing all the information requests in bulk from previous year's tax returns. So imagine Alex, you come to our platform tax flow, you just. Drag and drop the, the pre previous year's tax return from the customer. And boom, you got an email with like personalized for this person with all the documents that you need.

And this AI previously checked the client binder, so it's not gonna ask for what it already has. So that's, that's the first thing in bulk. The second one is, um, we're working on an agent to do tax prep. So once you have 50 documents, you just throw them all in and then. All of that gets reviewed, extracted, and put into your work papers.

This could be any Excel template, right? Like to, especially in international tax. Uh, you got work papers that are very complex, like complex calculations between each other. Um, like we're doing that. And the last one is an agent that can do data entry. So once you have the work paper completed, you can just take that same work paper, that Excel, that you, your firm uniquely uses.

Drag and drop it. And then these agents can open your tax software and go to the exact same fields where you need to input the info and put it for you. Um, of course it needs to previously be trained, so we're doing those three agents. If people wanna reach us out. Uh, we're, we're San Francisco based, uh, and you can reach us@jointaxflow.ai.

Um, and, and, and happy to help you, like, well start with tasks. That's my, my advice to, to everybody here in this podcast. If you're an accountant, the CPA. Start with tasks

correct, correct. And maybe we can even do a demo for the audience, like show them how this works.

you for the invite and see you round.