How AI Could Have Saved Macy's $150M and Prevented Financial Disasters - Ahikam Kaufman

Welcome to the Future Finance show where we talk about. I for one, welcome our robot overlords.

Does anyone wants to go back to the days before the calculator where you used a mascot or whatever? No one, right? I think after people would have the experience of using us. You don't want to go back.

Today our guest is Ahikam Kaufman, co founder and CEO of SafeBooks AI, the world's first AI powered financial data governance platform. With over 20 years in corporate finance, Ecom has played key roles at major firms like Intuit and successfully guided startups to acquisitions.

SafeBooks AI can maybe you start by just telling our audience why you founded the company, what you do.

I thought there was like this gap that always existed that people used to solve it by doing a lot of manual work. And I thought that today using AI, we can actually automate that and allow people to focus on the really important things. We can identify missing data, not just the existing data, but missing data that should have been there and it's not. Essentially what we did is we developed a unique graph technology which brings all the financial data and create a one big scheme, one big graph out of your financial data.

It's not like you're processing any journal entries, doing any write back. You're strictly just kind of doing that, monitoring and alerting.

I'd like to think that what we do, where we monitor your data, we don't inject any data, we only have a read only access to the system. The other thing you want to monitor financial data because financial data is money and if you have leakage, you leave money on the table. One staggering story, which I think is very interesting one, Macy's are about to do their earnings release that's like the week of Thanksgiving. And what they announced was staggering, they identified today.

Future Finance is brought to you by qflow AI, the strategic finance platform, solving the toughest part of Planning and analysis. B2B revenue align, sales, marketing and finance seamlessly speed up decision making and lock in accountability with qflow AI.

Welcome to Future Finance. Today we have Aikam Kaufman, co founder and CEO of SafeBooks AI, the world's first AI powered financial data governance platform. With over 20 years in corporate finance, ICAM has played key roles at major firms like Intuit and successfully guided startups to acquisitions. Ayikam, welcome to the show.

Hey, thank you so much for having me.

You're at south by Southwest this week. How's that going?

It's an amazing festival. You know, sometimes you go to tech conferences. Here you see a variety of Ages and a variety of people and Personas. So I think the fact that they, they combine the music and the film festival makes it a very, very interesting and colorful event. So the streets are full and it's like, it feels like one big festival. So it's very exciting.

That's great. Are you speaking on any panels there or what's your agenda?

Like, we were on stage for a pitch. They selected a few companies to kind of pitch the audience and that's what we did. So I came over here with our coo, Steve Selvik, and yeah, he actually did the pitch.

Cool.

Great. Well, you'll have to be sure and get by. What is it? Perla's, the famous seafood place in Austin there, right across the, right across the river. That'll be a good stop for you while you're there. If they're. They're probably so packed right now, I don't know if you'd even be able to get in.

Yeah, everything is very packed. Yeah. Actually traffic at midnight looks like midday. Yeah.

So it's like downtown LA or New York.

No, totally. Yeah, yeah, yeah.

I still remember the first time going through downtown LA at 1:30. Might have been a Friday or Saturday. I'm like. And it's still rush hour. What's wrong with this picture?

There's actually traffic at midnight. Yeah.

Well, enjoy the rest of the conference there. Hopefully you get to enjoy a little bit of music and have some fun. Not just the presentations, but want to jump in and ask you a question here. SafeBooks AI. You founded the company a little bit over two years ago. Can maybe you start by just telling our audience why you founded the company, what you do, just a little bit about it.

You know, I, I started my career as a finance guy. I had the privilege to work in four large companies. The first one was a company called Mercury Interactive, where it was a very super successful software company. Eventually we got acquired by hewitt Packard for $5 billion. But in the process we got into a compliance issue. And I've seen the dark side of corporate life. Of course, I personally was not involved, but the company had several issues and walked its way out of these issues with the sec. And then I became an entrepreneur and I did a bunch of things. We actually did a B2C company which was fun in the payments world, but I always dealt with compliance. Even when I was in fintech and payments, it's all about compliance. At some point I thought, and I continue to use it until today. So there was like an inherent gap between accountants and what they do on a daily basis and their ability to actually touch the foundation layer, the foundation of finance, which is the data. So if you think about that, companies today, they process a lot of data in, across disparate systems, in many systems, whether it's like you know, billing, elp, CRM, spend, management, payroll, H I S. So when you, when you try to, when you try to manage finance, you're dependent on a lot of system which essentially process pieces of the same transactions even. And it's your requirement as a finance organization, we call it internal controls, to be able to monitor and govern the data. But there's a gap because you're only an accountant, right? So you can do, you can deal with transactions but to deal with the data itself. So I thought there was like this gap that always existed and people used to solve it by pulling reports, doing a lot of manual work, putting into spreadsheets, and basically you need to document how you check the data. And I thought that today using AI and a lot of recent technology, we can actually automate that work and allow people to focus on, on the really important things which are like, you know, how do you remediate the data if you have to, how do you make all kind of decisions, whether they're accounting, operational decisions and not just monitor the data. So monitoring the data, which is a concept who's been around for 20 years in DevOps, in cyber, in IT, didn't get to finance. So what we are doing is literally replacing manual walk, manual specific repetitive work around what we call is financial data governance.

And what you guys are doing, I mean it's kind of the holy grail. I've been talking to the office of the CFO about AI and finance for several years now. And I mean with the real time data monitoring, automated reconciliations, your fraud detection mechanisms, and then integrating directly with ERPs, CRMs and all that, I mean this is, you're addressing problems I've been trying to solve as a consultant and as a, as a former CFO for years. And I'm, you know, I think that right now kind of in the run up and hype, the hype cycle of around generative AI, a lot of people have kind of lost sight of the fact that AI has been around for a while and I guess we'll call that classical AI machine learning. And a lot of what you're doing is classical AI. That's a lot, you know, it's deterministic, it's a lot more predictable and we, we understand it better than generative AI. That's generating new content. And I'm looking, you know, at the tools that you provide and obviously those are sort of machine learning tested approaches that have been around for a while. But I'm also like, when I see how quickly you can get a customer up and running, I'm imagining you have some generative AI components as well in there. And I was wondering if you could talk about, you know, maybe a breakdown of how you're using the classical AI machine learning in the platform and, and maybe how you're applying generative AI if you are, or if you could kind of run us through the technology that you guys are using today.

Actually, that's a great segue to what we do. So thank you for the question. It's actually you describe it very accurately. So a way to think about that. We developed several capabilities to be able to solve that problem. And I'll explain how we solve the problem. But essentially what we did is we developed a unique graph technology which brings all the financial data and create a one big scheme or schema and one big graph out of your financial data. And the reason we want to do it because we want to take that transaction from the CRM and connect the same transaction in billing at the same transaction in the lp, the same transaction in the payment gateway, and maybe the same transaction, the supporting documents that you need to monitor. Because people today are literally flipping pages and comparing order forms to the data in the CRM for a lot of interesting reasons, but they do that. So we created that basic technology with machine learning and then we run controls over that to be able to identify what looks okay, what looks wrong. And in the process of doing that, we're not just vetting or deciding ourselves, we're actually taking the checks, protocols, all the things that people are doing manually and we just automate them. But that's like the, the more like the ML part which you talked about. What we are doing is with AI, we are doing several things. So first of all, we, we don't just read figures, we read information or data. So think about entries, journal entries or description or expense reports where you have, you know, lingual data, you know, you have like descriptions or whatever. And we identify that the description doesn't match something. It doesn't match the date, doesn't match the amount, talks about something else. Maybe an entry is misclassified. So you see like an R and D entry which is in marketing or whatever. So our ability to understand the description of transactions such as journal entries or expense reports or whatever, and we could talk about why we're checking expense reports. But that's like, that's like one thing. The other thing is the ability to ask data questions. So we developed this module where basically we figured if you're connected to all the data, and I'm like the cfo, right? So the CFO doesn't get down to the engine room that often, right? So it's not like he's not dealing with what we're doing. But if we're connected to all the data, we can answer business questions on your data. So the ability to create a human interface where you can, in free text, you can ask a question and we will retrieve the answer is interesting. By the way, whether you are CFO or an auditor to that extent, imagine that you have like a door to the entire, the ability tool. It's like a vista point to the entire data lake of the office of the cfo. Another thing which you're doing, which is very interesting. So turns out we learned in the process of doing that, and that was a very interesting experience, that people spend manual time or doing manual work comparing documents to data, which sounds very, very natural and obvious for finance, but they're still doing that. So as a, as a mean of control. So for example, at some point when they close the books, they would compare all the, as much as many other forms as they can review or MSAs or whatever to how the transaction or the revenue was booked in the books, in the CRM or in the books. We automate that. So we use language models to be able to read documents and marry them to controls on the data. So I'd like to think that it's essential today to use AI. I'll give you another example. When we scan your data or your books, we have mini models who understand the structure of your transactions. And then we can identify anomalies, we can identify missing transactions. Let's say you get this lease charge every month. All of a sudden you're missing a charge. Obviously you need to accrue for that. I know it's all kind of boring stuff, but that's kind of the front and center of finance work, right? We can identify missing data, not just the existing data, but missing data that should have been there and it's not. So all these things we're doing, like with real AI and what we are doing with the traditional AI, back to your question, is more like the ability to run statistical queries across the data, find the anomalies that create a graph and all of that.

Okay, Paul, before you're going to need to step in here because I'm fighting the temptation to really geek out here and get into the Neo4J database or whatever with the nodes and edges and. Yeah. So I'm going to, I'm going to step back and let Paul ask a normal person question.

I want to give another example. I want to provoke you guys. And to me, to me it's a humbling experience. It's really a humbling experience, but it's like an amazing one. Now the reason we want to do that is because the amount of data is enormous, right. And human beings, we can't expect them to be able to monitor every single transaction. Now why do you want to monitor every single transaction? Because for many reasons, one of them, you want to trust your data, right. So you want everything to be in the right place. If you have just for example, if you have a transaction which is classified to marketing instead of R and D, you may violate the tax code because as an R and D expense maybe you need to amortize something that as marketing expense you just expensed. Right? That's like one tiny example. The other thing you want to monitor financial data because financial data is money. And if you have leakage, we actually, we can talk about leakage examples that we've identified. You leave money on the table. The staggering one staggering story, which I think is very interesting one, it's three months, actually three and a half months ago. Macy's are about to do their earnings release that's like the week of Thanksgiving. And what they announced was staggering. They announced that in their books, in the numbers they released already or they were about to release, they identified that they've under accrued for freighter calls. Right. The reason they have a freighter call is because I think their online business, they ship from the nearest store. So imagine that all your store are basically warehouses and you have to be able to manage the inventory from that to when you ship online. So you have to manage these costs. And what was amazing, so, so they, they announced that they've under recruit for like $150 million. And as a result they had to delay their earning release and they open strategies, special investigation and all of that. I'm not sure they even closed the matter until today. But. And I'm sure it was a but and I actually heard that the CFO is an amazing guy. But that's not what's I think mind blowing. What's mind blowing to me is what they announced to the SEC that it was a, a result of a single point of failure. So think about organization like Mises. You have like 500 people in finance, maybe a thousand people in finance. And as a result of a single person who actually did wrongdoing, he didn't, he didn't take the money. Right. But he, he was trying to hide the problems that was growing and eventually it blew up. But because of that, now you think about you as the cfo. How can you know about that? So in today's world, in cyber you have all these controls using automation where you can't ignore issues. People see those issues on dashboard and then they either take care of them or they escalate them. But in finance it doesn't happen because we are all playing with spreadsheets. So the ability to monitor the data can actually save your career, can actually save the company a lot of money from having to deal with compliance issues. And again, it's like you can see that a single point of failure caused this entire amazing company to delay its earning release on the week of Thanksgiving before the basis parade and all of that. And to me that's like an eye opening experience and another reason why we need to use more automation in internal controls, which is what we're doing.

Yeah, no, it's amazing the mistakes that we are made, sometimes the dollar amounts, the challenges. So you know, obviously you're helping companies, you know, manage, see all their data, understand that flow. But what particularly you're doing to help make sure they meet kind of the regulatory and compliance demands. Are you loading all those rules in? Because, right. Different companies have different things they may have to meet for compliance or regulatory. So how do you manage that in the tool with you know, doing different industries and things to make sure companies are compliant? Ever feel like your go to market teams and finance speak different languages? This misalignment is a breeding ground for failure, impairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet qflow AI, the strategic finance platform. Purpose built to solve the toughest part of planning and analysis. B2B Revenue QFLO quickly integrates key data from your go to market stack and accounting platform, then handles all the data prep and normalization under the hood. It automatically assembles your go to market stats, makes segmented scenario planning a breeze and closes the planning loop. Create airtight alignment, improve decision latency and ensure accountability across the teams.

Right. So I have this great example that I'm using If that's okay. So many years ago, I had the chance to, at a large company I work for, we did a contract with Palantir and I had the chance to work with the platform. So I'd like to think what we're doing, we have this platform which is geared towards being able to automate and monitor your financial transactions and automate the monitoring end to end. So what we do, basically, we take your existing wallpaper, we take your existing controls and all the checks you're doing today with some of them you do daily, some of them you do weekly, some of them you do monthly, some of them, it takes a lot of works and maybe you do a few times a year. And we just automate that in the system. So a way to think about it, we have the ability to ask the data any question and we can run it on a daily basis. And as our CRO Steve says, all the time. AI never sleeps. So it works 24 by 7. And this way we monitor every transaction. So we actually create it for the customer. And another principle that we have is we don't like deployments and our customers don't like deployments, so we deploy it for them. But we take their existing protocols, their existing tasks, and we just create these use cases. And because of our technology, we can create a use case in like a matter of hours. So actually, we're finding ourselves, and I think the other interesting experience that when the customer understand and realize what we can do, they add more and more because there's always stuff you never get to, and now you get to that. So the way we help them is by automating these checks and balances that they will do manually and automate the workpapers. So you end up with having all your data monitored way before close happens, right? So if you had an issue and the way the ERP captured an invoice was different than the way it was captured in the erp, you're not going to find out about it like a half a day before we need to close the books on a Friday afternoon, you're going to find out about it in real time. That's one way we are doing it. The other way is that all the documentation you need to produce, which is like the wallpapers, right? Essentially half of what happens when we say finance, close the books, they produce wallpapers, they produce all kinds of analytics and reviews, and all of that, we automate that. So think about machine creates for you, like these spreadsheets that you're used to. Literally, we create These spreadsheets and they flag, just like in the application, they flag all the transactions you need to pay attention to and you just get the same spreadsheet you used to produce manually. It's now, you know, it's now automated. So that's how we. So at the end of the day, it's all about manual work. It's all about replacing manual work.

We like that. As much as I love manual work, I like it being replaced, Glenn.

So I come, I. This is why you're so much smarter than I am. I'm. I've been just trying to explain all of this to the CFO office as if, let us go implement this under, you know, under your office. Whereas you've gone and created a product that does this because it's, I mean, it's not. You're basically, you're creating your own data pipeline, building your own data lake, setting all this up, whereas I'm trying to help them do that in their office. But to a cfo, this is all. We might as well be speaking Greek right now when we start talking about every, you know, what's going on under the hood here. So the fact that you've built this platform and you're able to implement it so quickly, I mean, you're, you know, as much as I do. And, and I guess what I'm trying to sell is the end product of what their internal sort of data pipeline and data lake and how they're accessing the data and what, what they're reporting on. Whereas you can come in with the product and just say, look, give, give me access to your erp, your CRM, epm, whatever tools you're getting them in. You're going to build your own data lake, be able to report on it. And, and so you are clearly far ahead. Your capabilities are clearly far beyond what CFOs even understand right now. And I'm wondering, as you talk to CFOs, I imagine the sale is not that hard because if they can quickly see the results and the implementation, is that easy. But as you talk to CFOs, what's your sense of, outside of using your product, how AI and automation are going to change the office of the CFO moving forward so that they do become more technical and, and do start using this.

I want to say a few things about that, but please feel free to weigh in when you think I'm like hitting a point or I'm missing a point, so don't be shy about that. So we would make it interesting. So I'LL say a few things. First of all, I'll start from the end. We're definitely seeing more interest in finance team to adopt AI and innovation, period. I think it's like after deep Seq. So if you like, if you think about after deep Seq and after like January and February where you're busy closing the months and they're like in the trenches once they came out of. You see any. I would even say that we've heard from a few companies that their finance teams are compensated as part of their 2025 goals to deploy AI and innovation. So it almost feels like overnight define the office of the CFO doesn't want to stay behind and they compensate people for that. So I think that's like a unique behavior we are seeing. Overall, I would say CFOs are almost divided into two. If I may, if I may, I want to say it very amply. I think you have a lot of CFO which are very strategic. They're coming from investment banking. They're coming from like I even seen a couple of weeks ago, CFO of a public company was a partner in a law firm for 20 years. So. But they're very strategic. For whatever reason, these kind of people, they're less aware of internal controls and the importance of internal controls. And then you have the CFOs who came from accounting, corporate controllership and all of that and they get it right away. The analogy I'm using with people, I'm telling them, listen, let's say in your house you have, let's say between five to 10 credit cards for whatever mileage plan and other rewards you have, you have all these credit cards. It's obvious in your house all these credit cards are operated by the right people for the right reasons. But if you don't read the statement to make sure you don't have a false charge or like a fee or like a double charge or like a fraud, then that's the difference between transacting and internal controls. And most people, most of us don't read those credit card statements because if you do then so that. But a company has to do that and compliance is not optional. So I'd like to think that those who come from the accounting world, they're more. And those who come from the investment banking world, it's a more of our learning experience or a lot of time they cascade it down to their teams. But this is why I talk about menial work. Because to me the way to penetrate it's not because of the innovation. It's just because of the sheer work that we can replace with automation. Right. But definitely now we're sensing more and more in finance team, they need to be a little bit more innovative and deploy AI tools. Now you asked another question. Where do you want to deploy these AI tools? I'd like to think that because of hallucinations and other things, it's very. The office of the CFO is a very sensitive, dangerous place. I'd like to think that what we do where we monitor your data, we don't inject any data, we only have a read only access to the system. We help you actually identify the issues as they are being as they happen. It's a much more convenient way to deploy AI as opposed to. Because we are not making any decision. We automate your work, we show you where to look for stuff, but then it's your data. You can always investigate that and resolve that. But we don't inject any data. You're not going to wake up and find a bunch of wrong data in your system. And I think that's like a very nice sweet spot to be in right now for us.

So it's not like you're processing any journal entries, doing any write back, you're strictly just kind of doing that, monitoring.

And alerting, if I may use that analogy. And it's high windows, but you have a lot of AI tools in the, in the medical field, for example, that help you scan MRI or cities, if I may. Those systems defined today as decision support systems, by the way, what's their main advantage even. It's an interesting case, but what people realized that even before helping you to understand the specific image you look at, it helps prioritize. So think about a night where a hospital processed 500 people in various whatever devices and. And now someone, a radiologist has to review 500 which one he reviews first. So the machine can help you just rearrange the line or the queue just to make sure you review those who are more urgent faster. And I think we're just like that, right? We help you focus on we what we say we scanned everything. We are so certified so you can rely on our controls. But basically you may want to look at these first and just make sure these are okay. And those are like the suspicious things that. Yeah, if it makes sure.

Look at these 10 first because we've done the. Everything else looks normal from running it through AI. These other ones look like they have an issue. So let's start here.

AI triage.

Yeah, yeah, yeah, no, makes sense. I Mean, that's a great analogy, right? Because if you look at things, you can see if an accrual is three times what it was last month, that should flag something. Or if there's an extra zero, like it was 10,000, all of a sudden it's a hundred thousand.

I'm saying even more. I'm even more than that. I think when you are public company again, we can talk about 404A and 404B. I'm sure you're aware of it or you heard the term. But you have a legal obligation to deploy internal controls over your financial data. Right. When you. Now you know. So if before you had like three people doing spreadsheets and someone may have been sick. Right. And option, as we say, compliance is not optional. Right. Even if people sick, you still need to do that. Right. But if the AI runs 24 by 7 and you deployed all those controls and you configure the system, you can actually rely on that and you can say, okay, I trusted the system to do this. It flagged that. Even if we missed something, you can still. You did your best effort and you made in a reasonable effort as required by Sarbanes Oxley or by the audit standards to monitor your data.

Yeah, makes sense.

But as we said again, it's a new concept. If you will actually Google financial data governance, I'd like to think we are the first unsponsored result. Now, Gartner started to talk about it recently, just recently, before the holidays. And they talk about DNA data and analytics governance. And Gartner actually determined that the number one, number one priority for CFOs in 2025. You can Google it up. It's what they call financial data governance or a financial DNA. They call it DNA Data and analytics governance.

Yeah, no, it's definitely super important and I love what you're doing because it's needed. Nobody wants to do it manually and we're much more likely to make mistakes versus let's let the machines what they're good at. Help help us and make our lives easier.

Does anyone wants to go back to the days before the calculator where you use the mascot or whatever? No one. Right. People say, hey, but maybe there are less. No one wants to do it anymore. So that's the same. I think after people would have the experience of using us or maybe in the future other platforms, you don't want to go back.

Agreed. All right, so we're going to move into. We have this fun section we like to do where we get to Know you a little bit. So how it works is we fed your information into AI, in this case, use Claude, and it came up with 25 fun and unique questions. So here's how I ask them. You get two options.

You can either have, by the way, for you it's the fun part. For me, it's the scary part, I guess.

Yes, exactly. So you can use the random number generator and I can generate a number between 1 and 25, or you can pick a number and we'll ask that question.

Okay. Thirteen.

Thirteen. All right, let's see what we have. Give me one second here.

Oh, that's a good one.

That's it.

I like this one. This, this isn't scary. Don't worry. If Safe Books AI had a company mascot, what would it be and why?

Making life easier. Would that work?

That, yeah, we'll give you that one that will work. What do you think, Glenn?

Yep, yep. So I tell you, we, and Paul and I, it's, we've been doing these questions. We wanted to mix them up every week and not just ask the same question. So we really, we used to do bake offs between the different, you know, Claude versus Chat, GPT versus Gemini and all that. Now we just kind of mix it up. But I think Claude 3.7 has gotten better. Gemini is still the worst by far. Google, you know, they got to catch up. But so Paul does the random number generator or gives you an option. And what I do on mine is I just, since we had them created by a, an LLM, I just have it. Pick one. And so we're going to let no human in the loop here. Okay, let's see. So the question that I've got is if you could solve one persistent problem in financial technology with a magic wand, what would it be? And everybody's talking about how AI washing companies and they pretend AI is a magic wand. So I think you maybe already are solving it. But is there one particular problem that, that stands out for you, that if you could snap your fingers and fix it, you would.

I think it's a data reconciliation, but you know, you could, it could also be like, you know, the, all the accounting resolutions, which is something we plan to do in the near future. But you know, once you capture all the data, how do you make sure that you're following the rules? But that's like, you need to know a lot of rules for that. And yeah, the reconciliation is like the low hanging fruit. Perfect.

Perfect. So, all right, well, this, you know, Paul, what this made me think is I Need, we need another podcast that's like for the super geeks only, like the engineering podcast. And then I feel like we could have, we could have a whole other hour conversation right there, just talking about the deep dive.

See, I would always glaze over, Glenn, because you'd lose me.

That's, that's why you need your.

I'll tell you, it's. It's very technical, it's very mechanical. But as we saw in the Macy's case, again, it can be scary. And I'd like to think that remote work doesn't help compliance because new members of teams can be less educated. They didn't rub shop, so there's a lot of. We didn't talk about the accountant shortage. If you will just Google accountant shortage in Bloomberg, you will see staggering numbers. So I'd like to think that I even have an assertion, and it's not politics, that with Dodge, for example, even if they would cut the workforce, they would double the penalties because they would say if you get caught, you're going to pay a lot of money. So I'd like to think that, you know, there are a lot, a lot of trends now that require automation. You know, the accountant shortage and remote work are some of the things, maybe even not being able. A lot of companies outsource that work to India and other places they would like to. I think it's better to use automations. So, yeah, we can deep dive if you want.

We'll have to do that next time. What do you think, Glenn? Deep dive next time.

Perfect. Perfect. We'll have it for the. We'll. We'll put a warning on it for the nerds only. We'll just.

Yeah, yeah, yeah. Exercise discretion before.

That's right.

Yeah, we, we could have, you know, future finance for the regular finance people. Future.

But we can talk about it. It will be still be interesting. Don't worry. Hey, guys, thank you so much. I can't appreciate that enough. It's been fun.

Yeah. Thank you very much.

Thank you for joining us. We've really enjoyed it.

Thank you so much.

Thanks for listening to the future finance show and thanks to our sponsor, qflow AI if you enjoyed this episode, please leave a rating and review on your podcast platform of choice and may your robot overlords be with you.