Skip to main content

The Impact of Artificial Intelligence on Tax Departments

CrossBorder Solutions’ CEO Donald Scherer and ‘Fiona Show’ host Matthew DeMello discuss why the changing global regulatory environment for transfer pricing necessitates the use of artificial intelligence to keep up. But unlike other industries impacted by technology, this technology also accentuates the role that humans will always play in the process.

Matthew DeMello: Before we get started, let’s talk a little bit about you first. How did you get into tax and transfer pricing?

Donald Scherer: I heard Mimi’s podcast the other day, she chose to come and work across border and get a transfer pricing. Mine was an accidental thing. Meanwhile, I was an international tax guy at KPMG and one day a pro a partner came up and said, “Can you write up a report?” It was the early days of transfer pricing and I found I liked it. It was more allowed you to be a little bit creative. It allowed you to write reports rather than work with numbers, which was never my strong suit.

So after going to working with KPMG, I went over to work with international tax services group with Ernst & Young in New York and London where I basically start to specialize in transfer pricing. CrossBorder Solutions One [the first iteration of the company] came from there. But that’s, that’s really my start and transfer pricing partner had a project, gave it to me and I kinda liked it.

Matthew DeMello: And where in that process did you know that you were going to do CrossBorder?

Donald Scherer: When I was working at E&Y, I had the idea to do a risk analysis program where we would look at 5471’s and 5472’s and we would then analyze the client’s risk of being audited in a given country. I did some really light programming work and I automated it. We were sending out these basically risk analysis reports to all these clients, in the early days of transfer pricing. A lot of them never even heard of transfer pricing. And that’s where I sort of got the bug of using technology with tax.

Matthew DeMello: And that evolved. You didn’t just launch into a cross border one with only that idea?

Donald Scherer: Well, no. After spending almost three or four years doing transfer pricing at E&Y — and I was, by that point, doing the studies, the functional interviews, comp searches – I started to see patterns that over and over again we were doing, doing the same thing. It seemed to me that if you’re doing it like that, then software could do it because you’re doing it over and over and over again. And that I thought it would be perfect for software. And, although I know how to code a little bit, I’m not a software per-se engineer. I looked around and I really only knew one person that knew how to code at that point, and that was my mother. The problem was that she had a full-time job and she was happily working on Wall Street and making more money than she’s ever made before in her life.

And so I came to her, I said, “We should both quit our jobs and go start this transfer pricing software company. And at that point you’ve got to realize that everyone was starting companies [at that time.] Every magazine you read, all day: they sat down and worked in their garage, came up with some sort of software, put it in a Manila envelope, sent it out, and became multimillionaires. So I said, “We should do this!” Then I looked at what I was making and what she was making. I was making a lot less. So, [I said] “All we have to do is sell like 20 of these a year ever, and we’re good. Then I don’t have to work for the Big Four anymore (back then, it was the Big Five) and you don’t have to work on Wall Street.”

And that’s really how CrossBorder Solutions got started. It got started by me somehow convincing my mother  to go down this journey.

Matthew DeMello: And the technology that went into CrossBorder One is a very, very different than the technology that went into CrossBorder Two.

Donald Scherer: We had that CrossBorder One, and this is, I mean this is silly if you look back now. My mother, although she was a software engineer, she was working on Wall Street running trading databases. So she wasn’t per se a Full Stack coder. She knew Microsoft Access, which is not by any stretch, a real development program. It is meant to be used by non-software professionals to be able to quickly prototype stuff. But that’s what she knew, so we said, “all right, we’re going to build it in Microsoft Access,” which is ridiculous.

No one builds commercial software in Microsoft Access. We thought we were going to do it in six months. If you remember back, I said, “We’ll do it for six months.” We didn’t have a garage, so we were working in her living room. We said “We’re doing six months, we’ll start selling it for $5,000. We’ll sell X number, we’ll make more money than we are. And that will be that. The problem is it took three years, cause again, 1. We didn’t know what we were doing and 2. It had never been automated. So we were making up a lot of stuff as we went along. There were no rules on what our transfer pricing report at that point should look like. Everyone was doing something different and so we kind of just said, “Alright, we’re going to make it up our way.”

And it took us three years, but we got it out. Our first client was Pepsi, our next client was Intel and went off and we were off running. That code line supported the entire product, through the life of CrossBorder One. And when we sold the company to Thomson Reuters, there were almost 1,400, maybe even 1,500 customers using that Microsoft Access product. I’ll give a shout out to my mother: It was almost bug free, which was incredible because literally you never see that there were no bugs and it that worked great.

Matthew DeMello: She made it all by yourself or did you have a team by that point?

Donald Scherer: She coded almost every line. My sister was brought on after a year because we weren’t getting it done. She did the reports. I did a lot of the verbiage.

So now, when you look at a transfer pricing report, the first 20 or 30 pages are country-specific verbiage. When we first started, that wasn’t there. I thought, “Hey, let’s make these things thicker!” Also, I needed something to do while they’re all working off hard on the code, of course. So I wrote those introductory second sections, and if you look back now, everyone uses every report that comes out worldwide has that. But that’s just because one night in a living room we said, “Yeah, I guess we should have thick introductions to the sections? So let’s try it!”

We were winging it as we went along. And because we became the standard and because so many companies started to use us, a lot of what we came up with is you’ll see today in the reports that are produced by CrossBorder Two, or even any of the Big Four firms.

Matthew DeMello: You touched on the sale to Thomson Reuters in 2007, but I want to even set the table a little bit up just a little bit before getting there. You were saying how that was one of the impacts, that CrossBorder One had on transfer pricing. Tell us a little bit more about that impact you feel it had when, when you sold it to Thompson Reuters.

Donald Scherer: CrossBorder One showed people that transfer pricing could be done in software. I mean, when we first launched this, a lot of people didn’t even know what transfer price was. I can’t tell you how much time we spent convincing the market that yes, you have to do this. It wasn’t like it is today where everyone knows that transfer pricing has to be done in every country. It just wasn’t there.

The other impact that the technology had was it also showed that people could do it on their own, that you didn’t need to work with a Big Four consulting firm and – back then, a Big Five. That this is something that can be viewed as a normal compliance activity that you could do in-house. I think that’ll sort of two big changes that we were able to accomplish, and we did that pretty well.

And then we sold it to Thomson Reuters in 2007 and they maintain that software. I mean, they upgraded it and I guess they’re still using it today.

Matthew DeMello: One Source, the Thompson Reuters transfer pricing documenter tool.

Donald Scherer: That came from our humble beginnings.

Matthew DeMello: So everyone [from CrossBorder One] takes a vacation. Everybody sees the world, or what do you do?

Donald Scherer: After we sold the company, I met my two kids. I actually knew my older kid, because my older son, Jordan, at that point, when he was one, two, and three [years old], we still lived in the city. So my wife would bring him to the office every day. Our, our loft was right outside our office. And so I spent actually a lot of time with him in the office and stuff. My second son, Sam – we had him after I moved to the suburbs, I left early to beat the traffic and I got home late. He got up later than I and went to bed early. So I barely saw him.

So the first year-and-a-half, two years after selling the company, I got reacquainted with Sam. I took him to all the mommy-and-me classes, like Gymboree. I did all that, almost for two years sam and I hung out. And then I did two other startups and I was actually doing my third startup when Thomson Reuters called us up and said, “Would you like to do something again with your old transfer pricing software?” And that was the genesis of CrossBorder Two.

Matthew DeMello: Did you think even a little bit before that with BEPS, with things changing that there was going to be a market opening or that you would want to jump back in?

Donald Scherer: Yes. I’ve done a lot of startups now and one of the things that we have seen is that there’s good startups and bad startups. Good startups are ones that go like a rocket ship and those are really fun. Those days that you look back on say, “Wow, that was like the best time of our lives!” I realized some startups that don’t do so well are not as much fun and you don’t miss them. We had [CrossBorder One] alumni dinners — when we sold the company to Thompson, we were almost 300 employees at that point, [in] maybe 2007. We had not-annual… probably bi-annual reunions and every time we would all get together we’d all say, “You know what, the market is changing.” There’s BEPS, we should really get back into transfer pricing and also —

Matthew DeMello: … get the band back together.

Donald Scherer: Yeah, get the band back together! And we’d all be like, “All right, now we have been there, done that” in feel. But finally after Thompson got in touch with us, we decided that maybe we should take a little bit harder look at it.

There was a lot changing. There were two things that really excited us about going back into transfer pricing. The first thing that made us think about really going back into transfer pricing was the localization aspect: so many countries were coming up with their own regulations. It used to be that every country did the same thing and you would change one little section of the report and then everything else would be the same because everyone basically followed the OECD guidelines. It was a unified standard. But now, all of a sudden – back in 2016, 2017 – all these countries, all of a sudden started doing their own stuff. When you have a hundred countries doing their own stuff that calls out for technology, right? Because all of a sudden it gets much, much harder to comply.

So that was a solution set that I thought we can actually help solve. It sounds nerdy that we’re a tech company and we were excited. Yeah, exciting people over here [sarcasm]! But at CrossBorder, one of the things that really sort of got us going was the changes that were going on in the underlying technology: what technology could accomplish. With AWS, and the cloud, and artificial intelligence, there seemed to be something that you can really do, and make it special, that wasn’t there before when we built a [Microsoft] Access program. These are things that “dreams were made of” so to speak. Right? I mean if you dream about transfer pricing… but if you dreamed about what someone could actually do and now, you’re looking at the technology and you can actually do what’s only been dreamt of previously – that  sort of got us thinking that maybe it’s time to get back into the business.

Matthew DeMello: You said you could tell the problem already necessitated technology. Did it necessitate artificial intelligence or…?

Donald Scherer: No, no. A little bit, maybe. We looked at the artificial intelligence is helping on the comp search, which is related to the localization because now every country has to do their own local comp search. No, the localization was that when you have 78 countries having their own regulations, it’s a complex solution set and you really need some smarts and some technology to handle that officially and effectively.

Matthew DeMello: Now [2019], we’ve gotten to the point of really getting to artificial intelligence. I think that’s just very much something that’s very different than the people who know about it than it is to the people who just use Amazon Alexa Dots. So, let’s nail down a definition: What does artificial intelligence mean to you as compared to me just saying “Hello, Alexa”?

Donald Scherer: I think that’s the $1 million question, right? If you go to 10 people on the street and you say, “What’s AI?” No one’s going to be able to say actually what it is. What they’ll say is, I’ve heard it, everyone talks about it all the time.” You cannot watch TV for an hour without someone talking in a commercial about AI. And I think if you look at it, it’s a very broad-based term with a lot of different meanings and I think it means a lot of different things to different people. And that suffers from what you see with a lot of new technology. People don’t quite understand what the capabilities are. They think that it could do more that it could really do, right?

And if you listen to the commercials, AI is literally going to take over the world. And we’ll talk about that [later]. But there’s always hype around this, so everyone’s always playing it up to be more than it is.

Matthew DeMello: There’s a global climate change doomsday factor to it.

Donald Scherer: I don’t mean to downplay its power, but then I don’t think a lot of people know what it is. And then they put marketing spins around it…

So I will tell you that dictionary definition, let’s start there. That’s the easiest way. And that’s the capability of software to imitate intelligent human behavior.

Matthew DeMello: And that would lead people to believe — especially when they interact with Alexa, I’m sure more than a few Echo dot Users have figured out you really can’t have a conversation with Alexa – that those are pre-programmed answers.

Donald Scherer: I will say there’s different levels and we could talk about that, but yeah, just on the short end, then I’ll get back to what AI is. But on the short end is: Yes, you can program Alexa to say certain things, but Alexa actually has the ability to learn to hear things and then find new information [based on those answers]. So she actually goes a little bit deeper.

So, let me then back that up for a second – so if we say that’s the capability of software to imitate intelligent human behavior: what you’re now seeing with that definition is incredible computing power being put onto this solution set. What computers can do now compared to what they could do three years ago, it’s just, there’s no limit to, to what the power that people have at their disposal. You use Amazon cloud, which is what CrossBorder Solutions uses, it’s unlimited. I mean, there’s no processing limitations, which allows you to do some really amazing stuff.

Now let me give you a little bit more of a nuanced definition. There are two types of AI, though we look at it here [at CrossBorder] and I think a lot of people do: there’s rule-based AI. Rule-based AI is basically [made up of] decision trees. “If x then y” [style-conditions] that have rules that say what you can do, and determine what path do you go down on your decision tree. The more rules that you put into the solution set, the more powerful the AI, right? Basically, a human in many ways makes decisions based on rules that they know. So rule-based AI is where you’re basically having a decision tree, you’re programming rules based on some sort of stimuli — whether it’s voice, whether it’s something typed in — whether it’s something that impacts the question of what the computer then or the software should do. That is what we’d call rule-based AI.

Now there’s then something called machine learning.

Matthew DeMello: Very different.

Donald Scherer: Everyone hears about machine learning, right? And then, what is that? That machine learning is a subset of rule-based AI where the software actually learns, on its own, how to handle that decision tree. It’s learning based on what, what inputs they’ve seen before. The classic example of a machine learning algorithm is Netflix.

Somehow you’d go onto Netflix and they tell you what you want to see. Now that’s based on a rule-based AI mechanism, right? That’s the screen that says, “Here’s what you should look at.” But then Netflix learned what you would have liked in the past, and so what’s in your queue years later is a product machine learning.

I think the real power of AI in general comes from this enormous computing power coupled with machine learning and rule-based AI. When you put all those things together, you will have a very, very powerful solution set that allows software to function much closer on how a human reacts to different situations. Does that make any sense?

Matthew DeMello: That makes sense to me. I think that’s a way better working definition to separate fact from fiction for the folks at home. So now let me ask you, why is AI with that definition so perfectly suited for tax?

Donald Scherer: I think most people don’t always say, “Oh wow, it’s tax. Now let’s use AI.” Right? I don’t think that’s the first leap that most people make. But if you think about it, it’s perfectly suited for them taxes, its own language. It’s its own very formal rule based environment. It’s very technical in nature. It’s very numbers oriented. It tends to rely on big data and it also tends to have very distinct problem sets.

All of these things allow you to create very sophisticated decision trees. And more importantly, it allows you then to have machine learning because it’s doing the same thing over and over again. So they can actually learn from the earlier inputs. In some ways tax is perfectly suited for it because you’re really doing your own language. So you can basically program the software to understand a specific language rather than a general humanistic type approach.

Matthew DeMello: And in general, what areas will tax departments do you feel will find Artificial intelligence helpful?

Donald Scherer: Today, it’s not local. And so you have very, very diverse set of rules worldwide. I think that’s one way that you’re going to see AI be very applicable to tax. But where you’re going to really see it, that’s the great solution set. That’s the set where they automate routine repetitive tasks that involve large datasets on solve. You know, where we’re seeing AI being used a lot is with classification issues: How do you classify expenses for correct treatment? How do you categorize sales for correct treatment? So classification problems using large datasets is certainly something that you’re going to see AI used for.

You’re seeing it with in the research area where, you know, traditionally you do a lot of research in the tax world, right? And it’s historically been done using books, portfolios, and then it went online. I think you’re going to start seeing a lot of research being done using natural language processing, you know, saying, “Alright, what are the tax rates in country X?” And then getting the answer.

I think there, there’s a lot of AI associated with that because the software has to understand what you’re saying if you’re not staying it in a very prescribed manner. Then, of course, there’s a global environment is a global knowledge base out there on tax. I think AI is perfectly suited to bring some of those issues to the forefront. So you can say, “Alright, what, what do I have to do in country X?” And you should be able to get that in.

Matthew DeMello: You mentioned a few times that you know “AI is perfectly suited for tax departments,” And I know that’s going to hit a few people’s ears –even if they understood our definition of AI — they’re going to hear that and say, “Is this coming for my job? Is this like driverless cars.” I know we found at CrossBorder, with our PSG group, is that AI works differently for tax departments. Or how do you see it?

Donald Scherer: I watch something I think on HBO two weeks ago and it talked about AI and how it’s going to take over certain industries. And it was scary! I mean, the problem that they were talking about were truck drivers. Did you know truck drivers are one of the largest industries in the U.S.? Turns out, hundreds and hundreds of thousands of people do this important work. And they were showing on this HBO special how basically trucks are going to drive themselves. Now, if it’s in two years — I don’t know if it’s in five years — but they’re going to drive themselves.

I just got a Tesla, and everyone can say what they want to say about Teslas. I drove from my house in Sarasota to Tampa [in a Tesla]. It was an hour and 15 minute drive. The car drove by itself every now and then I had touched the wheel to let them know I wasn’t sleeping. It changed lanes. It got off at exits. It stopped on the road when there was a red light – if I was the second car, if you’re the first car, it goes right through and that you don’t want. It parked itself! It literally parallel parked itself! I got out of the car, it locked itself! This is available today.

So what I could tell you, and this was the point of the HBO special, that in five years, no one’s driving a truck anymore. And so that is scary. That is fundamentally scary. They actually drove to a truck stop and they show these guys what’s coming and you could see their face.

Matthew DeMello: This is very different though because as you were mentioning with, with your experience at KPMG, there are automated pattern behaviors in tax, but that’s not the whole story.

Donald Scherer: So this is really different. I mean this is going to replace what people do. Amazon! They talk all the time about how Amazon is automating what they’re doing. I mean, there’s going to be displacement.

But I don’t think that actually applies in the tax world. They’re not going to replace it when tax, what I think you’re going to find is it’s going to more empower tax professionals rather than replace their jobs. It will allow tax professionals stop doing manual, repetitive tasks that and allow them to focus more on the higher level stuff.

There’s a lot out there that doesn’t suit itself well for what tax professionals do on a daily basis. I’ll give you an example and one of the reasons is because there are so many interaction intersections of different areas of tax.

So if you’re doing something with merger and acquisitions, you’d just say that’s a [data] set. But if transfer pricing is involved in it, if classification issues are involved… There are so many different interconnected sets to it that it’s going to make it very, very hard for an AI to assume that type of responsibility. Where you’ll see AI be used [more] are on these large repetitive tasks of big data, which should then free up the tax professional to focus on the stuff that’s more value-add.

Matthew DeMello: So, right there. Let’s say somebody is hearing this as a tax professional, where should they think about the future in terms of this maybe replacing their skillset.

Donald Scherer: The skill set is on value-added activities that take real thought, that take thought judgment. That’s just not going to be replaced with AI. I don’t think this is going to happen very cleanly. But I think you’re onto something. I think there’s going to be a lot of impediments to tax departments adopting AI, even for repetitive tasks.

As you just mentioned, there’s fear, uncertainty, and doubt. I think everyone’s scared of it and listen – [from what] you’d watch on that HBO special, I’d be scared too, right? I mean, I literally said to my kids, “You’ve got to watch this, because they were talking about doctors that are automated.”

And what’s a human being going to do anymore in medicine? Twenty years from now you’re going to go to the doctor, you’re going to give your XYZ information and [the doctor, leveraging AI] can say, well, “This is what’s wrong with you.” The whole diagnostics of a doctor’s job is probably going to go away. Surgery is being done by robots right now, right?

Matthew DeMello: Because surgery doesn’t require judgment, so long as you know which organ to remove or what operation to conduct.

Donald Scherer: But it is almost like a decision tree, right? It goes back to that decision tree [version of] AI.

Listen, there are industries that are going to change radically. But I don’t think tax is going to be one of them. But the repetitive tasks are going to go.

Matthew DeMello: And it’s those repetitive tasks [that will have to go], as you were saying before.

Donald Scherer: And one of the reasons that I don’t also think that I think adopting AI is going to be slow going into text apartment is that it’s hard to do. These algorithms are not easy to write. More importantly, they tend not to be right the first time. They learn over time.

So at first it sounded like, “All right, we’re going to spend — and it’s expensive — we’re going to spend this much money and then we’re going to get to perfect answer.” Now, that’s not what’s going to happen. What you’re going to wind up doing is spending x dollars to come up with a classification product or program, and then that’s going to classify them all right. And then you’re going to have to redo it again and iterate through. Whereas, the machine learning kicks in and starts to learn over time you get better results. But I think especially in corporations, the rule is “If I spend X, I have to have Y and I have to have it correct.”

I also think that tax often looks for a black and white answer. You know, they need that specific answer and I don’t think AI is great for giving specific answers. It’s very, very good for giving a range of results that then you can use your judgment to narrow it down to the right answer. But it’s never going to give you a black and white answer.

Matthew DeMello: And just to underscore that point — that is transfer pricing [not having a black or white answer]. And more so than other forms of tax compliance.

Donald Scherer: I think transfer pricing is in many ways perfectly suited for artificial intelligence. Let me tell you a little bit about how we’re using it here today at CrossBorder.

We developed a platform that we call Fiona. I think she talks during these podcasts.? I’m sure she does, I’m sorry. She’ll say something at some point. But the first thing that we do is we use natural language processing to feed the platform to provide access to our knowledge base. So this is, “Hey Fiona. What are the one of the documentation rules in Germany?”

Fiona: Of course, Germany follows the OACD guidelines, you know, a master file, a local file, and a country by country report. However, as you might expect, Germany has a few pesky additions, the German tax or authorities. Once your report do include an executive summary of the intercompany contracts, essential intangibles owned and used by the taxpayer and the names of company employees who have decision making power over business relationships. And if that weren’t enough, they also want a description of the value chain and the taxpayer’s contributions to it. Now, I am certainly glad I’m not the one who has to read through all of that.

Matthew DeMello: Well neither would I Fiona. Anyway, down back to what you were saying.

Donald Scherer: We have put our whole knowledge base on the Alexa platform from Amazon, as you can tell.

Matthew DeMello: And she answered us.

Donald Scherer: She did and she can do that.

I think the other thing that you start seeing AI use for at least we’re using it for is that we have very, very sophisticated decision trees that we talked about earlier that are rule based, that help our customers comply with all the different local regulations. Let’s say it’s collecting documentation. So if you have a legal entity operating in Mexico, it knows to ask the user what is your value chain versus if you have something in Canada, they don’t care what your value chain is. The software is smart enough to know that and you can see that in documentation. You can see that in comps or ranges. For instance, Australia uses one type of range and Canada uses a one-year [interquartile] range and the U.S. uses a three year range…

And the software is smart enough to know – based on the decision trees that we have set up, based on where you are – how you localize the transfer pricing analysis that you’re doing. I think the most exciting thing that we’ve done with AI at CrossBorder Solutions Two is that we have applied machine learning and rule-based AI to the comp search portion of a transfer pricing analysis.

Traditionally, Big Four firms spend 40 hours doing a single comp search, a localized comp search. And that’s the biggest expense of a transfer pricing project and it’s the biggest time sock of a transfer price in product. And the way they do it is they have an analyst that looks at each description of the comparables. Maybe they go to those comparables’ websites to make sure that the description is right that’s in the database. Then they look at that financial data and then they think about whether they should do it or not, and maybe they bring it in and maybe they don’t. It’s kind of a subjective call.

What we have is our AI actually reads to descriptions. And this is actually a great example, in the time it takes me to say this, the AI can look at 1.8 million comparables, goes to their websites, analyzes their financial data, and then comes out with the six best comparables. And it could literally do that instantaneously

Matthew DeMello: From the point you started saying this sentence. It’s really that fast.

Donald Scherer: And it’s done. And that’s compared to a human who could really only do a comp search by hand. And when you do a comp search by hand, you can’t do more than four or 500 comparables because otherwise it’s just taking too long now. Literally, you could look at the entire universe and come up with the best number for you.

I think we’re really excited about that. And what we use them machine learning to do is as we see these comps over and over again. The machine learns that, “Oh, when you asked for a contract manufacturer in Poland, these are the six that keep coming up and let’s now we know which ones are better” and it learns as it goes. That — that is some really, really powerful stuff on the transfer pricing side that we’re real excited about.

I think the software that we’ve developed now, or the Fiona platform, has the ability to really, really change an industry. So what was once a very labor intensive process normally done by outside consultants like the Big Four – I think now, all of a sudden, companies can have very localized reports that meet the regulations of every individual country, but they can basically get that a click of a button.

So it’s sort of like Turbo Tax now for transfer pricing and that’s because of the AI and the processing power that allows us to analyze these huge datasets in basically no time.

Matthew DeMello: You mentioned before that companies often go into AI thinking, “Oh, it’s going to solve my problems. Click of a button. Yay, done.” You explain how that’s not really the case. I assume that’s got to be from some personal experience implementing at CrossBorder. Tell us a little bit the growing pains of implementing and bringing AI to CrossBorder, how it’s changed the company

Donald Scherer: Well when we first started this company, we were learning this too in some ways. I mean we spent a whole lot of money. I said before, one of the big impediments to entering this space is it’s a lot of money. We spent almost $8 million building this platform – the AI, specifically – and it took us a lot of time. And then after we did it, we then used it internally for almost a year where we were using the same concepts, these rule-based concepts, to handle the compliance for our customers. But now, the results were great and we were kind of mind-blown about how great the results were. And then we finally now then put an interface on it where we can roll it out to our clients, let them use it on their own.

I think, today, we’re really excited about where it’s going to go. If you look out a year from now, I think you’re going to be able to see some stuff that is really like real Star Trek stuff. You’ll be able to say to your Echo Dot or to Fiona via your computer, “Fiona, I want to run a profit transfer pricing analysis for between these two countries, for this transaction.” Go and you’ll see a study show up in your email. I mean, that’s where it’s going to go.

And if you take that one extension further, it’s all, all this information stored in a database. Every study that we produce, I think we did 10,000 studies last year using Fiona.

One of the things that you’re going to be able to do if you’re a client or someone using the platform, you’ll be able to say, “Fiona, what were the results of the transfer pricing study of widgets between country X, the country Y,” and Fiona will go, “The arm’s-length range was between three and five percent” and they’ll be able to give you the results.

Or you’ll be able to say to Fiona, “Fiona, what’s the ownership structure of this company?” “Who owns, who’s the majority owner?” “Fiona, tell me about where I’m at most risk worldwide.” And she’ll be able to analyze everything and say, “You’re at risk for audit in these five countries.”

All of this data is going to be able to be processed and learned and it will be literally at your fingertips. I mean, think about, you know, those Star Trek fans. I know…

Matthew DeMello: Hey, we’re here.

Donald Scherer: … we’ll say “Computer,” and it comes out with the answer. That’s where this is going to go.

Matthew DeMello: That’s what we actually call an Alexa in my house.

Donald Scherer: Yeah. But to emphasize: This is literally where this is going to go. Where the data on someone’s situation is at their fingertips and can be told back to them. More importantly, you’re going to be able to tell the computer what you want, and all of a sudden it’s going to get done.

I think that goes back to empowering tax professionals because: Yes, we took off from their plate the repetitive task of looking for comparables or figuring out what questions you have to answer in a certain jurisdiction. But what we are giving the tax professionals is the ability to have all this information right at their fingertips so then they can use it in value-added ways and do the things that humans do best, which is make decisions that are relevant to their situation.

We’re going to extend this platform to R&D tax credits and to some others. Over the next year you’ll see probably three different product lines come out. I can’t talk about them now but it’s… it’s going to fundamentally change the way taxpayers deal with large datasets and sort of streamline their operations and let them focus on what they should be focusing on.