CPA Conversations Podcast icon


Apr 22, 2019

Artificial Intelligence: Accounting Profession Must Prepare for Impact

In a preview of the feature he coauthored with John Alarcon and Cory Ng in the upcoming digital-only special edition of the Pennsylvania CPA Journal, Accounting and Technology: PICPA’s Guide to an Evolving Profession, Troy Fine, manager, risk advisory services, for Schneider Downs & Co. Inc. in Pittsburgh, discusses artificial intelligence and machine learning. He shares how we use it in everyday life, the conveniences it can bring to the accounting workplace, and steps CPAs will have to take to remain relevant in an increasingly technological workplace.

If you’d like, you can download this episode’s audio file. Additionally, you can follow us on iTunes, Google Play, or subscribe to our RSS feed.

By: Bill Hayes, Pennsylvania CPA Journal Managing Editor

Podcast Transcript

Increasing improvements in technology are set to change the accounting profession in a major way, and nowhere is that more evident than in the area of artificial intelligence and machine learning. The topic is the lead story of our upcoming Pennsylvania CPA Journal digital-only special edition, Accounting and Technology: PICPA’s Guide to an Evolving Profession. Today we have Troy Fine, manager, risk advisory services for Schneider Downs & Co. Inc. in Pittsburgh and the coauthor of the feature with John Alarcon and Cory Ng. Troy, thanks for joining us today.

For novices in the audience, can you tell us, in a general sense, a little bit about the concepts of artificial intelligence, or AI, and machine learning? Outside of the realm of accounting here: what do the terms mean? Are they synonymous or is one an element of the other, like in the case of, say, blockchain and bitcoin? What do you think?

[Fine] That's a good question. It's kind of interesting: novices. I think most of us are still novices, even people who are claiming to be offering artificial intelligence products. I think it's in its infancy right now, that nobody truly understands the impact. Even I'm a novice in this. Most of us probably are. But at a basic level, it is more elements of one another. So, at a high level, AI ... if I look at artificial intelligence and I wanted to define it, it's really just teaching a program to perform a task like a human would and make decisions based off that. At its basic level, that's artificial intelligence. It's based on logic and rules. So basically “if this, then do this.” That's the basic AI. Now, AI is starting to get more advanced, starting to use algorithms, where you have probabilities and weights. This is where machine learning comes in.

Machine learning is kind of a subset of overall artificial intelligence. It's able to learn from experience and past data. So when new data comes in, these algorithms calculate this data. They add weights to it. They take the probabilities and they're able to predict an outcome. The way the algorithm learns is through trial-and-error. It's going to make mistakes in the beginning, but then it's able to learn and fine-tune that algorithm. Then, a subset of machine learning, actually, is deep learning, which actually is starting to take it to the next level. Deep learning is really when we're getting into how the human brain works, which is probably starting to scare people.

But, actually, deep learning is really trying to mimic how the human brain works. It's trying to create technology that can identify relationships and linkages in huge amounts of data, and then, based on that data, make a prediction, data that a human could not process on its own. It would be impossible for us to recognize those patterns and do those kinds of things. Deep learning, I think, is starting to come about. I think a lot of the stuff out there right now is machine learning. But deep learning is definitely that next level where it's starting to process things that humans just can't do on their own.

As you say, everyone's a novice on this right now. Just to demystify the concept here a little bit, are there ways that people currently use artificial intelligence and machine learning in their everyday life and maybe they don't even know it or think about it? What are some of those ways?

[Fine] We're probably using it more than we all think. A lot of us have virtual personal assistants. We use Siri and Alexa. That's actually artificial intelligence. Every time I talk or ask Siri or Alexa to do something, it's learning based on what I'm asking it. Right? If every morning you ask what the weather is, Siri's going to figure out that at 7:00 AM you're asking what the weather and traffic is, and it's probably just going to start telling you or sending you a notification without you even asking it. That would be the idea, that it's learning your habits, recognizing your patterns. Oh, you like to order pizza every Friday night? Do you want a pizza tonight? It's Friday, right? It's going to start to make those connections. Or take weather conditions. Normally, I do this when the weather is like this. It's going to start to be able to figure out that, hey, when the weather is like this today that John wants to go do this.

It's a little scary, but it's already starting to do that, obviously. Then, on social media apps, if you ever see where it says like, "Hey, do you know this person?" or "You might want to connect with this person." Basically, the social media apps have taken your profiles, they're looking at who you visit, what your interests are, what groups you are members of, and it's saying, okay, this person's similar to you. You might know them or want to connect with them because people you're connected to are connected to them already. It's things like that that I think people don't realize is actually artificial intelligence that they're using.

Search engine results are the same way. If I click on a link in a search engine, after I make a query and I stay on that link, the first link, for 20 minutes, the search engines recognizes that, “Okay, we provided a good list and they clicked on that first link.” Whereas if you go to the third page and you still haven't found what you're looking for, it then learns and says, "Okay, the results we provided for that query were not good. We have to fix this." So there's things going on that we're just really not aware of.

What ways can artificial intelligence, machine learning be used in accounting and what tasks can they streamline?

[Fine] I think this is the million-dollar question, right? I think a lot of CPA firms are trying to figure this out. I think, for our firm at least, we're kind of a middle-tier firm. We're the top 60. But we're starting off small. We don't have the capital to go and invest in our own products and build our own tools like the bigger firms do. They're investing huge amounts of capital in this, hoping that it pays off in the future. But for the smaller firms, I do think that you should go out and look for tools that are specializing in AI for the accounting industry. If you just do a quick Google search you will find companies that are starting to do this or claim that they're starting to do this. We're starting to use a tool, on our audits at least. Right now, we're just still in the proof of concept stage.

But we are trying to use it to take all this data from our audits and use this AI to say, "Okay, these transactions are risky transactions. You need to go look at this." So instead of us, in a typical audit, taking a sample of transactions, seeing which ones are risky and trying to find this manually, we can take AI, it'll take all the data from the past three years. It's going to learn, based on that client's general ledger and transactions, what is normal. And then, in year three, it's going to say, "Okay, based on what we know from this client, these transactions are risky and these are not usual for the circumstance." We're trying to see how this works. We're taking on a very small subset of clients right now. The good thing about this is the vendors are so new and they're new types of companies. They're willing to work with the smaller guys and try to make their products better. They're going to work with you. They're going to help you through this because they're trying to grow the same way you're trying to grow.

I think the middle, smaller to middle-sized firms, maybe the top 100 that aren't the bigger firms, should really start to invest in these products and take a risk, take a chance now. Because they're only going to get more expensive if there's more value being seen by them. They're going to start to see more value. It's a long answer to the question, but I do think smaller firms can at least start to investigate how they can use it on audits. Do I think it's going to be something they're going to be ... every audit's going to be using AI in the next two years? Probably not. We're not there yet. But I do think there is an opportunity there for this to create a better quality audit in the future.

On the flip side of benefits to the profession for CPAs, what would be some of the risks these technologies can bring? Anything that people have to look out for there?

[Fine] Somebody has to create these tools, right? An AI tool can't make itself, build itself, at least not right now. Maybe AI will build AI in the future. I don't know. But right now humans are still involved in building these products. So anytime a human is involved, there's always going to be bias, whether it's consciously or subconsciously. So if I'm building an algorithm and I'm trying to train an AI tool and the data I'm giving it is inherently biased because it doesn't include all types of situations or all variables or all factors that should be in that data, there is a possibility that the AI could be biased in this decision-making, just as a human, if I grow up in Pittsburgh, Pennsylvania, naturally, I'm going to be biased towards liking the Pittsburgh Steelers, right?

It's the same way. If the people who are building these products have biases that they are unaware of when they're building the tool, they might create results that are not good for the user. One example of this, and it's not even ... it's not an accounting example, but in recruiting. If you think about this, people are starting to use artificial intelligence to identify candidates and recruit people. Well, if I'm feeding it all this past data and the job I'm looking at is mostly American males who typically are hired by the firm, what do you think the AI's going to predict is going to be a better suited candidate? An American male, right? There's going to be issues there. We might get into the ethics questions, but that's just an example of how bias could be implemented into the process without the people building it or the people who are using it even realizing it.

I don't really have a good financial example or CPA example, but I thought that was a good example to show what I meant by bias. I think another risk from a CPA perspective is I think AI is not going to be a silver bullet. I don't think it's going to necessarily ... in the beginning at least, I don't think it's going to save time on an audit. It might, actually, if you think about it, if I'm using AI and now I have all these transactions that are considered risky that I wouldn't have caught before, that actually might create more time to do the audit. On the flip side, the idea is that you're performing a better quality audit. Because our job is to make sure we're identifying risk and mitigating it, and if I'm using the AI, it inherently should make my audit a better quality audit. If I'm doing a public company audit, the public should feel more protected, right, that these audits are being done better. But you might not save time on the audit. I don't think people should look at AI as, “Oh, I'm going to save all this time and I can charge clients less.” I don't think that's necessarily going to be the case. I think it's just going to be ... we have to look at it from are we getting more quality out of this product.

Yesterday, I was taking a look, and you see all sorts of these items on the internet these days, but I saw an item where someplace, I think it was in Japan, they had a robot who was putting together at a fast food-type place. The robot was putting together a food item for people. It kind of made me think about the question that I'm doing here as it relates to the accounting profession. Because you see these items and people are, of course, very fearful of the idea that robots are going to be able to take the jobs from people, such as one right there. I wonder how much of that builds into the accounting profession? Is that a fear that's rational to have that these technologies can come in and take people's jobs or will they enhance CPA positions?

[Fine] My opinion is that they will enhance it. I don't have any, obviously, numbers or data to back this up. If you think about the example I just mentioned how if you use AI and it's identifying risky transactions for you, that to me is going to make the person doing the audit ... it's going to make their job ... should make it easier. It shouldn't let them do a better job so instead of wasting our time on doing things that we thought were providing value we can now have a tool do this for us. Then we can focus on the value that the customer really wants us to provide.

I don't think it's necessarily going to just … we're all going to lose our jobs. There's just going to be more focus on different skill sets. Maybe your technical knowledge might not have to be there, but how am I going to tell this client ... I might have to develop, maybe, my emotional intelligence. How am I going to communicate to this client, hey, guess what? You have all this fraud on your books and before we were using this tool we never caught this. Right? That might be a difficult conversation to have with somebody because they're going to be like, "Why didn't you catch this before?"

The CPA profession's going to see things like that happen and they're going to have to kind of just develop different skill sets to be successful in the industry, having more advisory skills. How do I advise people rather than, “Hey, here's your audit report. We're done, right?” It's more, how do I provide value from advising? How do I help my client's business succeed? The accountant's not going away. We're always going to need somebody to audit a company. It's just, how do we use the tools to enhance our jobs and make it easier?

Are there ways that artificial intelligence and machine learning can be a threat to those standards? I know you talked about the bias stuff a little bit. But is there anything in the way of ethics that could be discussed?

[Fine] The bias, like you said, is definitely going to be an issue, I think, maybe not so much from a CPA doing an audit. If my algorithm has bias in it and I produce the wrong audit results, that's not good for your client. It's not providing value. Is it going to be an ethical thing? Probably not an ethical thing in that realm. But if you're looking at it from the recruiting example I gave, that's definitely ... there's going to be ethical things or problems that are going to arise from that. And privacy issues: I just read today actually that Deloitte released a report on ethics. They actually are suggesting that companies create ethics advisory committees and ethics policies to help fix these AI systems biases.

So before these big companies like Google go out and start building all these AI tools, they need to have a board of directors or members of a committee that analyze all these and say, "Hey, are we meeting privacy regulations? Are we doing anything that's unethical?" Because at the end of the day, if you're looking at GDPR in Europe with all the privacy regulations, I mean, anytime you're taking personal data now and running it through all these algorithms ... if I'm telling somebody they can't get a loan and I ran it through this AI tool because it predicted that based on all these 100,000 data points on this person that we shouldn't give them a loan because they're probably not going to repay it, do I have to tell the person why we didn't give it to them and what the algorithm said?

There could be some significant privacy and ethical issues in there, especially if people understand why they were denied. It might be something they didn't think people knew about them. Or it could be, hey, your grandparent committed a crime 20 years ago and was in jail and that's why. That went into this decision.

It could be crazy data points that are just so out there that people are probably going to have privacy issues with all of this. I think that's going to be the biggest concern, is how are you using my data? Maybe not for the CPA profession, from an audit perspective, but if the bigger firms are creating tools to help companies do things like this that are not audit-related, then they're definitely going to have to understand, hey, what are our ethical implications for building these types of tools?

I wonder if you could address the speed at which development of AI and machine learning is moving. How fast do CPAs have to get up to speed with these concepts? How fast do regulators have to come up with standards? In general, how long do we have until AI is really prevalent in the profession?

[Fine] That's also a million dollar question. I wish I knew the answer to that. But I think it goes back to kind of what I said earlier. I would start experimenting now on a small subset of clients, getting buy-off from clients, if you're going to start using this. It's not something that all of a sudden next year ... I don't believe that every single firm in the top 100 is going to be using artificial intelligence to do their audits. I don't see that happening. I just don't think there's going to be enough products out there that have been proven yet to do that. There could be some pushback from clients too, right? I mean, not that they want to hide stuff but if I'm analyzing all of your transactions instead of a sample, I'm going to find something that shouldn't have been done, right?

It's just inherent. So there could be some pushback from clients that are going to delay things. Regulation is probably going to delay things. If the bigger firms start doing this type of stuff, obviously, the government's going to take note of it and start to say, "Okay, hold on. Should we be doing this?" And that's going to hold up things. So I think it's a good idea to be aware of it, start understanding, hey, can we use this? But unless you're the bigger firms you're probably not going to see it widespread for a little bit of time. If you look at self-driving cars, I mean, right in Pittsburgh we have Uber. We're the heart of the self-autonomous vehicle, self-driving. We've had issues internally where they've had to take the cars off the streets for months. We see a lot of them driving around. Do I know the progress they're making? I don't know. I know there's a lot of regulation that's going to come into play. I would probably say we're all a little aggressive on how mainstream it's going to become.

Leave a comment

Podcast transcripts are provided as a summary of the conversation and have been lightly edited for the written medium. The transcript is not a verbatim representation of the interview.
Topic Suggestions
Have a suggestion for a topic? Want to be a guest blogger or speaker?

Let us know! Fill out this form.
Follow @PaCPAs on Twitter