What Does AI Mean and How Does It Relate to Accounting

Welcome to The CFO Project Podcast, where host Adam Lean dives deep into the challenges and opportunities facing today’s accountants, CFOs, and firm owners. Each episode is designed to help you break free from the “accountant’s trap” by exploring strategies, tools, and insights that fuel sustainable growth. If you want to scale your firm, elevate your client relationships, and build a thriving advisory practice, this podcast is your go-to resource.

In this episode, Adam sits down with Chris Farrell, CEO of Liscio, to discuss how artificial intelligence is revolutionizing accounting and CFO advisory services. From eliminating tedious clerical work to enabling smarter client communication, Chris explains how AI will reshape firm efficiency and client value. They explore the difference between public and private AI models, the privacy risks firms need to watch out for, and how tools like semantic search and automated workflows can free accountants to focus on strategy and insight—not paperwork. If you’re curious about where the future of accounting is headed and how to position your firm for success, this conversation is a must-watch.

Transcript

00:00:00:00 - 00:00:08:06
Unknown
Welcome to the CFO project podcast. Today we're talking all about what does I even mean and how does it relate to accountants

00:00:16:04 - 00:01:00:06
Unknown
Hey, Adam. Great to see you again.

00:01:00:08 - 00:01:23:22
Unknown
Sure, absolutely. So my name is Chris Farrell, and I'm the CEO of SEO. We do secure client communication and, file transfer and so forth for accountants. So we really kind of smooth out that whole path between firm and client. And we do that in a way that allows you to basically see everything in one view. So you're not scrambling, looking across all our different emails and various platforms for client info.

00:01:24:00 - 00:01:54:00
Unknown
So I got my start with Arthur Andersen way back when in the 90s, and got my CPA. There was in private practice or public industry for the decade after that. And then for the last 20 years or so, I've been working on building tech for accountants.

00:01:59:23 - 00:02:16:14
Unknown
Yeah, it's this thing that's all around us. It's affecting every industry. Right. And if you look at big players such as Apple, if you look at big players such as Google, new players such as OpenAI and so forth, there's a huge buzz around it. And I think we've had things such as Siri on our phones for a long time that are helpful, but they really haven't had a major moment yet.

00:02:16:18 - 00:02:39:15
Unknown
And there are other AI platforms, such as ChatGPT that have come out and have completely demolished entire, software companies. You know, if you ever were around the educational software, you know, market, you could look at companies such as Chegg, which is what all the kids were using in college. In high school, for a while, they literally their stock went down 99% in the last three years.

00:02:39:17 - 00:03:16:15
Unknown
And so Chegg would make, study guides and basically help, college students, for example, bone up, you know, learn languages or learn math, etc. online. They had great apps, they had great websites, etc. for educating people. And no kidding, Chegg. Their stock is down 98 or 99% in the last 3 or 4 years. Because OpenAI came out and just basically said, if you want an answer to something here, it is not always perfect answer, but you typically really close at an amazing price point and incredibly easy to use.

00:03:16:16 - 00:03:38:00
Unknown
So you can imagine a world in which, you know, if you want to get an answer like for a great cacio e Pepe, you know, recipe, you just go to ChatGPT, say, I want the recipe and it gives it to you. And there's no ads, there's no links, it's just direct information straight to you. So imagine that same thing for a student you know, who needs an answer to a math question or how to prove something out.

00:03:38:00 - 00:03:40:11
Unknown
They would just go there and get the answer directly.

00:03:40:11 - 00:03:59:23
Unknown
And I think AI is going to have a very similar impact across every profession, including accounting, which is I want to do X, Y, or Z, and you can just give it short form language and it's going to give you a great answer back that is going to be essentially, very much understandable, shareable and so forth in an instant.

00:04:00:01 - 00:04:15:10
Unknown
And so I think there's this whole idea that when you start to apply AI in a practical sense across what accountants are doing all the time, there's literally dozens of applications that are going to take away so much of the administrative burden that's been plaguing the profession

00:04:15:10 - 00:05:45:22
Unknown
for decades that are going to appear overnight. And I think we're really at that cusp right now where it's going to get to the point where of who we're going to see mass adoption, extraordinary privacy, and where UI perspective an absolute revolution, because how we interact with software is going to be fundamentally, very, very easy.

00:05:46:00 - 00:06:06:07
Unknown
Yeah, yeah, I think it's, it's a great question to ask because these things are almost like mystery boxes that came out of nowhere, and they're full of potentially goodness, and they're full of potentially badness or hallucinations. So the way you should look at it is they're basically probability engines. So they are going to look across the corpus of everything that they know.

00:06:06:07 - 00:06:23:00
Unknown
So there's a whole body of knowledge that they're filled with. So Microsoft is obviously sitting on a goldmine of of data. And so what it's basically saying is, okay, high protein foods, it's going to look across that entire corpus just like Google does, but instead of serving you, the link is going to serve you what it thinks is the common denominator.

00:06:23:06 - 00:06:40:23
Unknown
So if there's a million answers out there that say high protein food includes like eggs, fish, etc., it's going to say, okay, there's a very high probability because there's a million articles saying that these are high protein foods, that this is what the common denominator consensus is around that. So it's looking for the consensus on the data set it's trained on.

00:06:41:01 - 00:07:01:23
Unknown
So if you go take it and you train it on a garbage data set like Reddit or something, it's going to come back with a bunch of really wild results. But if there's a really good data set with, you know, a gazillion results out there and there's a very strong consensus, it's going to be incredibly accurate. So recipes, high tech protein foods, incredible accuracy around those.

00:07:02:01 - 00:07:47:03
Unknown
Yeah, yeah.

00:07:47:05 - 00:08:05:01
Unknown
Sure. Same thing. Just like I say with Google. Google went out and crawled all the sites, understand all the information. Microsoft has the same thing. Just nobody uses it really. They have Bing, right. And so they've done that. And then OpenAI has a partnership. They OpenAI makes ChatGPT. They have a partnership with Microsoft. They get the Bing data as well.

00:08:05:03 - 00:08:18:21
Unknown
And they are pulling data from other sources, including, you know, including things that we upload. So for uploading things, having conversations with it, it's using it as a training data set. And, you know, learning through acquisition of knowledge directly.

00:08:18:21 - 00:08:29:19
Unknown
Everybody listening. If you haven't already, sign up for our five minute weekly email with practical tips for accountants and bookkeepers to escape the accountants trap. Go to the CFO project.com/newsletter.

00:08:37:03 - 00:09:05:03
Unknown
I think it's a great thing in the hands of the right, you know, users. So a lot of companies are looking at it saying, well, you know, can we use this to essentially automate the accounting function? And, you know, if it's if you're thinking about pattern recognition, if somebody goes out and eats at Applebee's and we know that we've seen a ton of data out there across the entire body of knowledge that Applebee's is related meals and entertainment expense, you can start to really figure that out and start to do things, you know, kind of automatically.

00:09:05:05 - 00:09:19:15
Unknown
And so is it more and more of that training data comes in. Let's say there's a new vendor around the corner. It's a restaurant in New York City. It's going to see that and say, okay, I know this is meals without having to be trained with an account and saying, I'm going to match this vendor to this expense category.

00:09:19:17 - 00:09:29:17
Unknown
He has me great at things like that. Right. It's just going to say high protein food, fish, great matched. So I think you're going to see a lot of things that are repetitive kind of go away.

00:09:29:17 - 00:09:37:10
Unknown
One of the things I'm excited about at Lucio right now is everybody's renaming files. So a file comes in from a client and somebody took a picture of it using their iPhone.

00:09:37:12 - 00:09:52:20
Unknown
It's going to say like IMG underscore or some weird, you know, alphanumeric kind of string. Instead of that, we're going to look we look at the document now underneath it and say, okay, we can read what's on the document. And we see Applebee's and we see an amount. So I say, okay, great. This is, an Applebee's receipt.

00:09:53:02 - 00:10:11:12
Unknown
And we can essentially programmatically create a new file name for it and name the file automatically instead of a human doing it. And so there's a lot of drudgery around this kind of thing. And we think about that same use case. If you want to find that file, hey, find all meals and entertainment receipts from 2024. We know it's Applebee's receipt from 2024.

00:10:11:18 - 00:10:27:01
Unknown
We can just go out there and grab it for you. Just natural language search. So imagine how much work that's going to take off everybody's plates by being able to not be frustrated, trying to find things, be able to use semantic search, which is kind of what you want to do with Siri or anything like that, to like, Hey Siri, go out and find me X, Y, and Z.

00:10:27:01 - 00:10:53:12
Unknown
Find pictures about something like this in my photo stack. It'll do that. And so the amount of categorization work, the amount of filing work and so forth is going to go through an absolute revolution. And there's not going to be that much of that work to be done that just clerical in nature. I think that's going to be a real breakthrough and allow, you know, a lot more space for firms to interact with clients and actually do accounting work.

00:10:53:14 - 00:12:02:14
Unknown
That's just one example.

00:12:02:16 - 00:12:21:05
Unknown
That's right. That's right. I think the same thing, I think when you're thinking about, you know, CFO work and I was a CFO of a company, you know, many moons ago, it was the same thing if you wanted to go into a meeting with the CEO, and sit down and say, here's what's happening with the business, you're going to run your variance analysis.

00:12:21:05 - 00:12:38:06
Unknown
You're going to go ahead and do your double click through the transaction types and understand what was really driving, you know, expense changes, trends and so forth. I me great at going out there and taking a large data set. The key and the value is not in the CFO going in there and doing the clicks. The value is in what question do I ask?

00:12:38:11 - 00:12:54:19
Unknown
And then based on the data that gets presented, what conclusions do we reach in order to make recommendations to essentially improve the operations of the business? We're going to see, I surface a lot of the information for us because we're going to know how to prompt it. We know what to ask it in order to get those things.

00:12:54:21 - 00:13:16:05
Unknown
That capability largely exists today. The problem is you have to be really, really careful about how you use it. Because we talked earlier about how, you know, these systems learn they will take this data and use it as training data. So if you upload something to one of these large LMS, such as a ChatGPT, that data can end up in the public domain.

00:13:16:07 - 00:13:29:08
Unknown
So oh yeah, it's a large. Oh yeah. Pardon me. It's a large language model. Yeah. That's so they're just taking tons and tons of text and.

00:13:29:10 - 00:13:52:18
Unknown
They are limiting. That's right. And so there's gonna be two kinds of lm two kinds of this AI foundational tech that you're gonna have a choice between. So some will be what's existing. We call those foundational models or the big engines out there such as vs copilot, cloud by anthropic ChatGPT by OpenAI and so forth. Those are public models.

00:13:52:20 - 00:14:13:17
Unknown
Then there are private models, private models or models in which your data never leaves the system. So Lycia, for example, has a private model. So when you use us, we have the files, we have people's secure messages. We have all this information, we process it all within our boundary. And our business is based on privacy. So we never let this stuff go out.

00:14:13:17 - 00:14:36:12
Unknown
It never goes back to train a model. So if you upload a file and have us process something or for, you know, reading the context of your secure messages and making recommendations for replies, all that kind of stuff is essentially going to stay private. So when you think about technology to use to start to surface, this kind of stuff, gotta be really, really careful about which ones you choose.

00:14:36:14 - 00:14:48:08
Unknown
And consider really the privacy ramifications because there's a lot of risk out there. So anyway, I just want to make sure everybody's clear on that private LM versus, you know, public Lem or foundational.

00:14:56:03 - 00:15:13:08
Unknown
That's right, that's right. So instead of like, imagine this world in which, you know, you go out with a foundational Lem, it looks the entire body of knowledge on the internet. The internet's full of stuff, good and bad, right? We are very focused, and we're only interested in accounting. So when we train our Lem, we train it just for accountants.

00:15:13:08 - 00:15:38:08
Unknown
So one of the features that we have now in product, it's really exciting is, when you have a conversation with a client, we want to make it really easy for the client and for the, for the firm just to say, hey, respond to Adam, let him know that, I'm on top of the items we discussed. And fill them in on these two items and then let him know I want to meet next Friday instead of next Thursday.

00:15:38:10 - 00:15:47:06
Unknown
You just use that kind of shorthand. We'll look across everything we've talked about in that thread and create a personalized message in your voice

00:15:47:06 - 00:16:00:15
Unknown
that addresses the items we talked about earlier in the thread. It saves a ton of time, right? Because how much time, as in CFO advisory, how much time you spend going back through, remembering all the pieces and then composing a message.

00:16:00:17 - 00:16:20:12
Unknown
It takes a significant amount of time. People will spend ten 15 minutes on an email, but if it knows your voice and knows all the context, you can say, hey, I need to do this. And then it's going to give you largely the answer. It's going to turn you instead of from a full writer. It's going to turn you instead into an editor far faster, far better results.

00:16:20:14 - 00:16:44:08
Unknown
And again, extraordinary privacy. So one of the biggest things that when you look at the LLM world is with a private Lem design, the right way, you get to use the context of the conversation. You get to use the context of how you interact the client, the data you have through your interactions with them, and use that to inform how I will help you craft messages.

00:16:44:10 - 00:17:05:16
Unknown
So again, another breakthrough that when you start to look at where the time goes, right? How many people sat there with a gigantic number of emails in the inbox? I got to spend the evening nights and weekends getting a sifting through it. A lot of that work is actually going to become much simpler, and the AI is actually even going to sift through it and tell you what you need to do next.

00:17:05:18 - 00:17:21:12
Unknown
Yeah, it's available today. It's brand new.

00:17:21:14 - 00:17:36:05
Unknown
Yeah, we're we're very close out. And I think it's certainly within 12 months that you're able to say, take anything that's junk and go ahead and file it for me. And then what do I need to know. It's going to surface things like, hey, you have an urgent message from Marie who needs you to get to this thing by noon.

00:17:36:11 - 00:17:52:10
Unknown
So instead of you having to scan, it's just going to pull that thing out because it'll be it's observational. It learns your patterns and will be able to anticipate what you want to do. And we are certainly within a year. 100%, 100%.

00:17:58:11 - 00:18:15:05
Unknown
There's so much there's so much out there. As we talked about automatic file renaming, we talked about automatic message composition using context. There's going to be a lot more things. We start to surface a little bit about this whole idea that you can be have a conversation with it instead of having to move through UI. We took a lot of pride.

00:18:15:05 - 00:18:31:16
Unknown
This is hard to say. We took a lot of pride in our design. It's for a people to be able to find files, but that's going to go away. People are just going to say, hey lis, yo, go get my tax return from last year. That'll be the client experience. Literally. You open up the app and it's just going to be almost like Google.

00:18:31:16 - 00:18:53:00
Unknown
Tell it what you want and it goes and gets the thing right. And that's going to be everywhere. It's going to be pervasive and so, so much what we've done. Imagine just in accounting, how many times have how many hours of accountants spent creating file folders with subfolders and then having to find a file and find and navigate to the right folder to find it and put in the right place.

00:18:53:00 - 00:18:57:18
Unknown
And it doesn't always go in the right place. You have to go finding all that stuff's going to go away. It's all be abstracted.

00:18:57:18 - 00:19:08:11
Unknown
I'm not saying there won't be files and folders, but I'm saying is that's going to suggest where to put them. And these are just really trivial, mundane things that in the world of accounting add up to serious hours.

00:19:08:12 - 00:19:15:17
Unknown
Right? Yeah. Yeah.

00:19:15:19 - 00:19:31:14
Unknown
I know about it. I haven't watched it.

00:19:39:11 - 00:20:05:13
Unknown
Yeah. It.

00:20:05:15 - 00:20:29:10
Unknown
That's right, that's right. It's going to when I first started working at Arthur Andersen I'd come out of school and school. We studied accounting. The big misnomer about being a junior accountant is you're not going to do much accounting. So I stood by the copier by 6,070% of my day was, you know, non accounting work and you only do about 20% accounting work.

00:20:29:12 - 00:20:47:23
Unknown
And it was it was wild. I couldn't believe it. And none of us could believe we're all first years and say we're actually not doing anything here. We're doing clerical. But they had to have a ton of clerical in the accounting firms to support the audits, to support the tax and so forth. If you look at tax like the not so secret, it's an open secret, right?

00:20:48:01 - 00:21:11:02
Unknown
Tax software prepares a tax return. Full stop. Right? We can use TurboTax ourselves. So we're not paying for tax return. We're not paying for tax prep as consumers. We're we're paying for is all the stuff that goes into it. We really want the planning. We really want the, the assurance that we're going to get this thing right, the things the software might miss or we might miss, etc. that's what we're really paying for.

00:21:11:02 - 00:21:30:12
Unknown
Same thing with CFO advisory. Paying for the advice, not paying for all the time spent double clicking, looking at variances and trying to get into the detail. And so what's going to happen is it's going to get to the fruits of what we're actually laboring for and allow us to get much better data, much cleaner data, and actually help firms grow better with far less work.

00:21:30:17 - 00:22:23:16
Unknown
So we are literally at that point right now with AI and I, you know, we've done a lot of surveys on it. People know of AI, but they don't know the practical applications yet. And I think once that starts to really kind of find its way in the mainstream, which I certainly think is within 12 months, it's going to be an absolute revelation.

00:22:23:18 - 00:22:44:03
Unknown
Yeah, absolutely. It's like OpenAI has grown and gotten adoption equivalent to Google, and Google grew like crazy, right? It took 20 something years for Google. I think is 22 years for Google to get to the same scale that OpenAI got two and three years. Yeah. It's just it's completely and totally it's it is pervasive and it is unstoppable.

00:22:44:05 - 00:23:04:16
Unknown
It's transformational because all those links you're looking at when you do a search, all those ads, that's the revenue. That's how you make most of their money. That's getting disintermediated in a huge way. But I think if we bring it back to accounting and what's what we're going to look at here, which is, you know, very, very soon, we're going to see this whole idea that it's going to enable entirely different business segments.

00:23:04:16 - 00:23:31:12
Unknown
The CFO consulting gig was not possible before everybody moved their data to the cloud. How could you possibly do the data gathering right. And and get in there and do the double clicks and see everything, connect the systems, connect all the dots and provide information that entire, you know, business was enabled by a new tech. And so this new tech is going to be a another major step function improvement in enablement of that essentially that profession if you will.

00:23:31:14 - 00:23:44:05
Unknown
So it's really, really exciting. It's kind of one thing stacking and building on another. But this one's gonna be a big one.

00:23:44:07 - 00:23:54:09
Unknown
Yeah, yeah. What a bummer. What a bummer. There we go. Out of date or they need some changes. Yeah.

00:23:54:11 - 00:24:10:23
Unknown
Yeah, that that's gone. And so I think we're at that next point of like literally we've went from access which was the biggest blocker before to suddenly there's going to be observability built in. It's going to observe what the changes were helping form, which is gonna allow us to use our CFO skills and CFOs where the chief financial officer is.

00:24:10:23 - 00:24:37:20
Unknown
They weren't the doers. They weren't the accounting manager or the clerical. Right. It's it's going to put the focus really back on the chief. And I know as a CFO before, he was constantly digging, constantly double clicking to try to find that needle in the haystack. The needle in the haystack is going to be revealed by the systems.

00:24:37:22 - 00:25:02:23
Unknown
Yeah. Just come to, actually email me. So you have any interest. I love having conversations with, with CFOs, accountants. Just and anybody curious of all ilk. So Chris, Chris at Lazio life's CIO dot me. Well, you got to listen to me check out we have the AI stuff is mostly behind the curtain. We're rolling it out to another thousand, firms this month.

00:25:03:01 - 00:25:18:20
Unknown
And then once our customer base is all through it, then we're gonna announce it to the public more, more broadly. But come have a conversation with us. Curious to see how firms are using it today or, you know, have a discussion about how people want to use it in the future. But I think now is the time.

00:25:18:22 - 00:25:30:21
Unknown
It's never been more exciting, you know, time to be in technology and accounting. So, yeah.

00:25:30:23 - 00:25:35:22
Unknown
Awesome. Adam, great to see you. Thanks so much.