AI Town Hall

Transcription:

Seth Fineberg (00:09):

All right, we ready to rock? It's the last big session of the day before the awards and cocktail hour. You might be saying, Hey, you're not Dan Hood, who are you? Alright, so if you're here this morning, I'm Seth Feinberg. I am the Founder of Accountants Forward. I was a former editor at Accounting Web and also with Accounting. Today I started my own consulting firm to see this profession moved forward. I have over 20 years steeped in this profession and I'm very excited to be the moderator for today's really kick ass session. I'm so glad that you're all filing in here for the AI Town Hall. Maybe you went to a couple other sessions forward talking about AI. This is going to give you all a chance to ask the great panel up here. Some questions. I have some of my own, but we're going to try to get to as many as we can because I know you all have some questions about AI.



(01:10):

It's very, very early days. Come on in, come on in, come on in. We don't want to waste too much more time. So I'm going to get to our panel, to my far right. Over here we have Hitendra Patil, Paul Bailey from CliftonLarsonAllen, Liz Scott and Jacob Sperry. They all have a lens, a perspective on AI in the accounting space, and we're going to just jump right into questions here. So one of the first questions that a lot of people will ask about AI is very simple. What is AI? Well, this is what if you were to kind of go and look it up, this is what you get you ready? AI is the theory and development of computer systems capable of performing tasks that historically required human intelligence, such as recognizing speech, making decisions and identifying patterns. AI is an umbrella term that encompasses machine learning, deep learning and natural language processing.



(02:25):

There's not going to be a quiz on that. What AI actually really is, it is the next level of automation for all of you and a lot of the fear and trepidation for some of us who are up here on the panel and some of us in this room, we've seen it before. Relax, it's going to be your friend. Ultimately it's going to help you in your day-to-Day tasks. It's going to definitely change a lot of how you work in the same way that Cloud did in the same way that desktop did, but currently in probably some somewhat scary and unimaginable ways. And so we're going to be here to talk about some of that here with this great esteem panel here. Everyone want to just kind of briefly introduce just yourself, what your perspective on AI is? We're going to start over here with Jacob.



Jacob Sperry (03:28):

Yeah, so hi, I'm Jacob. I like to say I'm a Recovering Accountant, so I'm now on the delivery side of things at a tech company bringing AI solutions to accountants. I kind of view AI the same way I viewed Excel earlier in my career. I was pretty good in Excel and it let me advance faster, let me do more challenging work because I could tackle more complex things and harder things. And I like to think of AI as a similar lens is the people who are going to be better at it faster are going to have more opportunities to grow and more opportunities to lead out at their firms and in their roles. So that's why I think about it.



Liz Scott (04:10):

I think that the presentation that we saw yesterday Spark, I think of AI as being that spark as inspiration to what it is that I want to be able to create. I don't want to start with a blank slate, so it gives me a place to be able to throw some ideas and be able to flush them out. So I think that AI in my daily life is something that, that's the place that I'm beginning is those conversations, flushing them out in order to be able to communicate faster.



Paul Bailey (04:41):

Nice. Paul Paul Bailey with Clifton Larson. Allen. We're the eight's largest firm and so we may have a little different perspective on how we use AI, but I'll give you two examples. My daughter got married on Friday, which was a big deal in life and I have never had a daughter married. So I put into ChatGPT, I said, what should the father of the bride speech cover? And it gave me boom, boom, boom, boom, boom, boom. Now obviously I tailored that it didn't understand who my daughter is, so that would be one example, real life example. The other, when you think about our practice and the way we're doing it, how many people love to onboard clients and the whole process of getting it set up in our systems, AI can automate that. That would be an example of here's how we can embrace AI in an impactful way that would be meaningful to everyone in this room.



Hitendra Patil (05:27):

Hi, I'm Hitendra I'm with Datamatics Business Solutions, an offshore outsourcing company exclusively for accountants. And for me AI is additional income because I wrote a book on AI and that's still selling. Okay,



Seth Fineberg (05:43):

Love the honesty up here. It's awesome. Yeah, so there's definitely, I think there's still people in the room with microphones that can help you guys answer some questions as well. So let's jump into it. So we already started to hear the ways that Paul himself is using it and a little bit on his firm, but Jacob, how do you think about AI? What have you kind of delegated to it so far? We're testing around because kind of in that early sort of testing, kicking the tires phase safely and we're going to get into the security issues in a bit, but what ways do you do?



Jacob Sperry (06:26):

I mean, most recently I was fiddling around with it and kind helping it do an interest rate interpolation calculation, how to explain that to somebody, but a little bit backing up. So yesterday I did two of the introductions for the last two sessions in here, and so I took the topic description, I took the bio and I threw it into ChatGPT and said, write me an introduction for this. And then I said, make it 60% as big and then I read it off.



Seth Fineberg (06:53):

How did it do?



Jacob Sperry (06:54):

Yeah, I think it worked out well. Better than I could have written it. I'm not great at that stuff. It's good.



Seth Fineberg (07:00):

Liz, how are you delegating? What are you delegating to ai?



Liz Scott (07:06):

I'm meeting hop like many of you here, and I think that part of what I want to do is set the scene whenever I'm showing up to that meeting what the agenda is going to be. And so that's a place that I use copilot in order to be able to assist. The other part is during my meeting, I have it recording that meeting, it's listening to that meeting. And so at the end of it it's giving me a summary so that way I can actually send to that client a summary of what we talked about who's in charge of which action items. It's very clear and it's happening very quickly. I think that if you were to rely on me to get to the end of the day to remember every action item that I needed to do for each of those meetings that I have, I would have a very hard time.



(07:46):

So it has increased the speed at which I'm able to delegate to our team and it has increased the clarity with the team members that are working with clients as to what the deliverables are that we need to provide to clients. So I'm a big co-pilot fan, and so I'm using copilot inside of my team's meetings in order to be able to produce those. And it happens instantly. It's just like with ChatGPT, it's able to give me those action items, summaries, create the emails, create those agendas. I'm sending out agendas actually ahead of time so that way it's setting that expectation that we know what we're doing.



Seth Fineberg (08:26):

Nice. So we're hearing about writing, we're hearing about organizing, we're hearing about wedding speech writing.



Paul Bailey (08:34):

That's right. And maybe another business use case. And I love the copilot feature where you can actually get a transcribed message, but does anybody have challenges keeping up with email?



(08:45):

Absolutely, because the thing I love about it is I can go in and I can ask ChatGPT, while we have internally it's called C-L-A-G-P-T, we had to build some guardrails around what we have, which I think we're going to talk about. But I can go in there and I can put summarize emails in the last week and any action items I need to do and it will generate, here's the different action items that I've been asked to do in an email, the date in which it needed to be done. Sometimes I'm delinquent in it, but it's a way that I can avoid reading 50 emails. I can read a quick summary of here's the key things that came to you by email and what you need to do. It's a pretty powerful tool. Yeah,



Seth Fineberg (09:21):

Hitendra? Yeah, You're delegating it things to AI?



Hitendra Patil (09:25):

I recently asked the large language tool, Hey, I need to write an email to accountants so that they'll buy services from me. And the answer came, don't ask me just be who you are, because every other outsourcing company is using me to write those emails. So I'm just who I am, so I express myself. But delegating to AI in terms of ideation is what I use it the most for.



Seth Fineberg (09:57):

And I personally like it for meeting note taking. I'm taking a lot of meetings these days in the work that I do, and I like having a reed and Otter right there. Otter is really great. It kind of gives me a summary of the high points of what's discussed and it bullet points everything and it's pretty cool. Alright, so this is the friendly part of AI. Another thing, questions that people have been asking too, it's how much of a distinction should we be making between AI say baked into software? There's increasingly increasing amounts of the work that everyone's doing, the products that they use, it's kind of baked in there and AI that for good ill AI that firms are deploying on their own as you are with Cliff and Lars and Allen and a lot of firms are doing their own sort of brand. So how much of a distinction should we be making and also how much of the latter do firms need to do? Let's start with Paul on that one.



Paul Bailey (11:02):

Yeah, just a little perspective. So two weeks ago we announced an acquisition in the UK of a company that is strictly an AI development company. And we have looked at that to say it's not that we're going to be on the bleeding edge, but we need to be on the cutting edge of what AI is doing and it's going to allow us to do in our business and in the industry. And so we have looked at that to bring in the tools that we believe we can build ourselves in a way that's going to be impactful to our business. Our business looks different than most. And so we've looked at it to say, here's the different tools that we're going to have established over a short period of time to really improve the efficiency and the effectiveness of the work that we do. We've heard a lot over the last day and a half about staff shortages. We're looking at this saying there's opportunities that we're going to reduce 20 to 30% of the time. Our goal is that there's zero data entry in any work that we do. And if you could imagine a world in your firms where there's zero data entry, that would be pretty impactful. My guess for your business. That's the direction that we're headed.



Seth Fineberg (12:05):

So tasks, right? This is what we're talking about when we talk about automation. It's tasks, it's things that take time to do in our lives. Liz, in terms of you've been working with software for quite some time. The distinction between the two



Liz Scott (12:24):

I think of baked into software or going independently. I think of going to ChatGPT as something that's independent, and I think about that as being something that has a whole lot more security risk. I think of those as being open AI tools. I think of AI that's baked into software as being more of a closed system. So the chances of having more security breaches are reduced because if you're thinking about where is all this information coming from, each time that you're answering a question or each time that there's an update, all of that information is being scrubbed from resources that are out on the web. So I don't want my team to go and feed one of these open sources, a bunch of client information and feeding these data lakes. So that's the distinction that I make is is it open or is it closed?



Jacob Sperry (13:16):

Big risk.



Seth Fineberg (13:17):

Yeah, for sure. And I definitely want to get into that. I think everyone's kind of thinking in the back of their minds like, okay, but what should I be worried about? We'll get to that in a sec. But in terms of how you're using it, Jacob,



Jacob Sperry (13:34):

So we're a vendor in this space, so we sell these kind of tools. So Paul highlighted a couple of the core use cases that we hit on, which is around data extraction, but also data validation and coupling that with a workflow, having a preparer review or just a good clean way to know that the data's there. So to your point about how much of this should you be getting from an external source, how much should be building internally, there is a classic build versus buy decision of a company with any tech stack that you're doing. And so we're excited about the opportunity to bring to firms that don't have large r and d budgets or dedicated engineers to do that, of being able to bring some of that, some of the benefit from that technology out through an aggregated way and developing that technology ourselves and then selling it to our customers. So we're actively doing that today.



Seth Fineberg (14:29):

And in terms of some of the baked in, there's pretty much all of your core accounting programs and services that you're using. It's in there, a lot of mixed reviews. It's still very, very early days of course, but just hearing people just wanting to just shut it off in some cases. But on that note, and I'm going to go down the panel here and I'm sure you guys are all thinking this too. What is the worst case scenario for AI in accounting? And then we'll flip it. What's the best? Now we're already hearing some pretty positive stuff, but there's still a lot of risk and a lot of unknown as well. I don't want to lead any more than that, but what are we thinking? Worst case scenario, let start down you. Yeah,



Hitendra Patil (15:25):

I think the worst case scenario is when your clients think that AI can do everything that an accountant does, and that perception can be accentuated by the advertising that goes around AI tools. It's a lot of hype out there. And if that perception kind of strengthens, you have a problem of explaining, okay, here is what we can do that AI cannot. And that answer you got to be ready with. And I think that's a major challenge because it does happen that the general perception is AI is like a magic thing and it can do anything in the world. I've been trying to tell AI to put money in my bank account, it doesn't. So I have to speak right Yet.



Seth Fineberg (16:09):

But Liz, if you run into any sort of very any questionable scenarios where it's like, Ooh, I could see where this could go really bad.



Liz Scott (16:18):

I think that it has hit our industry so fast. I think that you were talking about in a course today that a hundred percent of accountants are in some way touching ai. That's a lot. There's no guardrails. So that's the part to me that is scary. I think that we definitely need to be able to have some bumper rails to be able to say, what are you doing? What are you not doing? What are you feeding to it? What information are you giving? Because of course, if you are having it draft some emails for you outside of a Microsoft environment, what information are you giving it about that client? So I think that those are the things that are scary.



Seth Fineberg (16:56):

You got to feed it,



Liz Scott (16:57):

So be careful about what it is that you're feeding it. But I do think that we need each of us as firm owners to be putting into place some policies. Yeah,



Seth Fineberg (17:07):

I mean because if anyone has ever tried or inside their own firms to see, oh, well, let's see what it does with this particular set of financial data. You have potentially leaked that out into the world. That's no good. Yes,



Jacob Sperry (17:25):

I could promise, I'm not saying this to try to generate leads, but there's a difference between going to ChatGPT and then going through a company that there are ways to anonymize the data, make sure it's secure. So there are protocols that we follow and we're going and using gen AI where there's a no retention, so we're not handing the data or the rights to that data over to these data lakes that are creating the LMs. So there's ways to do that that are different than my basically free version of Gemini that I play around with. There are different ways to interact with that,



Liz Scott (18:02):

And that's where I think of the baked into that close parameter.



Paul Bailey (18:06):

One, the scary thing is, I don't know if everybody in the room is doing this, but if your staff in your offices have access to ChatGPT, you really don't know what information is put in there and what questions that they've asked. Because the younger generation is engaging in this, and I actually think my son is using it for homework, but they're engaging in this in ways that most of us wouldn't think about and maybe don't understand the risk. And if you haven't taken that into account in a risk manual or the approach that you're taking or had your IT group locked down, so it can't be accessible off of the work computers, there's a chance that you've put client confidential information out into the world.



Seth Fineberg (18:41):

Yeah, there's definitely a lot of things out there that people think are a little creepy, little black mirror ish for those who know that series. People have created bots of themselves to answer basic questions and things. I really want to hear from you guys too. I got a ton of questions, but key concerns right now, top of mind, anybody, let's get this panel going.



Audience Member 1 (19:07):

So in an earlier session we were talking about the data security and how the engineers aren't even really sure how these models are doing some of what they're doing or why they're doing what they're doing. So you're talking about open and close models, but we talked about one of these models leaking data and they're not sure why. So how concerned do we need to be about that with, I mean, there's several vendors here that have AI tax prep software where you're feeding it the most sensitive information you could possibly think of



Seth Fineberg (19:42):

And people losing 10 40 business and things like that.



Jacob Sperry (19:46):

I could speak to the technical side of that unless somebody else, okay, I don't want to hog your time, but yeah, so there's a process that we follow. I'm going to try and sounds smart. So I'll first start by just saying I don't remember what this acronym stands for, but there's a process called RAG where effectively you're stripping down specific identifiers in the data. You're passing it to an internal database where you're converting the specifics of the question into a different format entirely. So what actually goes out into the gen AI isn't specifically identifiable, so it's kind of like hashed or mixed up. Then with the contracts that we have with the LLMs, there's no retention policy. Hypothetically, you shouldn't give them anything anyway, but if you did, it's not retained. It's not part, they're agreeing to use it within the model, but not in a way where it's learning or they're retaining that, but there's a scrubbing and cleaning that's done before that to hand it over in a way that is protected and anonymized.



(20:50):

We have a guard, we started out with lease accounting, so we have a lot of controller type customers using us for leases. A lot of maybe even some people here use this as we're the managed service provider for how they do lease accounting. We've got tens of thousands of leases in there. And one of the guardrails that you can have within our platform is you can have a gen AI conversation about your leases or about the documents that are in there, and that's specific to you within our company, it's about your leases. It's not about all of the leases in the system and that information doesn't bleed up. So there are methodologies that can be followed. They take engineers and structure to put in place, but it is possible to do that today. But actually the other side of that is that the model's slower to learn because of that, right? So no customer is going to sign a contract with us where we say, Hey, you give us all your stuff and then we'll give it away too. So the models are slower to learn because we're retaining and protecting as much confidentiality and privacy as we can. Sure.



Seth Fineberg (21:50):

Anyone else? Anything to add to the question? We got more?



Liz Scott (21:54):

I think that backing that up being somebody who's not a software developer but an individual with a firm, the thing that I think is important is that you've got these AI use policies in place, and that's what I think that we are needing. And also to work with tech companies that are doing things that you can ask them, what is your security policy? What is the policy of where the data is going? Very good question that you just asked. What's happening to that data? How is it being scrubbed? Those kinds of things. I think those are the kinds of questions we want to ask



Hitendra Patil (22:25):

Tech partners. I think when you use things like ChatGPT for finding something for the client's situation, ask yourself a question, do I need to really applaud the whole financial statement or a tax return to get one answer that I'm looking for? Or can I ask a very specific question by myself anonymizing the information that I'm uploading to a third party tool? That's a better way to look at things as such. If you upload, let's say a social security number into ChatGPTs, it is not going to just relate to somebody else in the world. It's going to tokenize it. So once it goes in the machine learning algorithms, it's unidentifiable, but it resides in that transit database for some time 30 days or whatever. And that's the danger who has access because that's the real raw information sitting there and they have to keep it to make sure the databases are not misused. So at that point in time, it's just like any other system, immigration that uses your passport information, visa information to put in the system. If that system is hacked, that information goes. That's what this little in-between database is, and that's why you don't want to put I information in any third party tools. So again, find the question for which you want the answer for. You don't have to put an affidavit into the system. Here is the whole information.



Seth Fineberg (23:44):

Right? Heather, you had a question? That was, is it okay for me back there too? Okay, go for



Audience Member 2 (23:49):

Okay. So one of my biggest concerns is the reliability of these tools. So we see these tools popping up that have CPA in the name of the app, and if you read the user agreement for these particular apps, they basically indemnify the company against any bad information or misuse of the information by the user. And so my big fear is for CPAs that use their due diligence and actually research the answers and understand what it's saying because we have that expertise, I think the general population that's going to rely on this data is at risk because they don't know the difference and they're also not reading those license agreements that say basically this tool could give you misinformation and they're going to rely on it. So I mean, that's my biggest concern. Is that something that you're seeing as well?



Paul Bailey (24:52):

I think it's a great comment. One of the things that we've done is we refer to it as human in the loop. And so we are not extracting human interaction from a decision. And so we'll use the tool because it's a great resource to go out and you can do tax research, you can put in very technical question, it will spit back to you, here's all of the internal revenue code, all of the regs, but you need the individual with the expertise to be able to go through still in the loop, it's accumulated it and now we're going to read it, we're going to review it, we're going to make sure it's right. No different than if I had a manager or a senior pull that same memo together. I'm going to be able to take it and review it.



Audience Member 2 (25:28):

And are you also advising your clients about use of AI to answer technical accounting questions?



Seth Fineberg (25:35):

I was going to say I think it's on the CPA to kind of go. Okay. Be upfront. Sorry. Yeah,



Paul Bailey (25:41):

Yeah, no, I agree. So clients that want to use it, yes, we're giving 'em advice. Hey, that's great intel. You can go get that, but you still need to have somebody who can interpret it. I think it was your presentation earlier that said, look, regulations that change today aren't going to be known in the system because it hasn't learned it and it's going to take a period of time before AI is learned. And so that's where the human interaction has to be there so that we can apply, well actually this is the new regs that just came out and here's how you're going to interpret it.



Seth Fineberg (26:09):

Yeah,



Hitendra Patil (26:10):

It's like today what software use, let's say AI was not there, you're still using automation, analytics and whatnot, all that software. If a client comes to you and say, you're just using software and generating this intelligence and insights, what the heck I should be paying you for? You're just using the software, then you tell, look, it's my expertise, experience, the ability to connect the dots, all of that goes in AI is just maybe a faster tool for you to do it. So I recommend that you tell your clients, look, we are going to be using ai, but we are responsible for what we tell you. That's it.



Audience Member 3 (26:43):

Good point. There was a question in the back there. Yeah, I just want to follow up on, I believe it was Heather's question about the service agreement. One way to I guess transfer risk when you're engaging a third party is to have an indemnification clause in a service agreement. So if there's a breach of your system or some liability issue, it's on you, you're going to indemnify me. Or at the very least have some sort of clause in there that requires them to have some sort of coverage, some sort of insurance, whether it's cyber insurance, liability insurance, something where you can transfer to risk. And I know a lot of firms are getting more successful at doing that with a lot of other third party providers. Is that the same? Is there same traction with dealing or contracting with a third party gen AI platform?



Hitendra Patil (27:37):

Yeah, I think third party gen AI tools not trained on accounting specific things. Obviously the risk is much wider because you just have no idea what those tools got trained on. If somebody comes and says, we took the entire IRS website and trained my tool on that, you might actually feel that, yeah, this should be good enough from a research point of view. But I think the more important factor is what is it that you use AI for and what's the impact on the client's financial or tax situation? I think that is where it goes to. If it generates, let's say a wrong financial statement, somebody checks it, the downside is limited. But if you don't check anything and use it for significant decision making at the client end, you better not use just the AI tool and rely on it. You obviously have to put in your own expertise in that.



(28:31):

From an insurance point of view, I do not know if insurance companies will cover much wider unknown AR risk. They would definitely want to know which AI tool you're using, what privacy policies, those tools have you considered those? Are those in your processes or not? So essentially insurance will just check whether you're just blindly implementing AI or you're doing it very consciously. You've done lots of groundwork and all of that. And other than that, I don't think there should be too much of a concern on that indemnification because if you control this animal to go in a particular way, I think you should be safe enough.



Audience Member 3 (29:07):

Yeah, but I mean, I hate to say it, but I see it's great, but I can see it breeding a certain intellectual laziness from people who are going to rely on it and they're not going to do the human review to figure out whether it's accurate. And if it's wrong and you produce work product, the client isn't going to sue the third party Gen I platform. Correct. It's going to sue you. Exactly. So is there a way in the service agreement, because you're going to enter into a service agreement with a third party platform, is there a way to transfer that risk and go, if it's your fault, I'm still going to get sued. But you're going to share in the risk too, and that's where I think an indemnification clause would serve



Seth Fineberg (29:49):

A lot of people.



Paul Bailey (29:52):

I look at that similar to an accounting software. And so if I have a tax software solution and I put the number in the wrong box and it created a penalty for the client, is that the software's problem or is it mine? We still are going to have the fiduciary responsibility. If we think about AI as being an augmentation to our team, that ai, we have the responsibility no different than you have for supervision and review of your associate. You're going to have that same responsibility. And so that's the part, and it's going to be a training, and I agree with you, it's a risk that people are going to get lazy. They're not going to want to do it, but we've got to have the protocol in place to make sure that any product that comes out of that, that it's augmenting the person that has the judgment and has the technical ability.



Liz Scott (30:35):

And I think it's doing the research in a lot of these situations. I can ask a question and AI is going to be able to pull up an answer for me, but you've still got to go back through and verify it in the situation like you're talking about is producing a product. There's still got to be an associate who's going through and reviewing what was the output of that information before it's given out otherwise, what is the purpose of our roles, our jobs? So we're the experts.



Hitendra Patil (31:02):

Many years ago did we hear this trust but verify? I think that applies now.



Liz Scott (31:07):

Yep. Trust but verify. Yep,



Seth Fineberg (31:09):

Absolutely.



Jacob Sperry (31:10):

Just specifically how that works in our solution is the AI. Let's say you're automating a test of details, you're looking at 70 transactions, the system will flag for you. Here's the stuff we automatically matched, and then part of the cycle is actually having a human being accept that to look at it and say accept. So that's the preparer step. And then there's a second reviewer step. So we view it more as the AI is not the preparer, it's not extracting a human from the process. I'm going to steal that by the way. That's great. It's not extracting a human from the process, but it's enabling and making that human who's being the preparer more efficient in that task and so that the responsibility would still sit there at the prepare level, but that's just functionally how these companies can exist because an indemnity clause can just destroy a small company.



Seth Fineberg (31:58):

So for what we can see right now, obviously, and it's good to be talking these things out because there is still a lot of unknown in the short and possibly mid term, what do individual accountants and even smaller firms need to be doing? I mean right now it still seems to be in this sort of testing phase, but carefully and with some information, hopefully you're getting some of that up here, but could we kind of go down the panel here and say, Hey, look, we do want to be more efficient always. And we're starting to talk about some ways where, yeah, this is probably a good scenario for accountants to be using AI.



Hitendra Patil (32:53):

I think the simplest thing to do as a starting point is look at the solutions that you're using at your firm currently, accounting, payroll, tax, audit, whatever is your core line of business. And in that solution, talk to the vendor and say, what is the AI thing that you are implementing within the solution baked in or otherwise that's your, you're talking build because you're producing all of your revenue from those software. Don't go outside of it. Whatever comes in, those software is going to be applicable to you one fine day, whether you use it or not, maybe they're already using it behind the scenes, you must know. So first thing is being aware of those things that's right within your house.



Paul Bailey (33:31):

I think for a lot of the firms that aren't going to develop things internally, it's going to be who are your vendor partners and how well are they adopting the software? And then rethinking about, I mean, how many of you have changed your tax software in the last five years? Everybody hates doing that. So there's probably not a ton of people that have. And so who's that vendor partner going to be that's going to help provide AI solutions and what is that going to look like? I think that's going to be the key. We're probably in this room, there's probably less than 3% of us are going to go develop our own internally developed AI solution. It's going to be off the shelf. And what are the solutions? What's the quality of the vendor there? Are they reputable? Do they have good controls? Do you understand it? I think that's a big part of it.



Liz Scott (34:13):

Carbon, they just released last week an excellent study and it was talking about firm size



Seth Fineberg (34:19):

Practice management platform. If you guys dunno who carbon is.



Liz Scott (34:22):

Yeah, great workflow tool



Seth Fineberg (34:25):

Workflow. Yep.



Liz Scott (34:27):

The thing that they released in their study though was talking about firm size. So the larger the firm, the more likely the team has been trained on AI tools. What I found interesting about that is whenever I went out and I looked for AI training in our accounting industry, there's not, so I wonder what that training actually looks like. So in some firms I think that you're going to have some actual resource documents, there's going to be some official training, and then in other cases I think it's going to be, hey, look what it can do. Did you see what this dislike did? And it's going to be like these kind of wow little show and tell moments. So I think that it's



Seth Fineberg (35:04):

About where we're at right now,



Liz Scott (35:04):

Right, I think it's going to be very important to do what you just said, rely heavily on your vendors. I think that a part of that is going to be our IT security solutions that are out there. They're great resources whenever you're thinking about what is that you should be thinking and considering and what should be some of the measures that you should be going to be able to find a good AI partner.



Hitendra Patil (35:29):

One good question to ask your vendor is when they say it's AI, say what percentage is AI? What percentage is human assisted AI? And see what the answer comes back.



Seth Fineberg (35:41):

That's a very good point because there's definitely solutions out there that fit that scenario.



Jacob Sperry (35:46):

I'll just throw it really quickly. KPMG had a study that came out a couple of weeks ago that said 77% of CFOs expect the audits to be getting more efficient, either moderately or very more efficient as a result of ai. So it's only a matter of time until you start to get questions. How are you getting more efficient? How are my fees going down? How's the scope of this work getting better? How's this getting more automated? So I think staying aware of that and finding small ways to drive small efficiencies before, because we are very early days on this stuff, there's going to be small opportunities to gain efficiency before there are huge ways to gain efficiency, but it is going to be an expectation that we're all going to be feeling from our customers and that's why we need to go to more fixed fee arrangements and more of a subscription based revenue model because as we get more efficient, if clients are expecting us to give them our efficiency, it's going to be a world for



Seth Fineberg (36:38):

Our industry. You can't rely on that hourly. If anything, I felt that the coming of cloud, when cloud technology was able to do things in real time within minutes, things that it took a lot longer to do, it definitely really started to spell bad news for the billable hour and the more efficient that you are. So you're definitely one of the ways that I personally see it changing the profession is the way that software and automation already has. It's just making you more powerful. It's making you more valuable because it is taking tasks away as long as you don't feel like what you do and the value of what you do is in a task maybe at one time, accounting was, we talked this morning a little bit about just even the word accounting and what being an accountant is. You all have the opportunity to say, no, I'm actually, this is what I do, this is who I am. Yes, I am an accountant and I am in accounting. But so ai, would you all agree that AI is going to sort of reshape what accounting is?



Liz Scott (37:51):

Absolutely. I'm going to say I think that what AI is, the possibilities that I wanted. When I first went to work for a firm in that CPA firm, I wanted there to be a lot of advisory services. I come from a family that had small businesses and so when I found out that that was not part of our job duty, I was devastated. I was like, how on earth can we actually effectively do our job if we're not actually offering advice? So I see a lot of these tools that have AI built into it and it's allowing us to be able to start to offer those kinds of conversations. There's some reporting tools out there and inside of those reporting tools, it's looking at the data and it's able to see trends. We're super busy. We're not able to always see month over month over trends.



(38:40):

So some of these tools are actually doing some nice jobs of color coding. This is a good thing that you're seeing right here. You had increased profits this month, why it then maybe would tell you had better sales that month. And so it would give you things to be able to talk about with your clients. And that's the piece where I feel like now we get to those conversations that I wanted to have years ago with clients and now it makes it where the firm that I worked for at the time, what they said was, we have the fear of getting it wrong and so they didn't want to offer bad advice. But if it's coming from trends, if the words are already being generated as to what to say to these clients, that's allowing you to have these really positive conversations. And I think that that's what powerful and



Seth Fineberg (39:28):

Valuable ones, ones that people are willing to pay for. I think there's another question in the audience.



Audience Member 4 (39:35):

So I have built some automations with no code and OpenAI and it is wonderful. So it's made my processes so much more efficient. One of the things that I keep in the back of my mind is that this technology is incredibly inexpensive right now it's $20 a month for the ChatGPT paid plan. So one of the things that I think about is that it's not always going to be that cheap and you're building your AI tools proprietary within your firm, which is really smart by the way. But at what point are we putting ourselves at risk of we build these automations, we start relying on them, they're part of our business model, and then they jack up the price to a point where it's not sustainable for us and it affects our practice, it affects the going concern of our practice and our clients. Do you guys think about that at all?



Paul Bailey (40:32):

We know we spend a lot of money with Microsoft right now and we are a Microsoft shop and the prices every year continue to go up and it's one of those, we had conversations about it of we've got to be willing to pass through pricing increases to our clients. This room, maybe not everyone in here, but generally speaking, we suck at price increases. We did during covid. We have the last five years we have for the last 50 years. We need to be thinking about that differently than we ever have because we're going to be more efficient, costs are going to continue to go up. They're going to be costs different than what we have today. Right now, we probably all could say the profitability of a firm is a third to our payroll. Well, that may not be the case long term because we're going to have other costs that are coming into the equation in a huge way.



Seth Fineberg (41:15):

So let's address the, I don't know if it's quite an elephant in the room, but when you talk about ai, the first thing, and I really want to put this to bed, we are at the firm growth forum. We're looking for ways that we can change and we can evolve as a profession. I really want to kill the discussion of AI taking your job. Let's get into that while we have a few minutes left. Specifically, at least how I see it is that as long as your job isn't defined by a task, you should be just fine. That said, what jobs is AI potentially going to do away with? Cause you to pivot, evolve, what have you. We're already hearing evidence of that out in the market today, particularly with some 10 40 work and other things that really automation can just, it'll just out automate that set of that sort of job task. So let's get into it.



Jacob Sperry (42:29):

We'd like to say that AI's not going to take your job, but an AI empowered accountant might take your job. So from a competitive perspective, we're competing. Some of us are competing amongst each other, right? We're competing for the same work and being able to do that more efficiently, being able to have a better quality of life for your employees because the jobs are better because you're more AI enabled will help you retain the right talent and you'll be able to go to market and win. That doesn't mean that it's not going to go through some level of chaos. That's part of the benefit, the efficiency of growth as part of that result. But sure.



Seth Fineberg (43:05):

Liz, what do you see it doing away with and what do you see it evolving into or benefiting



Liz Scott (43:13):

The 55 hour work week go away,



Jacob Sperry (43:16):

Getting down to 55 or getting,



Liz Scott (43:20):

I think that what we're going to get to is I think we're going to get to a situation where somebody goes, why are you so inefficient? So what you were just saying, hiring somebody who is an accountant who is capable of utilizing these tools and understanding them, that type of worker is going to be much more valuable. They're going to see some of these tools that are out there. They're going to be able to use some of these workflow tools that have AI built into it, and they've got really solid processes. They didn't have to start that from scratch. They were able to use a tool for that. Boom. They just sped up hours of their job being able to use reporting analytics. I feel like those are all those key components of where our profession is going to get stronger. The jobs that I worry about are the bookkeeping jobs. I think that those are going to have to be redefined. So I'm not sure where exactly those are going. I think that maybe you even have some storytellers that are inside of that group. I loved it yesterday whenever Terrell Turner was talking about he doesn't need to hire a whole bunch of super degreed workers. What he needs is people who can interpret the data in order to be able to talk to the client.



Seth Fineberg (44:31):

Have you seen anything like that in your firm where you're just like, yeah, there's actually stuff because your firm's big enough where you can maybe see, yeah, the way that we did things, we don't have to do it that way anymore.



Liz Scott (44:43):

The way that we did things a month ago is different.



Paul Bailey (44:45):

Yeah, a hundred percent. And I can think of one. I think AI is going to make our industry and the careers in accounting a sexy career. Again, we used to have that we haven't for a long time, and I think this can bring it back because people can engage on, I need to be doing something other than re-keying data and moving something from one box to another box. And so I really think this is an opportunity that we're going to elevate our industry and the appeal to the college students because we're embracing technology in an impactful way. I could go back to when I started, I didn't have anything to do. So hey, go learn how to do a tin key. There's probably still people in here that use a tin key, but most all of us are in Excel and using Excel. And then you think about, well, we went from typewriters to word processors to a computer



Seth Fineberg (45:31):

Desktop.



Paul Bailey (45:32):

Yeah, exactly. So we've moved through those phases. Generally speaking, the people that we're able to adapt are still very successful in what they're doing. And we as an industry are going to have to be able to retrain and redevelop our people so that they can apply the skills. They have great skills. How do you redeploy those into a way that we can leverage AI differently? It was commented, I think it was on the young professional panel that what you learn in school, you don't actually apply in the industry. I think that's going to change because what we've done with accountants that start in the industry as, okay, you're going to go and you're going to create a tax return and here's the W2 and then all of this. Go do the 10 40 follow last year, same as last year. Well, that process isn't going to be needed anymore. And so we're going to have to be able to train and get our people engaging with clients and providing value and insight into what we're doing. And if you can provide insight that's data driven, it's going to be way more impactful than what we've ever done before. And I think that's how we can keep the value of our industry up.



Seth Fineberg (46:28):

So it's no longer chime. It's your expertise, it's your insights.



Hitendra Patil (46:33):

Yeah, I think the simplest point of starting to look at AI as to what is going to do for you is go and look up the dictionaries for two words, accountant and accounting. Whatever it defines in there is what is there on AI's radar to gobble up, obviously that needs to be done. Okay. It says collect, organize and interpret financial information. That's the definition I believe from the 15th century. A dictionary says that actually Luke. So if that is what is going to be done by ai, what next comes after that interpretation is where your work should shift towards. And that's like your sketch, your brains, if you're doing that work that's defined in dictionary, you're in danger.



Seth Fineberg (47:18):

They said that about cloud too. I mean cloud, oh, if you don't embrace the cloud, you're going to be gone. I think that to a degree, I think it has been embraced. We don't even necessarily even talk about it anymore, but it's been said before about pace of change. Never going to be slower than right now. But I think with this level of automation and the degree that it can automate things, it is going to ultimately redefine what accounting is. And so I think we have an opportunity right now at this time, at this date, to really start re-imagining how we work, how we want to work. We've got about a minute or so left. Any final thoughts down the line? Otherwise we've got some awards to get to and then some cocktails as well.



Paul Bailey (48:10):

I dunno if I'll be able to read this or not, but there was an Accounting Today article that came out and it talked about the number of firms. It said 12%. This is referring to the profession. 12% said they see no role for AI in accounting at all. 27% said they don't want it in their personal lives either. And it may not be surprising then that 70% of accountants think AI is advancing too fast. We've got to get on board. We can't be that industry that wants to sit back and let everything pass us by. This is the opportunity for us to really engage. And if you're one of the 12% that don't think it's going to apply to our industry, I would make a bet with you today that five years you're going to change your mind or probably even six months from now, you'll change your mind on that.



Seth Fineberg (48:53):

I definitely remember that with cloud, it was just, oh yeah, it's going to pass.



Liz Scott (48:57):

I think we are the finance industry. I think that whenever you look inside to see how businesses are doing, we know we could tell during covid those stories. We know what was happening in those individual lives. And so I think that we are the ones who are seeing what AI is going to have as an effect on business and society. So I definitely think that we should be embracing it because that's how we progress and teach.



Hitendra Patil (49:23):

I think till such time, ChatGPT charges $20 a month per user, you're not in danger because they need more paying users. So they think they're not going to replace you as long as they're charging you $20.



Seth Fineberg (49:38):

I mean, it's not always going to be, but that is



Jacob Sperry (49:43):

I think one of my first jobs was 10 keying a payroll register, breaking off the tape and stapling it to the print up of the register to make sure I did lots of that. Eventually I put that in a spreadsheet and every time I did the payroll register, I made the formulas a little bit better. And then eventually I was taken into the office of the CFO and he explained to me that what was a 20 hour week job doing payroll math was now a five hour CSV export and I didn't have a job anymore. He gave me a month pay and I had a new job three days later. So actually worked out great, but on the back of that, but on the back of that, the skill that I developed from that led me to the next step in my career and that skill with Excel when I was finally A CPA, it helped me to advance faster in my career because I knew v lookups better than most of my peers did. And I really think that AI is going to provide that same opportunity for the people who are early adopters and creative and imaginative. It's going to create opportunities to stop working so much on the low key side, but also to be really dynamic and change your practices and change firms. So I get really excited about it



Seth Fineberg (50:50):

Long term. Yeah, we as a profession really need to collectively have the hive mind of working less is a good thing. Automation is good. Automation is going to help you actually be more valuable. And let's guess what that is. Ultimately what you want your clients to pay you for. It's your knowledge, your expertise, your unique view on financials, on business planning, on wealth management, on all kinds of other things you could be doing to help your clients because that's why you got into this profession to begin with, isn't it? To help. That's what AI and other levels of automation to come are likely to do. I want to thank Hitendra, Paul, Liz, and the AI in the room. Jacob? No, he's a



Jacob Sperry (51:51):

I'm a fact.



Seth Fineberg (51:53):

He's a real guy. But thank you to the Trullion for sponsor having a session. They're here. Definitely come talk to him about what they can do for you as well. I've been Seth Feinberg and I cannot wait for this award ceremony to come up. Is that Dan who I see back there? Yeah. Now he's not here yet. He is awesome. Thank you to this awesome channel. Thank you. Thank you. Thank you for your questions too.