Kelly Fisher, chief practice officer at Top 100 Firm Wipfli, offers a practical look at artificial intelligence, current use cases, and how accounting firms should be thinking about it.
Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.
Dan Hood (00:03):
Welcome to On the Air With Accounting. Today, I'm editor in Chief Dan Hood. For a long time, there's been a lot of wild and irresponsible talk about artificial intelligence and its potential impact on accounting, but as we come closer to having a meaningful understanding of what AI is and what it could do and what it might become, we're also getting a better sense of what it might mean for the profession here to help us with all that, help us come to a closer understanding of that is Kelly Fisher. She's a chief practice partner at Top 25 for Wipfli. Kelly, thanks for joining us.
Kelly Fisher (00:28):
Thanks, Dan. Excited to get a chance [00:00:30] to talk to everybody about the, I'd say pragmatic application of AI with all the hype that's out there.
Dan Hood (00:36):
Excellent. Well, that, that's a great focus for this episode because pragmatic approaches are few and far between. There's a lot of hyperbole. There's a lot of, like I said, wildly irresponsible talk and a lot of fear and uncertainty and just lack of understanding. But also, as you say, there's some people who are predicting that it will, I heard people predicted it will end death will, there will be no more death because of ai, and I'm like, I figure I'm going to set my expectations [00:01:00] somewhat below that. Maybe you can help us get narrowing on an even more realistic definition, but why don't we start by talking about, when you think about AI and accounting firms, what are you excited about? What are you scared about? Where is your window of feeling about it? Is it the greatest thing in the world? Is it the worst thing in the world, or is it something somewhere in between?
Kelly Fisher (01:20):
I definitely fall on the optimist side of the house. I would say from the doomsday, yin and yang side of it, I'm definitely more on the light side of things. [00:01:30] I do think that it's going to help the world frankly, solve some of the tough problems that we're facing for us in the profession. Honestly, I think what I am most excited about is the potential that AI has to kind of magnify, let's call it the passion and purpose that I think exists within the profession that maybe has lost some of its light over the past couple of years just because of the complexity that we're all dealing with between [00:02:00] the regulatory changes that are coming down, trying to stay up to date on everything, transform our own firms, help our clients work through everything that they're going through. Also, I think that all of us, from an accounting firm standpoint and all of the people that we're working with, it's just been brutal, frankly, the amount of data we have to process, information we have to process, and I think that it has caused us to have to spend a lot more time maybe on things that we consider a little bit more internal [00:02:30] and not spend as much time on the things that I think drew a lot of us to this profession in the first place, which is helping our clients out there.
(02:39):
I think you look at what accounting firms across the country can do, frankly across the world, whether you're a smaller firm, I mean, we all have the ears and minds and hearts of the C-suite. I mean CEOs and CFOs of Main Street companies that are impacting our communities to the biggest organizations in [00:03:00] the world. I mean, think about the more time that we're spending with people helping them, that kind of multiplier impact we have, and I don't think we highlight that enough as a profession.
Dan Hood (03:10):
No, no question. And it's weird because everyone's aware of it in the profession. When you talk to other accounts, when accountants talk amongst themselves, they all understand this, but they're not as good at getting the must out to non-accountants is maybe the way to put that. So yeah, Eddie, it's interesting. You talk about the number of things that are getting piled accounting, the [00:03:30] amount of work that's getting involved, the amount of change that's happening, and when you look at the way technology has helped firms keep abreast of that for the last 30 years. Just basic technology, not ai, but it just keeps piling up, and so hopefully AI can help us with it going forward. With that in mind, how do you see AI being applied in accounting? What are the use cases, most common use cases when you think about it, what comes to mind?
Kelly Fisher (03:53):
Yeah, I mean, I think with that whole kind of freeing us up aspect of it, I think a huge part of it is just that internal [00:04:00] monotony aspect, whether it's the processing, the information, I mean, there's a lot of use cases and applications that I think accounting firms can and should consider, and it's not even using that edgy gen AI type technology. I think most people maybe don't realize how long a lot of the technology from an AI standpoint's been around. I mean, a bunch of this stuff has been available for frankly decades. I think what has been great about [00:04:30] it is that it's highlighted stuff that I think a lot of us could or should have been doing for a while to take a lot of this monotony out of things, particularly around automation. So I think that that's probably the first thing from a use case standpoint that a lot of firms should take a deeper dive on is looking into basically anytime you're moving information files data from one system to another across people notifications, that's usually probably a very [00:05:00] quick easy flag for you to think about building an automation, and there's so many kind of low code tools that have become available on the market in the last couple of months that there's a really low barrier to entry on some of those use cases.
Dan Hood (05:16):
Right. Let's dive into that a little bit. Interesting. There is definitely, it seems to, there's an overlap between basic automation, just let's having a computer program doing it and AI doing it. Is there a difference or is it shade into one and then [00:05:30] out the other? Is there a different application for AI in automating tasks than just a basic software program?
Kelly Fisher (05:37):
Yeah, I think there's definitely different layers to it. I mean, with some of the stuff that we've been building, obviously with the low code solutions, I'd say that's at the most basic level, not a ton of risk as long as you have the right stuff in place from a cyber standpoint and your data infrastructure, what I think has been able to become more readily available because of the generative AI movement, particularly around [00:06:00] what large language models have been able to do, is power by putting some of the generative AI capabilities on top of some of the machine learning that's already been in existence, it takes you to a different kind of level on what you can do from a use case standpoint. I mean, we've done a number of things over the past couple of months from a Wipfli standpoint, like partnering gen AI with machine learning to read invoices that we read, [00:06:30] notices that we get from an IRS standpoint, and it drafts the first response for standard notices from looking at things.
(06:38):
We've done a tool where we're leveraging generative AI on top of the Python library, so there's an Excel plugin now that has Python, which gives you access to all of those libraries. You can use generative AI for people who didn't grow up in a programming background where they can use natural language to say what they want to do. So the gen piece is small. [00:07:00] Then you just leverage the machine learning and apply it against something like an Excel database, which a lot of our audit team is a lot more comfortable with. So it's really kind of putting the pieces of the tech together that'll give you a lot more of the use cases that frankly you could deploy at scale today,
Dan Hood (07:18):
But that translation right of natural language into something for Python. As you say, not many people are familiar with Python, but that's a huge translation role. That's amazing, [00:07:30] and it's one of the things that gen AI seems to be at this point seems to be one of its key skills is either generating copy one way or another or translating vast quantities of information into something useful, but that's very cool. That's a neat specific example. Any other use cases you see beyond that, beyond automation and that connection of data?
Kelly Fisher (07:50):
Yeah, I mean, I'd say on the more provocative side of stuff or using a little bit more of, I'd say that UpToDate more LLM model things. We're seeing [00:08:00] it in our management consulting area for helping companies look at things like do the first cut of reviewing their internal policies and procedures and making recommendations on how to change them from a risk management standpoint or even to help them be in compliance with things like grant requirements. There's a lot of things that used to take a lot of time on just processing and reading through stuff, whether it was lease agreements or partnership agreements. That's some of the stuff that companies [00:08:30] and accounting firms could look at expediting today, and frankly, it gives, I think, associates a chance to upskill a little bit faster because that was stuff that frankly, a lot of times people were reserving that work for the third, fourth, fifth year and up individuals because you're not going to throw a partnership agreement at a new hire necessarily and think you're going to get a great executive summary. So leveraging tools, this kind of spit out that first draft at times and gives accessibility [00:09:00] to people a lot faster, which I think is going to be critical for us to stay competitive
Dan Hood (09:06):
And see, it's interesting when you hear a lot of people talk about ai, but particularly gen AI right now, it frequently comes in, it's a first draft, it's something for somebody to review. It's send that out, you got to look at it first, but it's a great way to start. It'll get you an outline, but not a finished product kind of thing in the sense of this makes perfect sense be Gen AI is still barely over a year old, or TBTI should say specifically is little [00:09:30] over a year old, but that's the one that brought it to all of our attention, and it's still a relatively young field, and all these models are relatively young, so there's a sense of they kind of need to grow up and they need to ingest more and learn more and that they'll just become more powerful. But there's also a sense then that what we can do with AI now is probably very different from what we're going to be able to do even relatively short periods of time ahead, 2, 3, 4 years. When you look ahead, do you see big differences or do you see that changing as it becomes, for instance, as these models become more mature and more knowledgeable?
Kelly Fisher (10:00):
[00:10:00] Yeah, I think though if you just look at the amount of capital that's infused into the AI space, I'd say a lot of the largest organizations in the world with the deepest pockets and arguably some of the smartest minds are investing in this technology. So yeah, I think that it's going to have an exponential kind of growth and capability rate, and that's going to obviously give us a pretty significant opportunity [00:10:30] when we talk about it internally. From a Wipfli standpoint, I think the last thing we should do is think that technology is going to be the limiter. It's usually our own processes and our own people from a change management standpoint. So I think with the investments, the tech, I mean what for us as we look down even from an internal standpoint where we think it's going to make a big difference over the next couple of years, we look at it in a lot of different kind of patterns.
(10:56):
I think it's, hey, the solutions we can bring to our clients is one category, [00:11:00] our own internal processes, procedures, things like that is going to change. And then I think the third is I think it's going to have a significant change in our learning and development of our people simulation scenario-based training, a lot of stuff that all of us had to work through and have to live through it live. These kinds of technologies are actually going to make it a lot safer and a lot more rapid for us to put associates through things like the angry client, the [00:11:30] missed deadline. I don't want to pay the bill or the mock boardroom where you're getting questions fired at you from executives. So that I think is coming a lot faster than people would probably realize.
Dan Hood (11:44):
That's a very cool, I like that. That is a potential use for it. As you say, you can't practice that without getting a client angry. You can't practice that, and you don't want to deliberately go out and get a client angry, and there was been, for years, there was the sort of notion that young [00:12:00] accountants had to do a lot of scut work and do a lot of base work. You have to do 10,000, 10 forties to learn how to do a texture. You really only need to do about 110 forties, but we need to do 9,900 more, so you got to do 10,000. But that ability to train people and to generate, I'm assuming we're thinking about, right, the AI just starts generating scenarios and you just say, how do you react to this? How do you handle it? That's fantastic. That's very cool. Are there things you want AI to be doing for you in five years where you're like, I really want it to take care of this for me. I hate this, [00:12:30] or things where you're like, this is a natural fit. Once it's better, it's going to be a perfect fit for this. I can't wait for it. Is there anything like that where you're like, oh boy, oh man, I have a long
Kelly Fisher (12:40):
Of things that I hate doing that we could talk about?
Dan Hood (12:45):
Well, let's make that the ones that you hate doing that are also perfect fits for ai. Give us one of those.
Kelly Fisher (12:50):
I think from the accounting firm standpoint, I think the broad category of knowledge management is probably the biggest thing from an AI standpoint. If you [00:13:00] look at probably coupling that piece of the, bringing the passion back into the profession and what we deal with from the external factor, but then think about what all of us deal with from where information and knowledge is stored. Sometimes it's in people's head, sometimes it's in your slack or teams chat. Sometimes it's buried in email or your tax software or your audit work paper or in methodology. And that is a big tough challenge for us from a profession standpoint and what [00:13:30] AI is I think in the future going to be able to help do and as all of us kind of change the way we think about things. Even meetings, if we start thinking about something like 60% of people show up to meetings because they're afraid that they're going to miss something, they need to do their job. Think about changing those things into digital assets that can be queried instead of, I got to show up to this thing. So I'm most excited for the kind of knowledge management aspect between what we all have [00:14:00] from a internal, our own way of doing things with what's out there from a public knowledge, information crowdsourcing. That I think is what I am incredibly excited for. From a use case standpoint,
Dan Hood (14:14):
That could be very cool. I mean the potential for that, as you say, particularly, not just internally, I missed this meeting, but somebody had their AI assistant in the meeting taking notes so I can get all the information I needed. But also that AI assistant can then go out in the world and say, you were talking about this. Here's what everybody [00:14:30] else is saying about it kind of thing. That's very cool. There's a lot more we're going to dive into on this, but real quick, we're going to take a break in a second, but real quick before then, hey, I mentioned with these a great big firm, you guys are one of the biggest firms in the country. Do you think the large firms are going to be using AI differently from small firms? We mentioned the investment. Obviously there's going to be large firms just to be able to make bigger investments, certainly currently, but do you think that's a thing that will persist? Are large firms always going to be using [00:15:00] AI differently from smaller firms? Is that going to even out? Is that playing field going to get leveled?
Kelly Fisher (15:04):
I think that this is one of those spaces where it could level out maybe faster than some of the prior technologies. Obviously, the capital deployment and accessibility to information gives the large firms, frankly, the advantage. I think that what the smaller firms could have an advantage on is potentially how nimble they can be on some of the stuff. How fast to really deploy [00:15:30] a lot of the AI solutions at scale. You got to have your infrastructure set, you got to have your process, your procedure, you got to know where your information is. So with the right kind of focus, I think that piece of it could be an easier problem for some of the small firms to tackle. There's also, I mean, there's a very kind of public, we don't know who the winner's going to be from a closed model versus open model from a solution standpoint, which obviously has a huge impact on the cost [00:16:00] of some of these solutions. So we see that race going on and then what you need to do from an internal standpoint. I think that different challenges on both sides, but that's one of the things that from a larger firm standpoint, we always think about not losing the advantage that we have on those kinds of things to a faster, more nimble, small organization.
Dan Hood (16:25):
Right. Yeah, no, I'm not trying to get you to give away your secret sauce though. I'm totally trying to get you to give away the secret [00:16:30] sauce, give away the advantage to all the small firms swore all over you. You got to keep an eye on 'em at all times.
Kelly Fisher (16:35):
Yeah, they're lucky. They have some things. If they deploy the limited capital in a more strategic focus, I think that there's not a reason why they would have to be kind of left behind. It could also just lead to those organizations may be leveraging a little bit more of the AI that's built into applications versus necessarily developing their own solutions.
Dan Hood (16:59):
And that's [00:17:00] probably just one of the bigger things. There may be a gap in time between if you're developing it on your own and putting a lot of money into it, you get it earlier, you get a bespoke, that sort of thing. But once third party vendors start getting in there and start spreading those tools out, they may be a little less bespoke. They may not be as deep as some of the larger firms tools, but they'll be out there and then the nimbleness comes in and that's when you really need to crush them.
Kelly Fisher (17:23):
Yeah, I think there'll be a little bit of an early mover advantage, but with the pace that everything is moving game.
Dan Hood (17:30):
[00:17:30] Alright, I want to talk a lot more about what Wipfli's doing particularly, but we're going to take a quick break before we do that. Alright, and we're back. We're talking with Kelly Fisher, chief practice partner at Wly talking about ai, all things ai, where it's going, who it's going to be good for, who it's going to be bad for. No, it's not really going to be bad for anybody. It's the gist so far of the conversation and there's tremendous potential and going to be more coming down the pi. But I want to dive a little more specifically into what your firm is doing with ai. [00:18:00] What does your AI strategy look like?
Kelly Fisher (18:03):
Yeah. For us, our AI strategy, I would say, I think the first time that I was in the boardroom presenting on AI was in the fall of 22. So we've been at it for a while from a conversation standpoint. We look at it and tackle it. I'd say from a couple different angles, the kind of use case aspect of it, we deploy a top down methodology, a bottom up. So we have an innovation center [00:18:30] where we allow all of our associates to kind of come up with ideas. So we get our use cases from that. We're also incredibly active in looking at some of the capabilities from out of the box or what our key partners from an application standpoint are doing. So very early on we met with leadership of all of our critical partners and applications to make sure we understood their roadmaps from an AI standpoint to help give us [00:19:00] some kind of visibility into where the white space was, where the gaps were, or maybe where something wasn't going to be as fast as we needed it to. So I'd say that's kind of the big picture aspect of the use case. Then we think about the where are we going to innovate a little bit more on the edge. We have taken the approach of we're not going to do that in any of the direct client facing solutions that we spin up because of the risk tolerance and risk mitigation for us. So a lot of the kind of edgier [00:19:30] experiments, I would say are all around internal facing at this point. So it's in helping deliver some of our client service deliverables, but we're not spinning up the AI enabled bot to directly interact with our clients and give advice. That's just not part of our strategy that we're willing to do at this point. So we look at it across the spectrums, what the purpose is and if we're going to build or buy.
Dan Hood (19:59):
Gotcha. [00:20:00] But that's just for now. In a couple of years you may all be handing all your work over to ai, but that's
Kelly Fisher (20:05):
Once the early, again, for us, it's just that we're doing a lot and we're doing it for a lot of client facing things. I mean, one of our tools is specifically around our client experience that we've built. So it's definitely meant to make it better, but we just won't turn it on for them to have the first interaction with it right away.
Dan Hood (20:27):
It is all still relatively new as you say. I mean, [00:20:30] you said 2022 and said, so we've been doing it for a while, and I'm like, well, in AI terms, that is for a while, but it's two years in accounting terms. That's why we've barely vetted it at all. So it's strange how much it's coming just that short period of time, but also relatively young. It is. And given that, I want to talk a little bit about the pitfalls. Talking about, in instance, just the notion of don't try it out on your clients first kind of thing. Are there these things you would warn firms [00:21:00] about as they think about getting into AI or things they should be bearing in mind as they approach these tools and this potential sort of, well, you talked about how it may change your processes, so you may be rewriting everything you do in your firm from ai, but before you do that, think about this for instance.
Kelly Fisher (21:17):
Yeah, I mean obviously especially for a profession that's built on trust and arguably transparency, that is a highly talked about topic for us in our leadership room and in [00:21:30] the boardroom. So I think some of the things that accounting firms have to think about is at least getting your own leadership team board educated to an extent on understanding some of the risks. There's obviously cybersecurity risk with the tools that you're leveraging. Make sure that you're sitting there and have a use policy for your associates help on this. So make sure you at least have the use cases and your [00:22:00] own fair use policies out there. Whenever anybody is using or participating in one of our pilots, there's mandatory ethical AI training they have to go through before they get access to it. I think that the not even understanding what the ethical risks are inherent in these systems is something that could be a big pitfall for firms if they're not addressing that.
(22:23):
The tech debt, you could spend a lot of money and spin up a lot of solutions chasing [00:22:30] some of this. So making sure that you tie it to your strategy and then the regulations huge, right? I mean, the EU advanced legislation just last week, that's going to impact a number of organizations on top of executive orders got signed in the US in the fall. I mean, a lot of countries are sitting there and states for that matter. For us, we have state regulations to tackle also is to just make sure you understand who's using what and [00:23:00] where you're actually deploying it and make sure you're not going to step in it from a regulation standpoint, from a regulatory risk.
Dan Hood (23:09):
Exciting as it is, right? The potential for potential downside. If you mishandle it, it can be pretty significant. And as you say, I think we should all expect a lot more regulation around this as it gets more and more powerful. I know you talked to clients about mean, apart from protecting from your own initiatives or not telling phrase, but I know you're advising clients on AA as well. Can you talk a little bit about how that works [00:23:30] or what's advising them on how you're advising 'em around ai?
Kelly Fisher (23:34):
Yeah, I mean, we started out, Wipfli's approach is definitely one of pragmatism, I would say. So most of the time our clients, their first question is we serve primarily the middle market from an organizational standpoint. So usually their first question is, how far behind am I? Because I think just what they hear and see and what's out there from a news cycle standpoint. So usually it's [00:24:00] just helping them understand they're not nearly as far behind as they think they are on deploying some of the solutions. So that's the first piece of it. And then it's honestly sitting down with their leadership and helping them work through the AI impact on their strategy. I think that just like we talked about earlier with what you see is going to come through some of the applications just embedded in it, is helping our clients understand kind of what makes [00:24:30] them special and different in the market. They can't lose sight of. It may evolve, but at the end of the day, access to a lot of this information and tools is going to, everybody's going to catch up to it at a point. So get what you want from the early mover advantage. Deploy it in the ways that help give you a competitive advantage or help you solve some of your problems, but don't lose sight of your value proposition as an organization, because that's still tremendously [00:25:00] important in thinking about these technologies.
Dan Hood (25:05):
Gotcha. I'm curious, have you ever run across a client where they said, do I need to worry about ai? And you said, no, not yet. Are there any industries you look at and go, Hey, AI is not going to make a big change here for, I mean, I think we all assume it's going to change everything at some point, but are there industries you look at and go, yeah, you don't have to worry for a couple of years, you're good?
Kelly Fisher (25:22):
No,
(25:26):
Just because I think that some of the, there's obviously just [00:25:30] like in all of our own firms, there's different industries that go through the adoption curve. I would say there's some on the front end versus the backend, but I do think a lot of the early use cases for the ai, you can apply it across functional areas fairly quickly, especially with some of those longstanding. So whether it's hiring, recruiting, vetting people, helping your finance department, creating collateral for marketing, a lot of that stuff is [00:26:00] not, what you do is unique from an industry standpoint, but the use cases for functional areas, I think can be deployed quickly, even if the field services aspect is going to be a little bit slower to adopt some of the technology.
Dan Hood (26:16):
Gotcha. Alright, well, so you heard it here. It's inevitable, it's inescapable. Get to work on it right now. No, it's very exciting stuff and I think you've helped us get a clearer picture of what it's Ken and will do for accounting firms. So Kelly Fisher [00:26:30] of Wipfli, thank you so much for joining us.
Kelly Fisher (26:31):
Yeah, thanks for having me excited about this and what it's going to do for the profession, so thanks for letting me talk about it today.
Dan Hood (26:38):
Excellent. Yeah, well, I think we're all a little bit more optimistic. I'm less Skynet and more friendly AI assisted in my thinking now than I was 20 minutes ago, so thank you for that.
Kelly Fisher:
Thanks. I aim to please. Thanks, Dan.
Dan Hood [27:00]
Excellent, and thank you all for listening. This episode of On the Air was produced by Accounting Today with audio production by Adnan Khan, rate or review us on your favorite podcast platform, and see the rest of our content on accounting today.com. Thanks again to our guest and thank you for listening.