AI in Accounting: Speed, Risk and Judgment (Ep. 20)

AI in accounting is changing how technical papers, reports, training content, and audit resources are prepared. But faster does not always mean better.

In this episode of Financial Reporting Conversations, Wayne Basford and Judith Leung discuss how AI in accounting can improve efficiency, support technical research, and help professionals produce more polished work. 

They also unpack the risks: convincing but flawed analysis, hallucinated references, weak application of IFRS principles, and accounting papers that look right until someone tests the reasoning.

🎧 In this episode, you’ll learn:

• Why AI in accounting can speed up technical work
• Why professional judgment still matters under IFRS
• How AI-generated accounting papers can mislead auditors and preparers
• Why facts, contracts, and critical thinking still drive the right answer

Financial Reporting Conversations is brought to you by Basford Consulting helping professionals go beyond compliance and get financial reporting right.

For technical insights, training, and resources that make the unknowns in financial reporting known, visit basfordconsulting.com

🔗 Connect with us:
LinkedIn: Wayne Basford & Judith Leung
YouTube: @BasfordConsulting
Website: basfordconsulting.com

Transcript

SPEAKER_01 0:00

Welcome to Financial Reporting Conversations brought to you by Basic Consulting.

SPEAKER_00 0:04

We're here to make the unknown announcement of accounting, auditing and climate standards known so you can avoid the blindfredding mistakes and do financial reporting better.

SPEAKER_01 0:13

Each episode will impact what the standards really say, what they mean in practice.

SPEAKER_00 0:18

Whether you're a preparer, auditor, director, or litigator, our aim is to help you get it right. Hello and welcome to Financial Reporting Conversations. AI is rapidly changing the way we work. Companies across industries are integrating AIs into their workflows, and we are already seeing large organizations pivot in a big way. For example, Alassian recently announced it was laying off around 10% of its workforce as it shifts more heavily towards AI-enabled operations. And many professional services firms like order firms even now include AI tools as part of the technical resources. People are asking AI technical questions using tools like ChatGPT to draft accounting position papers, write reports, or even help analyze accounting issues. And full disclosure, much of this script is also written by AI. So the question we're exploring today is how would AI change the work for us as technical accounting gigs and our expert witness work? So, Wayne, lately I know you've been experimenting with AI tools quite a bit. How are you using them today?

SPEAKER_01 1:29

I'm using AI a lot, and it is it is actually quite scary what you can use it for. So two aspects I'm using it for are one on training, creating training and the possibility of training is unreal. And the ability to get AI to I've just managed to and I haven't watched it, but I for the sake of the world, I have just cloned myself. So I have just gone and created a AI avatar that is cloned and it's cloned me at 59. Now, if anybody's gone and looked at my LinkedIn photo, I which I think I froze myself at before 40, I think, if you look at that uh LinkedIn profile. It is really weird that I can actually clone myself at an age, and I really wish I'd done it sometime in my 20s or when I'd got more hair. But and then I've actually gone and cloned my voice, which I'm not sure is a good idea or a bad idea, and should have put a much more acceptable home counties accent on it, or a much more acceptable neutral Edinburgh accent, which I I could do. Um and that just creates an amazing possibility that I can now potentially create so much content, not having to record these videos, just get AI to produce text and then get my clone to speak it. That is unreal. And it's gonna be from a our perspective, you know, the the positive things about AI, you're gonna get AI, generative AI, everybody's gonna get fired. There's gonna be no jobs for anybody unless you're a plumber or an electrician. So forget being a chartered accountant, forget being a technical chartered accountant, big picture, five years, ten years, fifteen years. There aren't going to be people involved looking at laptops if you're on that view of the world. The positive side of AI is education is gonna be unreal. It's gonna change the education sector completely. People are gonna get individual training if they're struggling with English, struggling with grammar, struggling with a particular concept. You don't need to have a lecturer that you might not get contact to in a university, in a school. You're just gonna have AI educators, and we are going we we we Baster Consulting are looking at AI and how that's gonna work on technical accounting. From a day-to-day perspective, I'm an accountant, I am not a journalist, I am not a lawyer, writing essays was never my great strength at school, and I can actually use AI to produce a much more polished document quicker. I use AI uh speech to it, so I can chat to ChatGPT or shout at Chat GPT, as I do, because we have our we have our little moments, and I can say my ideas, I need a persuasive argument, tell them debits are on the left, tell them it fails fix for fixed, uh, tell them da-da-da. And it does it so much quicker, and in a much more polished manner. And this is really playing around with training, that I can I know for 30 plus years experience of saying, all right, how do you address the risk? Would you how would you design procedures to vouch the five-step approach in Inifras 15? And if I asked ChatGPT that question, unfortunately, its answer will be significantly better than the answer I'd get from certainly a grad, certainly a senior, and I won't challenge managers or or partners. And it's amazingly powerful. And but where it is at the moment, I can go to it and say, Ah, but I think you've missed this. How do you address that? And then in a very polite way, it was oh, you're right, you're you're spot on, and then and I go recorrect it, and it does it, and it adds that last step or it explains that last step. When I'm bouncing some technical things off it, you see that AI at the moment, people are actually having relationships with AI. The number of marriages that are breaking down because they've suddenly discovered one of these AI tools understands them a lot better than any partner they've ever matched. That there's some people that are delusional that they've they're now convinced that discussing with their AI friend, they've solved world hunger or they've solved something amazingly powerful. And they just get obsessed because they are every day they are talking with their genius AI. That's actually the genius AI is telling them they're the undiscovered genius. Right now, this can happen in accounting. You know, I can ask it a question on this because I'm being idle and but and it comes up with the most convincing argument. I do enjoy chatting to AI because I don't have to be woke, I don't have to be nice, I can shout at Chat GPT and tell it it's got completely wrong, which it is, and it apologizes normally. But it's a dangerous world we're going in.

SPEAKER_00 7:49

Yeah, well, I also like you, I also well, not to maybe I've got less time, but um I also use AI um to you know put together our quizzes for our training program, design actually our webinar flyers was um quite impressed about I did we saved up on our designer to design our flyers and you know write LinkedIn posts and you know being a sort of like a bilingual that like I'm neither good in Chinese or English, like it just helps me polish um like yeah uh position papers and then corrects like those tiny mistakes. So it is definitely a great tool that we can have to sort of uh improve our efficiency and make things, especially us with being a small boutique firm, improve a lot of our resources. And I know you know, um big firms, uh firms also have their own internal AI tools that can answer technical accounting questions. And I guess that's sort of what we want to, you know, talk about is that big question for our profession. If AI can already write our reports, um, you know, we use it to generate training content. Does it replace us? Maybe we need we can retire.

SPEAKER_01 9:10

Yes, it will. Most definitely, AI will replace us, gen AI will it replace us. The just the big debate is when. And sitting here as a 59-year-old, I'm happy I might be able to make it to 70. The power of AI is gonna be just unbelievable. And you know, the thing for us, will we have anybody to actually train? How many accountants are gonna be out there that need our advice? It's gonna be very, very difficult at the moment. You've got everything is facts and circumstances, so at the moment, can you give AI all the facts and circumstances? That at the moment, I think, is its weakness, but you can speed it up. You know, I can get AI to read documents faster for me, search documents faster for me. I last week, clients who come to me and was asking me information, and it was a public announcement. So it's da-da-da. The public announcement says this, the client wants to account for it this way. And annoyingly for me, and this is me being my old dinosaur self, so I'm gonna put this accounting paper together, and I want to be able to cut and paste out of the PDF, out of the announcement, and I can't cut and paste out of the PDF. So I started off by feeding that public announcement PDF into ChatGPT and said, please turn it into a Word document. So effectively, so I could cut and paste it. Which he did like that. And then I said, Okay, tell uh find me the clauses that say this. Tell me what you think this is, and cr and write me a paper, write me a paper on how it works. And he did it. He did it in seconds, he did he did it in two minutes by the time he's thinking and journeying it out, which would have taken me two, three hours. He got the answer wrong, yeah. But it it's speed, you you know, and I I was listening to I listened to this guy who's the diary of the CEO. Remember Chat GPT in particular, or I think a lot of these AI models, they want to give you the answer they think you want. So they're actually programmed and they've learnt to get to know their user. And the classic thing he says, he this guy's um a Man United fan, and he uses Chat GPT a lot, and he asks ChatGPT who's the best Man United player ever. And ChatGPT comes back and gives him the answer that he likes. I think it's messy. He's got a good friend who's also a fanatical Man United fan, and asks the same question, and it comes back with a different answer. And I was actually, I asked this, I asked Chat GPT, and by the way, this is where the the question I'm asking you is whether you're in exploration or whether you're in development. Based on a public announcement that really says they've gone into development, have they switched that switch between IFS 6 and IS-16? And it writes it very convincingly, no, we are still in exploration. And I go to it, but how about this? How about surely when they announced they were going into stage two? That's the point where it's commercial viability. Very good point, comes ChatGPT, and rewrites the argument completely and re-wrote it into a into the correct argument. And then, all right, what is the key decision? Put clearly, this is the clear decision mark that the entity needs to decide when it has gone into commercial, when it's commercially viable. But I was still driving it because if I'd have just let ChatGPT do it, it would have given me the answer it thought I wanted.

SPEAKER_00 13:48

If you're finding this discussion useful, please take a moment to click like, subscribe, and share. It helps others in the financial reporting community discover financial reporting conversations and keeps you up to date with every new episode. For me personally, um, some people have I've reviewed a lot lately, I've been reviewing a lot of papers who looks like it's been written by AI. Um and sometimes the conclusion, it could be, I guess, good for a map from a management perspective when they are trying to arrive at some at a conclusion, but um, but a lot of times the analysis is lacking and it can, if you look at the reasoning behind it, it is flawed. And sometimes when you don't understand, um, I've looked at a paper where they've said how to account for revenue and you know identify the five-step model, step that they've laid it out. They say step one, identify the contract, step two, identify the performance obligation, step three, step four, step five, recognize revenue. But when you do, when someone that doesn't understand the what identifying performance obligation means, then we're picking up terms in the contract that are more legalistic rather than the actual what have I promised a customer. It's yeah, it's all the my obligations are all the precedents before I need, you know, I need to get, um, you know, my obligation is to make sure that I've got this license that I've maintained, I'm licensed all the way. That's not what we are looking for in um revenue recognition. So we still need to understand the standard and what it is. So for me, I'm thinking, yeah, like it doesn't take away our training because we still need to train people to understand what they need to look for.

SPEAKER_01 15:42

You can see the problems that are happening in the courts, that there's so many court cases where AI has been used to generate the brief. And it's all very convincing. And the lawyers look at it and it's got a footnote that cites it to this case. Amazingly convincing, and and far more you know, case law and citing it than there is in the canon of IFRS. And then when they look at it, it's completely wrong. And you know, this Silk's in trouble for this, and certainly the solicitor that prepared him the brief would be absolutely getting killed for it. So at the moment, these accounting papers, very convincing, possibly fill the length, they talk about the five-step approach. There's a real risk it's complete rubbish. And I do, you know, from an audit risk position, it will be, and do you remember this horrible phrase I've got there, auditors just file, they don't audit. I do suspect possibly a junior accountant within the department will go on to ChatGPT and create an accounting paper on this is how we do revenue recognition, this is why uh it's not a share-based payment, this is why it's not a derivative liability, and it will look amazingly convincing. That paper will be submitted to a senior, and that paper will be yeah, yeah, yeah, five-step approach, fix for fix, tick, tick, tick, tick, tick. That paper will be filed, the audit opinion will be issued, and then somebody will work out Chat GPT got it wrong, and the auditor did not check ChatGPT, and because it was such a convincing document. And then somebody like me or you will say, Well, that clause clearly that's linked to this, and it resets and there's a ratchet, and the performance obligation is this, da-da-da, and they're not distinct, and we'll just tear it apart because at the moment some of these some of these accounting papers are are the equivalent of being written by a five-year-old, but in a very convincing style of English. So it's gonna be in for the next as we transition to a world without us, it's gonna be a very interesting time.

SPEAKER_00 18:11

Yeah, I mean, for me, I sort of I don't know where you know generative AI can end up doing the judgment for us because you know, I forests are principles-based standards, and necessarily that means, you know, they want us to, they want professional, they want accountants in applying the standards, exercise professional judgment. And that is, you know, you need to consider all the specific facts and circumstances to a transaction. And sometimes, you know, I guess it's you know, rubbish in, rubbish out. You need to understand the standards and how to apply, and then be able to feed like Chat GPT the right information for them to even sometimes to come up with the right answers.

SPEAKER_01 18:55

To the power of AI and the longevity of the technical accountants, the IFRS canon isn't very big. You know, there's not been that much written. So feeding all the accounting standards into any model, feeding the basis of conclusions, feeding the illustrative examples, feeding all of the IFRIC agenda papers, all the staff papers, it's nothing. You you know, chat chat GPT could it it's it's it's not not even scratching any surface.

SPEAKER_00 19:35

And I think some of the accounting tool and the firms, like their own proprietary ones, is already doing that.

SPEAKER_01 19:43

Think about the way we do things and the way we speak things up. We you you know, we know this, we know the answer. How do we back that argument up? You go and look at the the big firm's books, you've got the AI, we've got the subscriptions on most of them, and you go and search it, and you go and find chapter 14, and you go and find that example. That takes you a few hours. If we then log in and we'll pay another subscription to the firm's AI models, it will find it for us. So their tools, based on their literature, which is great literature, we will save ourselves two, three hours for it. And it may actually be more diligent than me or you, certainly more diligent than me, that it will refer me to another chapter, another example that I've flicked over as I've been desperately in my old speech, you know, looking at some of the old books. So that aspect of it is gonna get so much better, so much more thorough, and the analysis will be great. It's principles. That analysis is gonna be better. And ChatGPT can work with principles very well. Now, remember, I think possibly ChatGPT and the AI models may actually run a run a program that tells me completely where the principles don't work, where this uh standard clearly goes against the framework and where are the inconsistencies. You may get AI will improve standard setting and inconsistencies going forwards. Um where I think you're touching on at the moment, as is our as is constantly ours, and if you ever see the the caveats in all of our advice, based on what you've told us, this is the accounting. Now we have questioned. The preparer, we've questioned the CFO, we've questioned and discussed with the engagement partner or the manager. At the moment, how does our questioning of all those facts and getting this how do you get that knowledge into Chat GPT? At the moment, we read contracts, so you know how do we get on top of it? We read the announcement, we read the contract, we read the side amendment, we need the amended amendment, and we talk to the audit teams, we talk to the partner, we talk to the client, we hear various facts and support for it. And then we make our technical advice.

SPEAKER_00 22:48

Yes, that's right. Like, and sometimes I think it's also like have because when we approach our work, we we do it from an open mind. It's you know, like you said, like a chat GP already had that inbuilt bias in in it. And sometimes when I, you know, um I play around with Chat GPD, also like, you know, seeing what they're saying when I write an accounting paper, and then they come up with an example, and then I say, where is the example in the standard? And then it's like it's paragraph X, and it's actually quoting it. I was like, it's actually not there. And then they will say, You're right, it's not there.

SPEAKER_01 23:25

What's the phrase hallucination? You know, be careful Chat GPT hallucinates. But I'm I'm thinking that's a very generous phrase. I think Chat GPT lies, that it comes up with an argument, and I believe it, and I believe, but I actually want, you know, and remember, sadly for me, I use I arguably use Chat GPT a lot more with auditing than I do with accounting standards. And it's yeah, and or ethics, and it's it says this, and I go, No, it doesn't. And then, oh, thank you for correcting me. You're right, it doesn't. And you know, this is dangerous at the moment. I I am the subject matter expert and I understand the subject better than Chat GPT, but it you know, this hallucination, it's it's lying, but it's very convince lying, which is going to be scary for lawyers, is going to be scary for accountants, auditors. And you can see this problem, you can spot the lies. When you see the problem that's, I won't name them, some of the big four consulting firms have had putting reports out to government government. The problem is it's lied. And the people that, you know, it's gone through the process, but hopefully, before those big four have charged half a million dollars for the report, somebody within the big four read the report they issued. And the people that read it didn't spot the lie. You know, it's dangerous.

SPEAKER_00 24:55

Yeah, so thanks, Wayne. That was um, yeah, good conversation on AI on topic, which is likely to be another tool that will stay. And where for us, you know, it is a great tool to use to help us do our work better um and more efficiently. But um, like I think as this conversation has highlighted, there's still new professional judgment, experience, and critical thinking. At the moment, it's still very important to so um, yeah. So you will certainly, but it would certainly be interesting to see how where we go, this journey continue to develop and how it will continue to shape the way we work um as accountants and advisors, and how our industry adapts and changes to this new technology. So thank you for joining us for this episode of Financial Reporting Conversations. And um, yeah, if you have any thoughts about how AI is influencing your work, we'll love to hear from you. So until next time, thank you.

SPEAKER_01 25:55

Thanks for listening to Financial Reporting Conversations. For guidance on applying accounting and auditing standards or to access our online training programs, please visit batsfordconsulting.com.

SPEAKER_00 26:06

Don't forget to like, subscribe, and share this episode with your colleagues and contacts. We'll see you next time where we make the unknowns in financial reporting known.