AI In Accounting: The Update Episode

AI is everywhere right now, and if you're an accounting firm owner, you're probably somewhere on the spectrum between genuinely excited and quietly terrified, sometimes within the same afternoon.

Aly joins Amy and Will around the campfire for a proper, no-fluff conversation about what AI actually is today and why understanding that matters before you let it anywhere near your clients' tax affairs. Spoiler: generative AI is a probabilistic pattern engine, not a truth machine. Think of it as a fast but overconfident junior who's great at drafting and terrible at knowing what they don't know. Your job is still the review, the verification, and the accountability.

We also dig into agentic AI and automation, and why handing over your logins or letting an agent roam freely across your tech stack is a cybersecurity and privacy conversation you need to have before it becomes an incident report. For firms operating under confidentiality expectations and growing compliance demands, client data security has to come first, full stop.

And for anyone feeling behind? You might not be. Sometimes the tech just isn't ready yet, and holding your ground while you map your processes and cut waste is actually the smart play.

Come for the AI chat, stay for the sanity check.

Subscribe, share this one with an accounting mate who needs to hear it, and leave us a review. 

AAAAA IS PROUDLY BROUGHT TO YOU BY OUR SPONSORS
Practice Protect | Protect your clients’ data and put your security on autopilot with our all-in-one, managed-for-you platform.

ATO Mate | Fast, simple, secure ATO document processing for accountants
and SMSF admin

FOLLOW US
www.accountingadventures.com.au
Accounting Adventures (@accadvpodcast) | Instagram
Accounting Adventures | Facebook
Accounting Adventures | LinkedIn
Aly & Andrews All Aussie Accounting Adventures (@accadvpodcast) | X
podcast@accountingadventures.com.au

CHECK US OUT
ALL IN Advisory | Your squad of award-winning accountants, tax wizards & biz visionaries, perfectly tailored to elevate your biz. Let's soar together!
Illumin8 | Purpose-led cloud-driven accounting humans
Clarity Street | Bringing clarity, time & ease to your practice. Your accounting systems, people and processes working together. Simply. Seamlessly.
Dowling Buckley | Accountants showing biz owners efficient and compliant tax strategies for balancing ownership, investing and lifestyle.

MUSIC
ENTENTE (@ententemusic) | Instagram
PRODUCTION
Dav...

Transcript

Welcome And Guest Allie

SPEAKER_01 0:00

Hey Amy. Hey Will. What's going on?

SPEAKER_02 0:03

Well, uh lots is going on, especially in the world of what we're talking about today.

SPEAKER_01 0:08

Well, I don't want to spoil the surprise because I feel like you've got something special to lead us all into very quickly.

SPEAKER_02 0:16

Only a very special guest.

SPEAKER_01 0:18

A very, very special guest.

SPEAKER_02 0:20

Very, very special guest who is bursting at the seams to talk about the topic today. She was like, oh my god, please, please, please, can please can I be on the please please. No, no, no, we we invited. We were like, yeah, sure, man. Whatever. I crashed this episode, people. So we have Ali joining us.

SPEAKER_04 0:39

Hello, everybody. Well, I'm so excited to be back on the tech episode. And it seems that I seem to crash this um topic every time.

SPEAKER_01 0:47

It's not a crash. You're always welcome. We want your insights. The people love Allie's insights on tech. Correct.

SPEAKER_04 0:55

So we'll have you whenever you can. I love my own insights on this one. We'll roll with it. Do we need to go to some theme tune music? Yeah, let's do it.

SPEAKER_02 1:21

So the topic that we are unpacking today is the ever-evolving on a daily basis. That's how rapidly, almost hourly, minute by minute, second by second almost you could say, is the topic of AI. Yeah. And what it actually is, what it means, what's going on in this world, and we are quite literally going to wing it today in terms of what we're talking about in relation to everything. I did uh recently uh from a clarity street perspective, we did a what is AI. Or actually, no, the topic was AI is still a bit shit. That was actually the topic of the webinar that you've had to come out. And can I just say we had the most amount of people sign up to that webinar that we've ever had. So unpack that. Where does that come from? It's a hot topic. It's a hot, hot topic. Massive scared. Yeah, people are scared. And they are curious, but like they're scared, they're curious, they're excited. There's a raft of different emotions associated with it. And the reason why our topic was called that is because it still is on the day-to-day. Like everybody has an idea of where it's going to go and what it will do, but right now it actually is still a bit shit.

SPEAKER_04 2:29

Yeah.

SPEAKER_02 2:30

Plain and simple.

SPEAKER_04 2:31

It's a yes and a no. And this is, I think, the frustrating thing with AI is sometimes it could be so incredibly brilliant, and you're like, oh my gosh, I can see where this is going to go. And then literally the next second, it's the biggest piece of shit you've ever used, and you're getting frustrated with it. So um it it the dynamic is quite strong. So true. Yeah. And it's kind of like the way that I describe it, I think I've described it from like this the whole time. It's kind of like having a junior who's super, super confident, like in some ways very capable, but in others just absolutely dead set stupid. Um, so the things that you think it should be so good at, it just isn't, and it's frustrating, yeah. And you can't when it goes like that.

SPEAKER_01 3:15

I love that analogy. That's like a junior that is overconfident. It's yeah, overconfident.

SPEAKER_04 3:20

Yeah. And and well, like when you question them on the confidence and where they got the information from, they'll stand by it. Yep. Right. Yeah, yeah, yeah, yeah.

SPEAKER_01 3:28

And we'll put resources at it. We had one today, and we've got like all kinds of different models across all the different platforms that we dabble with to until we get it right. But just quickly, I just chucked it into generic ChatGPT today. I had an FBT question just on the difference between the two the two methods for a vehicle. We've got a client who's buying a Ferrari. Um, advice aside as to whether a bit, not the Bentley this week. Different one. We were we were trying to just basically very quickly like explain the difference between the two um motor vehicle methods, and chat GPT spat out absolute rubbish and missed something.

SPEAKER_04 4:03

No good on legislation.

SPEAKER_02 4:04

I agree. Now, I know the reason for this is because we did a we actually did a bit of research for our webinar, so which I'm going to bring across to the podcast, which is because AI is a probabilistic pattern engine. Yeah, it's creativity where it's where it shines. Correct. So what it does is it brings out, like based on all of the information that you've provided me and all of the information that's out there and the parameters that we think are correct. This is the most probable answer we think is going to be correct. Yeah. And that's what AI actually is, a probabilistic engine.

SPEAKER_01 4:39

At what point does it become 100% probability? And it's okay and safe to use? Does it ever get there?

SPEAKER_04 4:45

Or is it just constantly that is the existential question, and this is where you know, oh man, I'm gonna go all over the place with this episode. But this is where they're saying, you know, software is dead, but software is good noise, it's repetitive, it's rote, right? And this and a lot of accounting jobs, that's what we need repetitive rote process. Whereas AI is creative in its nature, that's why it's so good with copy, so good with advisory potentially, right? Because there is no right and wrong with the creativity, but there is right and wrong in accounting and in legislation and in numbers. And that's where you think that, oh my gosh, I'm gonna be replaced. But there maybe it's the creativity that will be replaced and we'll be left with the numbers and the process.

SPEAKER_02 5:30

Which I get, but I think you still need the like you still need the human beings for the inputs to begin with. So like I think that's the the the kicker there from the creativity. Like I get it, but it does like I get where you're going with that, but you still need to give it something to begin with, which requires the human.

SPEAKER_01 5:47

Yeah, we were we were tinking around with uh an AI bot, robot, um, agents last week, and we were looking at trying to figure out and it's all context, right? We're trying to ground ourselves great. Robotic process automation is where you are basically mimicking a mouse movement on a screen and it's going from grid point A to grid point P B exactly every single time. We were going, all right, well, sometimes the green button that we want it to navigate to is going to move around and be on different places on the different pages. Can you restrict the AI agent to go and find where that green button is and press it? And that's the extent of its creativity. And that's about as far as I'm willing to trust AI at the moment is go find the green button, make sure you're only pressing the green button, don't do anything else. Everything else I've found that it's just like it just makes me uneasy with where it can go.

Agents, Security, And Client Data

SPEAKER_04 6:39

And look, we have to put the caveat that we're this is being recorded in early March 2026. And look, yep, it can go in leaps and bounds, and and over the December, January, February period, there were some really big shifts in AI around coding specifically, and that's where there's a lot of coding jobs kind of um they're worried. Yeah. So you know, what we're talking about today, it might be able to do tomorrow, but right now, without the checks and the processes and the human, it is a very dangerous technology. And there was also this huge proliferation with ClaudeBot or Maltbot or now OpenBot, right? Um, where it people were giving it its login information and it was going in and logging in and taking all this information, and it's like, what are you doing?

SPEAKER_02 7:22

Like we actually have to have facial expressions right now, like I'm about to fall off my perch.

SPEAKER_04 7:28

Why would you do that? I mean but because they were getting so excited about what these um agents were able to do. But if when you consider the security risk, the hacking risk, right? Yeah, like we live in an environment if we're CA, CPA, you know, we're dealing with private information, we're getting AML stuff in coming in, like we don't and are not allowed to do that type of thing. And so, which is why we've relied so heavily on the large tech to take away a lot of the security issues and risk issues, right? We we're de-risking by using them. So by creating our own agents, by by shifting and scraping information from one software to another or through our systems, are we breaching privacy, right? Do we need to update our engagement letters? Like these are the things we actually need to start thinking about and making sure that we've got the right processes and uh systems and operating processes in place so that we're not doing the wrong thing. Because I'm concerned.

SPEAKER_02 8:33

Yes, I am massively concerned. I am massively concerned about how I think I said this on I've said this a number of times. So uh just the level of and I will use this word, sorry, the level of dumb fuckery that is out there in the world and has been for the last six or so years that we've seen with lots of world events that have happened and that are still continuing to happen if anybody's watched the news in the last three months at the beginning of 2026 or two months even. Um and it just it's the world say I too. Yeah, it just scares me that those human beings are feeding this inform, you know, this massive, you know, beast of a tool with all of this information and the outputs of it and hence the probabilistic output, this is why we're getting cracked. This is why it's unreliable, why it's not secure, all of those kinds of things. And you're kind of like, well, what is true? Oh, that's a big question.

SPEAKER_01 9:22

And surely there's also commercial considerations as well. Like I imagine if you're one of these platforms, you want to at some point establish a pyramid of value for how good your um each of your products are, so that your enterprise customers that are paying you hundreds of millions of dollars for subscriptions and access are getting some sort of privileged higher tier or better output than just the generic models that we're all the rest of us are getting on. So, and I've seen no transparency or disclosures around the quality and differences between the models that are readily accessible to everyone. But like, yeah, that's where you get into military grade. I'm sure those military contracts are getting a higher tier or better access. So then it makes it difficult as the average punter to try and navigate and figure out with the limited capacity that we all have to figure out how to get started and how to test it and get comfortable with what it can do.

SPEAKER_04 10:19

Yeah, and I think this is where you need to kind of seriously look at your own skill set and say, if you're not a coder, if you haven't built you know stuff before, is now the time. You can spend some time. I'm not sure it is. Like I like Yeah, I'm not sure it is. And you know, I think we go through these, there's the there's the innovators, the people that develop it, there's the early adopters that go through the beta phase, then there's the majority. I think we're in early adopter phase. So if you have normally an early adopter, maybe don't don't don't yeah, maybe don't try and become one now. Wait for the software that you use to develop the AI tools. Um, and not just the stuff that they're gonna say, oh, it's AI, use it, use it. Like the actually the stuff that uh resolves a pain point, provides efficiency. Like our software providers are all across it. They will develop something. It might not be in the timeline that we want and we might get a bit of FOMO, but decide where you are. Are you an innovator? Are you an early adopter? Are you the majority? Right? That doesn't mean that you can't dabble in it and learn, but be very, very careful with the identic AI that's carrying out and the claw bots that are coming because you need to know what you're doing.

SPEAKER_02 11:25

I want to unpack that one, but what I wanted to unpack is also what is the desperate need for it? I think that's a bigger question, and as accountants, you two can answer this way bigger way more than what I can, which is what is the desperation and what do you think the better output is going to be for your clients if you have AI versus not?

SPEAKER_01 11:44

I haven't seen efficiency in workflow myself, and I'm keen to hear your perspective, Ali, but I haven't seen efficiency in workflow. I have seen efficiency in research, and when you have something that you need to go and learn or you need to skill up on or you get a curly question, the ability to navigate all the different phases of how you need to go and basically take you know go from A to B as quickly as possible and making sure that you are getting the right answer. I have found um a lot of success in getting from A to B very quickly from a research perspective. But from uh productivity and how it's being leveraged in different software platforms right now, I've actually found it to be more challenging given that you have to double check and you have to invest time to implement. I haven't got any returns from an efficiency perspective yet.

Where AI Helps Today

SPEAKER_04 12:36

I'd 100% agree with you. Um I haven't been able to get the efficiency gains either. And you know, I think even in the early days of AI with the Chat GBT, which wasn't wasn't that long ago, but I'm still getting those benefits. So, you know, the tra so where am I getting benefits in all in? The ability to transcribe meetings and um get quick actions and outcomes from that. My favourite the ability to draft client emails in our in our tone um very quickly, the ability to um level me up with my marketing, because I do all over my own marketing and copy and blogs and content, um, the ability to well, even with my advisory, I it's leveled me up with my advisory um in relation to you know pre preparation and preparation delivery, yeah, yeah, yeah. Yeah. So but it hasn't helped me on workflow, it hasn't helped me on the pain points of onboarding or offboarding. But I have seen, I have seen some very techie people who are innovators who have been able to get, for instance, IASs to, you know, pull the ref reports from zero, chuck into a spreadsheet, deal with the exceptions, chuck it into zero. Um, I've seen reviews of um uh accounts, for instance. So I have seen some of that stuff now because I don't use it. I'm I can't say whether it's accurate, whether it creates more time, less time, what are the little things around the sides that you need to do that maybe then create issues? Yeah. Um, but I think you know, if AI comes in and it's agentic and it can work like an employee, I think we have to rethink the way that we're doing things. We're not just going to then take our process and chuck AI across it. It's actually going to be a complete rethink of the process itself and how we obtain that information and how we communicate with clients. So I think, and all of the ways that it's helped me is once again through the creative side. But I haven't been able to see any benefits of the AI being chucked into products at this point in time, giving me much return at all. Um, but that's not to say that it won't.

SPEAKER_01 14:43

What a lot of what you um were just mentioning, Ali, sort of triggered me to think that actually a lot of the value that I have got out of undertaking similar processes, like you were talking about onboarding and offboarding and all the pain points and friction points that exist and capturing meeting minutes, action items, and things like that. These are actually a lot of the value maybe that I'm getting right now is in the forced exercise of having to apply a lean methodology approach across the firm's existing processes. And it's the efficient the the main efficiency in value add right now is being forced to actually look at existing processes and mentally prepare for all right, how is this going to be evolve and improve when systems and platforms are able to come and alleviate some of the challenges that do exist? It's not there yet, but actually I'm looking at our processes through a way more leaner mindset, and I'm choosing not to adopt certain technology yet and do things maybe a little bit more manually and old school because I know that if I overinvest in technology now in certain areas, that will prevent us from being able to leverage technology and automation in the future when it does come out.

SPEAKER_04 15:52

And is that a mindset though? Like can I so the first question is can AI do this, test it? And if it can't, okay, we go back to the old.

SPEAKER_02 16:00

Yeah, but I'm gonna challenge you on that one, Will, only because I know that you are a new fledgling um Oh god, I'm getting challenged. Oh no, we've been called to the principles of the city. Um it's more around, but are you doing that anyway because you're going, oh my god, we've been in business just over 12 months and shit, our processes are a bit crap and that kind of thing. Or are you actually genuinely going that plus hang on a second, there is an AI lens that I need to throw on this as well? Do you know what I mean?

SPEAKER_01 16:26

Yeah, so the I'll give a very recent example. So, I mean, we're constantly writing notes about our end-to-end process from start to finish and all the different um mini steps that go into it. Uh and so we're like writing it out and documenting it, and we're looking at each of the steps, and we're going, which of these steps can be automated today? Human automate now, later. Yeah, exactly. And so we're talking about in every step, and so we're just getting good at documenting the steps, and now we're going, well, actually, what of these steps can be removed and eliminated right now? Forget technology coming through that'll be able to automate it or if it's already there. Is it even necessary to begin with? So we're almost just taking like a consulting approach to lean methodology, lean six sigma, whatever the the um the approach is, but like what what excess process and waste exists in our current process that we can start to eliminate now that is ideally gonna hold us in better stead for when we do want to evolve our our workflow and adopt more technology in the future. So, yes, it is like because we're early days, but it I'm definitely finding it is a shift in mindset of looking at the process through the lens of we know things are gonna change in the next 12, 24, 36 months, therefore we're we're not hesitating to solve problems today, but we're forcing ourselves to be even more lean and minimal in what we're doing and how we set ourselves up rather than oh yeah, we'll just chuck that app on and it'll solve our problem today. Actually, can we redesign our process because we know that there might be some efficiencies in the future, but we're happy to just lean it up now and do some small things manually for now, but overall we're not over investing and over processing or over-engineering our process. So, yes, take your challenge. I think it is because we're new. A little column A, little column B. Little column A, little column B.

SPEAKER_04 18:17

I might say I'm eight years in and I've taken the opportunity to kind of look at every system and process and see what AI could potentially deliver. And look, in most places I've been wildly disappointed, I'm not gonna lie.

SPEAKER_01 18:29

But in others, I've just your processes are so good, Allie.

SPEAKER_04 18:32

Yeah, well, do you know what? So underwhelming.

SPEAKER_01 18:36

Pat yourself on the back. Go on.

SPEAKER_04 18:39

I look, I I and that's I think the challenge is that there are times that I'm just blown away. Like, you know, I was trying to automate some of my marketing um calendar the other day, and the first 80% was just brilliant. I'm like, oh my gosh, this is I I can't believe I've been doing it this way the whole time. And then I couldn't even, I couldn't finish that last 20% and it drove me batty. And I was like, oh but that's what I'm funny with AI. 80%. It does the heavy lifting, but then you then have to take over and it has its limitations, but that's not to say that tomorrow it might not be different. But I definitely think I'm having a lot of conversations with other people in the industry around how can I implement AI into my business? Um, what can I do? What are others doing? And the other really interesting thing, and I'm taking this approach, and I've heard others are too, I'm not wanting to change or do anything with my current tech stack just in case I can eliminate that later. And so I have I have a real um stagnant point of view with my software, yeah. Because I don't know what they're gonna implement. And there have been one or two pain points in me that I really want to fix, and I'm like, I just don't think I can do it in the next six to twelve months just in case.

SPEAKER_02 19:49

Yeah just in case. I think that's a that's a really good spot to uh pause and maybe go and hear from our wonderful, wonderful sponsors. This episode is brought to you by Business Automation Works, the team behind ATO Mate and Omble.

SPEAKER_01 20:08

Now you likely already know ATO Mate as that solution that has completely transformed our practices, manage ATO correspondence, reducing processing time by sometimes up to 90%. Yeah, it's awesome. But did you know that the team has also built a new solution called Ombul? It is designed to give your practice total visibility by consolidating all the real-time data from the ATO into one powerful dashboard. If you've mastered your documents with ATO Mate, then it is time to see how Ombull is going to help you stay proactive with client lodgements, payment plans, and any outstanding account balances and debts. So if you are not using ATO Mate, then get on board. And if you already are, then take it to the next level and take a look at Omble.

SPEAKER_00 21:05

Ellie, uh, if you have been paying attention to any kind of news lately, you would have seen that there are some significant cybersecurity challenges that go on in the world around us. Not for big guys, but for little guys and everybody in between. Whether it's hackers trying to get access to our data, where they're trying to change passwords, get cash, whatever it is, it's it is real, it is here, and it's something that as accountants we really need to be aware of, right?

SPEAKER_04 21:29

100%, and that's by it all in. We have Practice Protect. It is the foundation on which we stand in our cloud solutions. It offers superior cybersecurity for the modern accounting firm, right? Which is us, right?

SPEAKER_03 21:42

Superior.

SPEAKER_04 21:43

It tells access to client data, you can manage your risk, you can stay compliant with an all in one platform. And a little birdie tells me that some insurers, if you use this product, will give you a discount on insurance. Now, if that's not a reason to use it, I don't know what it is. What do you think, Andrew?

SPEAKER_00 21:58

I think that's incredible. And I Believe they have over 6,000 cloud integrations and counting. So whatever platform you got, they probably can protect it.

Adoption Strategy And Hype Cycle

SPEAKER_02 22:12

Hey, um, I feel like what we were just talking about, the potential pause on tech and changes to tech based on what AI is capable of right now is a really good place to start. Um, Hallie, you just left that in a really good way. Will, what we Will was like chomping at the bit to go, go. No, I wasn't. I was I just keen for the conversation to continue. But we had to hear from our wonderful sponsors because we did them.

SPEAKER_04 22:41

Without them, you know, and they're resolving pain points today. Literally.

SPEAKER_02 22:46

I also want to talk about some use cases, some AI use cases that we came up with, uh, which is probably for uh right now rather than a futuristic. Anyway, have I lost your train of thought yet for you, Will? No, I was just agreeing with Ali.

SPEAKER_01 22:58

Like, but is that a risk? Like, are we are we at risk of falling into the trap of tricking ourselves into not doing anything and just punting the decision down the road when maybe is that a little bit of bias creeping through? And are we being a little bit slack by not actually confronting it and trying to go one step further in the decision-making process? And like, what are some additional criteria that we perhaps all could benefit from in making these tech decisions rather than just I'm actually not gonna do anything for the next six or twelve months? I'm just gonna wait and see. Like maybe that works for some firms, but maybe for others, you're just perpetuating some bad processes or some software that's maybe not doing things or it's actually holding you back from having a practice that is doing more or getting to where you want it to go. So yeah, this thing don't do anything.

SPEAKER_04 23:49

Is that really I get your point, but I have a good idea? For those that are starting out for those that are starting out on the journey, don't pause. I like I think you know, you've you've got to take a step somewhere. For those that are established and a long way down that pathway, and we've done like all in, for instance, has done every integration you could possibly imagine. We've automated whatever we could, we've built those systems and processes. If I can see that there's a system that I'm not particularly happy with, you know, they haven't really had advancements, it's quite expensive, but I can kind of see that AI is kind of getting there. I'm like, I think I'm gonna hold, you know, like there was a really big decision where we were gonna we were gonna flip over something pretty major in our firm, which would have cost time and money, and we're like, with some of the IS stuff I'm seeing around, yeah, I reckon I can hold for a little bit longer on that rather than going through the pain of a full transfer for not that much. Solving every pain point. Does that make sense?

SPEAKER_02 24:46

So to counter that with what we do um from a clarity perspective, we say basically, first up, it needs to solve 80% of your problem. Like one piece of tech is never ever going to solve every single thing. So if it takes 80% of the boxes, good, go with it. Don't implement something just purely for like a small handful of clients. That's the other thing. Like it actually has to service a whole large proportion of client work that you're actually doing. But if I could give a slight different spin on this, we have had many, many, many, many firms still getting many firms that are have been on the bandwagon waiting for myop or waiting for APS. Yeah, okay. So and this they still exist. So you're right. Yeah, and I am aware. And there is a lot to be said for those firms because there is an element of, well, if it ain't broke, don't fix it. And I'm not against this, right? I think there is definitely efficiency gains, but there is also definitely um some comfort in knowledge of how the systems, even though they feel broken, they still know them. So they know the pain points, they know how to fix them, and it's just easier and simpler to throw a human being at the problem rather than overhauling their entire tech systems. My analogy is this though, for all of those firms that have waited and waited and waited and waited and waited, in the 17 years since I've been doing this throughout various different roles and things like that, I've only ever come across one firm that has changed to the world of zero from I think it was my ob or something, and they changed the world of zero and then have since changed back. Wow. One firm that says it all, doesn't it? So I know I know of a couple that are like, why did I actually change this? Is a bit shit, blah, blah, blah. But they're not like, okay, I'm gonna change back because I hate it that much because it's that crap. Do you know what I mean? So that obviously means, in my opinion, that you can wait and you can wait and you can wait. And I think that this is definitely a different topic to a degree, but also how long do you wait for?

SPEAKER_04 26:40

Is that type of firm going to go from zero to hero though? Like they're not gonna transform their whole businesses with AI if they haven't even. Of course they're not the comfort and going to cloud.

SPEAKER_02 26:49

Of course they're not. But I think but the to your point around the should we wait, and then Will saying, but you know, will we be in jeopardy of waiting too long, those kinds of things? I think at the end of the day, it it comes down to that whole pain point situation. You will change when the pain is greater than not. You know what I mean? Like you'll just be like, right, totally, I'm over it. I want to change. And the holding off and waiting is not necessarily a bad thing. I mean, ultimately, if you're like, I'm gonna reassess it in six months, or I'm gonna reassess it in 12 months, that obviously means that the pain of the problem right now is actually not that great. Or it's not that great that you really want to change it. You can still deal with it for as long as what you need to.

SPEAKER_04 27:26

And I want to slightly go on a slightly different vein here. Um, people are throwing trillions of dollars into AI.

SPEAKER_02 27:36

Yes.

SPEAKER_04 27:36

And the returns aren't quite there yet. So there's impetus for them to make it sound like it can be and do more than it can. And so I think the challenge here is to try it for yourself and see if it works for you and see where the benefits are for you. But also understand there's a massive marketing machine with maybe six people at the top waiting wanting to transform everything and wanting your money.

SPEAKER_02 28:00

And it has to work for you, yeah, and that's where you have to try before you buy. Um, my little thing, one of my takeaways was don't listen to the AI AI hype influences and just follow blindly. Like, do your own due diligence, look into it, and also don't just get swept away in the FOMO of it.

SPEAKER_01 28:17

But that's been the case with, and there's like models, it's Gartner's got the model. I think James Bergen did a great talk from Zero did a great talk on the Xerocon. It's the trough of disillusionment. Is that what it is? Yeah, yeah, yeah, yeah. And it's like, yeah, there's just so much hype, hype, hype, hype, hype, but then it is a pretty, and I think it's been tried and tested over plenty of tech trends. Gartner does this model and they plot them all. And I think everyone that's gone and done the AI hype cycle where it sits with Gartner against the Gartner hype cycle, it's like it's very much in hype phase, and we haven't gone through the trough of disillusionment yet. So it's either gonna be the trend that bucks all trends of from a technology perspective, or we are gonna go through a trough of disillusionment, and then we're gonna actually find people that have commercially cut through ROI positive use cases where it can be deployed efficiently and effectively, not just a majority.

SPEAKER_02 29:09

Or it's not gonna matter because World War III is like quite legitimately gonna hit, and we're all gonna be it's gonna be not a thing anyway, so who knows? Yeah, I know. Just the current economic climate.

SPEAKER_04 29:20

I think back to the cloud conversations, you know, in the early 2000s, right? And the and the hype cycle that was there, and then the the you know, the blockchain mongering.

SPEAKER_02 29:31

When blockchain was gonna be the next best thing.

SPEAKER_01 29:34

NFTs.

SPEAKER_04 29:35

Look at the NFTs and I wasn't on the blockchain or the NFTs, but the moment I saw cloud, I saw how it could work. I do agree. And so I was on the cloud train. Yeah, I am on the AI train. I and I think it's similar to cloud. I think it's the next evolution, and it's not quite there yet. We're doing it very early, but it may deliver it quicker than what we've got, than what when what the cloud stuff was. So I do actually think that it's got validity, yeah, but we have to realise where we are on that cycle, and it might go quicker than what we expect, and it might be bigger than what we expect, but there is definite benefit once again. You have to try before you buy where is the benefit for you at this point. Um, so I definitely think it's that next iteration. We've definitely because it's I don't think it is the NFT blockchain, you know, um crypto. It's not that.

SPEAKER_02 30:25

I think the reason for that though is because this is like AI, like the WWW, and then like cloud, this is accessible for everyone. Yes. And I think that's the reason why it's gonna boom, is because everybody has the ability to have be impacted by it or make an impact with it.

SPEAKER_04 30:44

And I think if we look back at the internet, that was around for quite some time before it became um available to the masses through the search engines, right? Through the Googles of the world. So, you know, I I think this is the early iteration of the internet, but we don't have the tools for mass production yet. But that's not to say that they're not around the corner. And I do kind of see that coming out with the qualification of things, so with the bots and the um agents. So it's I can see it.

SPEAKER_02 31:11

Something else that people aren't talking about, though, in relation to this, is the cost of actually housing the data warehouses. Oh, I have seen some very interesting topics around that one, and just the severe amount of money that it is actually going to cost them. The energy consumption and how they have to be replenished every five years. So apparently it costs stupid amounts just to create them. They need to be replaced every five years. And it was something like Microsoft, Google, Apple combined don't even cover a percentage of what the cost is to build one of them in terms of what their revenue is. So it was actually covered.

SPEAKER_04 31:48

And that's when we go back to the return on investment, right? Are we in the hype cycle, right? Uh until this thing actually starts to do some return on investment, are we are we putting too much capital into it?

SPEAKER_01 31:59

Maybe. Just on the energy one, like I looked into this when it when um electric vehicles came out, and I know that this is also a thing with semiconductor chips as well, but like don't we can't underestimate the speed of innovation at which energy and resourcing is going to be able to find efficiencies through what that industry is going to do as well. So it's not a case of this is like I used to live in Canada, so it's an ice hockey thing, but don't skate to where the puck is, skate to where the puck's going to be. And there's also going to be an innovation trend of energy efficiency when it comes to AI resourcing and what's powering that. I don't think we can assume that it's gonna stay where it is today. Like there'll be industries that innovate in that we'll put up Elon Musk, we'll send stuff up to space, there'll be data warehouses.

SPEAKER_02 32:49

Yeah, we'll just we'll just dump all those electric cars that are sitting in China in car parks into space and you know, and all of the wind turbines that are no longer viable, that have been powered by diesel, you know, we'll we'll dump them up in space as well on Mars, maybe. I think it's a great idea, Will. I love it. Because if you do you know what I mean? Like you said, if we're talking about the Sorry, we're all over like if we're gonna talk about what happens with like energy and data sources and things like that. I agree that there is an evolution, and yes, in theory, it should make things better. But equally, the fallout of that is that there is also a lot of waste. Heaps. And that's probably where that was more my point that was going to that. Um, that little thing that I was just saying, a single, so a single one gigawatt AI data center costs 80 billion. Big tech plans to build a hundred of them. That's eight trillion in infrastructure, more than the entire semiconductor industry has earned in its lifetime. To finance that, you'd need 800 billion in annual profit just to cover interest. No company on earth makes that, not even the thrill tri the three trillion giants. Um, yeah, and AI data centers only last five years before hardware becomes obsolete.

SPEAKER_04 34:00

That's crazy.

SPEAKER_01 34:01

It's great, great. Starting point as at today. Yeah, well relatively new industry. Sure. Processing power is like the demand for it is exponential, like it's exploded. Understandable that hardware is gonna take a little bit of time, there's gonna be a lag to catch up. But yes, hugely like right now, the cost to Earth of this hype cycle is probably quite significant and maybe something that a lot of us are underestimating, and there will be flow on cost of the city. 100%.

SPEAKER_02 34:32

Uh the water? No one's talked about the amount of water that is required to cool these data centres, and they're just yeah, there's just so many.

SPEAKER_04 34:39

And I think you know, when we talk about it, that's that's part of the conversation, isn't it? You know, in even regulation, um, you know, that's another part of the conversation, but I feel like the tech is far ahead of anyone.

Safe Use Cases And Hard No’s

SPEAKER_02 34:51

Ali reining us in, Will she's definitely raining us in. No, it's great. Like I agree, it is. We could go down rabbit holes for days. No, I agree with you, Ali. And there's I you know, from that, in terms of some of the use cases, something that we came up with in our little webinar, I figured we we might sort of end on this topic potentially. Is some of the safe things that it's really good for? The you shouldn't, but you could, and then maybe the unsafes. So um safe formatting documents, rewriting emails for tone clarity, like you've mentioned, summarising internal meetings, drafting marketing content, website copy, maybe job ads, training outlines, administrative checklists, non-client-specific spreadsheet formulas, agenda structures for meetings, um, and helping curate the content for this web um for this episode and the webinar that we did as well. Anyway, um we didn't use it for that. Of course not. Um you probably shouldn't, but we could. To create this list. To create this list, yes. Um we probably shouldn't, but we could. Uh internal partner discussion prompts, idea generation for planning um angles, and then devil's advocate questions, report reviews, those kinds of things. Devil's advocate. And then it's great for that. Yeah, and then the unsafes, drafting tax advice, applying legislation to client facts, uh interpreting law for decision making, technical memos, scenario modeling influencing uh client outcomes, preparing BASS or tax returns, financial statement preparation, audit work papers, summary of ATO rulings, client-specific financial modelling and app development. That's the app development is a bit of a maybe, maybe not. I don't know. That one, I reckon that's probably in the middle bracket. But do you agree with any, or sorry, do you disagree with any of them?

SPEAKER_01 36:34

Letters of advice, um, so drafting advice, I'd put into two buckets, not blindly writing advice, but um as a research assistant or research tool where you're going and verifying sources and then doing your own click-throughs. Like I will quite often very quickly be like, oh yeah, which legislative reference do I need to get to for um show me the like five or six Div7A references, quote them, give me the links, and then I'll go click through and then I'll find my through the legisl my way through the legislation from there rather than trying to go, all right, go to Oslink, go open up Income Tax Assessment Act of 90 of 36 and then go and find it manually. It's kind of like shortcuts that way. Um, and then I sometimes often ask it to challenge me on my thinking. So do that. I do that, but yeah, chucking client details, specific situations and context, absolutely not. Have no, have no. Yeah, so I use I I use it as like a research assistant, and I'm comfortable and fine with that, but I know that ultimately I need to be, I'm signing off on the advice, I need to go and double check all the references, and I need to get good at spotting and working with AI, understanding where it has its failure points or where it could be leading me um down the wrong path. So that's another skill set in itself. But other than that, yeah, I'm cool with it.

SPEAKER_04 37:49

Allie? Yeah, same. Um I it's kind of where I'm at with it at the moment as well, although I am trying to challenge the status quo and and trying to push hard on how I can um use it, but um, that's where I've got it so far too. And I am wildly concerned about the risk. Um excited about the potential about the risk, excited about the potential. Absolutely excited about the potential, and I'm excited about potentially how our software can evolve to um use it, um, to be honest. Uh I think that's where we're going to get a lot of our gains in industry. Um, because they know their product well and they know you know where the pain points are. So hopefully they'll be able to make their products more efficient and and and build out um faster um get better products and new innovations.

SPEAKER_01 38:40

I do think I agree. I think that software vendors that have it is their job to solve problems for industries and they're incentivised to do it, that's where I'm sort of banking on the most amount of progress. I don't think that I'm gonna develop a robot using OpenClaw that is gonna um have huge efficiency gains and do exactly what I need it to do. Like it'd be nice for me to figure something out like that that saves me some time, but I'm sort of relying and trusting on the fact that there are entire industries within software that focus on the accounting workflow that will be able to come through, having invested their own resources and capacity to solve some of these problems for the rest of us. And I'm okay working with software vendors and investing in software when those products do come to life. So I'm expecting that that trend will come through over the next 12 to 24 months, but I'm just not really seeing it yet today.

SPEAKER_02 39:32

Neither. And I think to answer the question around, you know, will AI replace humans, etc. Well, I still think the answer to that is no. However, there was a question that I saw on a recent little interview that asked AI, and apparently there is a um AI tells us that I can't remember the specific date, but it's in 2046. Apparently we're all gonna be dead. Humans will no longer be dead. 2046, apparently. I do the same thing. Obviously, there's so much true. It was very specific. It was very specific. No, it actually gave the date. I think it was like March 7th, 2046 or something. I was like, what? How do you ever what? What if it anyway? Um on that note, final words to wrap up.

Closing Thoughts And How To Learn

SPEAKER_04 40:13

Um give it a go. Yeah. Try before you buy. Um don't don't think you're behind. Um, don't get the FOMO. Um, always put risk and the security of client data at the forefront. Um and you'll be just fine.

SPEAKER_01 40:29

Yeah. I would say watch videos. Like if you get 15 minutes, you're gonna be frustrated if you try and figure it out for yourself. So don't underestimate investing some time up front learning from people that specialise in doing DIY videos and stuff on YouTube that can step you through certain things so that you are building up core capability and skills in how to navigate rather than try and tinker with a specific piece of software or something like that, because you can probably get frustrated really quickly when it doesn't work out exactly the way you want it to. Um, so yeah, I would say invest some time in upskilling from people that are putting content out there on how to build up that core capability and skill set for how to use some of these tools and platforms efficiently.

SPEAKER_02 41:12

Uh mine is you're not behind, the tech just isn't ready yet. So, you know, AI is amazing with so much possibility, but currently it's still just a bit shit.

SPEAKER_01 41:25

I'm so glad you went last.

unknown 41:29

All right.

SPEAKER_03 41:30

Thanks for joining Allie guys.

SPEAKER_04 41:37

Hey, hey team, it's Allie from All Aussie Accounting Adventures. Thanks for joining us for this tech edition and joining Amy and Will. We hope you've picked up a few great insights and ideas. If you'd like to keep following the adventure, make sure you check out our website and find us all across the socials. We are everywhere, you know us. Just search All Aussie Accounting Adventures. Whether it's conferences, the website, the podcast, social media, there's always something happening with us, so make sure to follow. And don't forget to share the episode with your accounting mates. We love that stuff. So thanks for listening, and we'll see you on the next adventure.