The CA ANZ AI Fluency Handbook, or guidebook, which I do recommend you have a look when we circulate that after this webinar, defines AI fluency as ‘the ability to use AI ethically and effectively’. So it’s about harnessing AI to improve processes and outputs in our daily work, and really to support that data-driven decision making, which is really important in terms of providing advice and high-quality work to our customers and clients.

– Cameron Anderson, FCA & General Manager at XBert

In this episode, we explore AI fluency for accountants and bookkeepers and what it really means in practice. From governance to prompting frameworks, we unpack practical steps firms can take today.

We are referencing the Chartered Accountants Australia and New Zealand AI Fluency Playbook https://www.charteredaccountantsanz.com/news-and-analysis/news/ca-anz-launches-ai-fluency-playbook

Key Discussion Points

• • What AI fluency means for modern bookkeeping firms
  • Traditional AI vs Generative AI explained simply
  • The AI and human partnership model
  • Governance, ethics, and client disclosure considerations
  • The RTRI prompting framework: Role, Task, Requirements, Instructions
  • AI use cases across client communication, reporting, and operations
  • Using AI for financial analysis and contextual insights
  • Capacity planning and workflow intelligence
  • Starting small and building sustainable capability
  • Why data quality underpins every AI initiative

Key Takeaways

• AI should augment human judgement, not replace it
• Start with low-risk, easy-to-validate use cases
• Build AI capability alongside human skills
• Transparency and governance matter
• Small, consistent learning builds long-term fluency

Apps & Tools Mentioned

XBert, Xero, ChatGPT, Gamma, Otter.ai, Fireflies.ai, Gmail, Microsoft Outlook

Scroll down for a full transcript (This transcript has been generated and transcribed by AI)

Heather Smith

So welcome everyone. Thank you so much for joining us. I’m joined by Cameron Anderson, General Manager, expert, and we should be talking about AI fluency for accountants and bookkeepers. So by attending this session, I’m just going to just quickly run through some of the things you’re going to gain from it, you’re going to gain some quick win AI use cases that going to save you hours and boost your data accuracy. You’re going to get an understanding what AI fluency means for the modern bookkeeping firms learn some simple prompting frameworks to get consistent and reliable outputs. We’re going to talk about practical governance and what that looks and feels like, and we’re going to discover how expert can help your team combine AI efficiency with human oversight.

So I encourage you, I encourage you, to ask any questions you have in the chat area, which we’ll monitor through the session, which we welcome now.

Can I start off, Cam, by asking you to share with everyone who’s joined us today, who you are, your expertise, and what you’re bringing to the session today?

Cameron Anderson

Yeah, thanks, Heather. And welcome everybody that’s on the call. Thank you so much for taking the time to join us. So I’m Cam Anderson, the Chartered Accountant and General Manager here at XBert. So I lead our sales, marketing and customer success teams across the business, and I’m very excited to be joined by Heather today to take you through some kind of practical frameworks that hopefully you can apply straight away into your business and take value away pretty quickly. Fantastic.

Heather Smith

Thank you so much Cam. Now, for those of you who don’t know me, and I do recognise a lot of names, but I’m a chartered accountant as well. Like Cam, I’m also a certified bookkeeper. And I spent many years at certified bookkeeping, bookkeeping events, professional bookkeeping events, and running bookkeeper meetings. I’ve run a management accounting practice for close to two decades now. I specialise in navigating the business or the accounting app ecosystem, and I produce an accounting Apps newsletter and then the counting apps podcast, which I encourage you to subscribe to wherever you listen to your podcasts.

So that’s enough about me. Let’s move to the next slide, and let’s talk about what we’re thrilled to see you all. I think it’s the next slide, isn’t that? Oh, we’re going to talk about the resource that we’re using for presentation today.

Cameron Anderson

Yeah, I think so, yeah.

Heather Smith

So we are using a resource that was produced by the Chartered Accountants, which is called the AI Fluency playbook, and we’ve tapped into it to try and give us some structure to this talk, and we’ll provide the link to it on the follow up newsletter that you’re going to get.

Now, what I like to do start the foundation of this conversation by explaining what artificial intelligence is, just in very simple terms. And if you’ve heard me talk about it before, you will have heard this. AI is data combined with algorithms. And as bookkeepers, we are the custodians of vast quantities of small business data, so we are so well placed to take advantage of all that’s coming down the AI pipeline, because we know where that data is, we understand that data, and we have access to that data.

Would you like to take it a bit further, Cam, and talk about what AI fluency means?

Yeah.

Cameron Anderson

Thanks, Heather. So the ca ai fluency handbook, or guidebook, which I do recommend you have a look when we circulate that after this webinar, defines AI fluency as the ability to use AI ethically and effectively. So it’s about harnessing AI to improve processes and outputs in our daily work, and really to support that data driven decision making, which is really important in terms of providing advice and high quality work to our customers and clients.

Heather Smith

Absolutely, I’m very sort of one of my mantras I’m always talking about is data driven, data informed decision making that we can now make quick and we can make courageous decisions using the data that we can surface for our clients.

Did you pull out their traditional and generative AI, Cam?

Cameron Anderson

I did. So if we think of kind of traditional AI and kind of what we’ve been accustomed to until kind of the most recent couple of. Is, is really prediction algorithms. So they’re largely rule based. So if x happens, do Y or Do this, do that generative AI is really a fundamental shift in the way we work. So we’re no longer limited by rules and algorithms. It’s really open ended. So people, you know, I think, largely have been conditioned to use Gen I, like Gen AI, sorry, like a search engine, so giving less context short words in order to get the best kind of narrow their search results back, whereas Gen a Gen AI, is far more conversational, so we’re actually better off to interact with it like a human. We’ll kind of give a few tips on that as we go through

Heather Smith

absolutely thank you for that, Cam, if we can move over to the next slide. So we’re going to talk about how AI Artificial Intelligence augments, but but does not actually replace the humans, and we really should see it as a partnership. So we’ve got here, sort of it enhances our human judgement, rather than substituting it.

You, many of you will have heard, heard the term copilot. I like the term strategic enabler, how it gives us a base or a foundation, but we don’t have to accept it. We can judge it. We can we can look at it, check it out, see if it’s suitable or not. So for quite some time now, as we write, our sentences are finished off for us. Sometimes they’re perfectly acceptable, sometimes they’re non nonsensical, and we have to review them.

The other thing that many of us are using chat GPT, and we might go create this LinkedIn post for me, but we shouldn’t copy and paste it. We should use it as a base, as a base or a foundation, to maybe unpick what our thoughts were about the topic, or maybe draw it, draw out and sort of give us a framework to talk about the topic.

It is kind of like working with a smart intern. So when you’re working with someone who’s smart but but sort of at the intern level, it doesn’t have the experience the nuances that you need to work with. You need to constantly give them feedback and to train them. You can’t expect an intern to work in isolation.

So for example, one of the popular tools that many people will be using are transcription meeting note takers, so a tool that arrives at your meeting, transcribes the meeting, surfaces activities from the meeting, surfaces insights from the meeting, such as, how long did one speaker speak? How long did the other speaker seek? Some of the tools, such as fireflies, will actually surface the manner in which you spoke, was it professional? Was it neutral? Was it kind? And it can give you feedback on that.

But what will happen, and many of you, like anyone who’s not had this happen is, if you are talking about, I’m using a software called Xero, and in Xero, it does this, and it spells zero with a Z constantly. So it does really well, but it still needs oversight. It still messes up US and UK spelling. I don’t understand why it doesn’t know how to spell my best friend Tanya’s name. It always spells it incorrectly. Why? Why is that the case?

But I’d like you to drop into the chat where you’re using AI, Oh, okay. Well, one better look at Cam. You take it over. You describe the poll for us.

Cameron Anderson

Pat, just put a quick poll in here. What are you using AI for today? And it’s very broad, so financial analysis, data checks, capacity planning, not yet using it, or content creation. So we’ve got kind of sitting at 50% of not yet. And then the second highest selection is content creation Heather. So I guess, in terms of kind of our exposure to Gen AI today, and predominantly, probably ChaGPT and the likes that would be, where we’re getting our content creation from, I imagine,

Heather Smith

yeah, absolutely, absolutely, goodness, quite a number of people. We’ve had 60% of people respond, but we do encourage everyone to just have a quick jump in there and respond. How are you using it for data checks, financial analysis? Not yet.

Using it, but quite a number of people still in the not yet using it does. Can everyone see the responses now? Can Yes, yes, they can.

Well, from those responses, you can actually see that content creation is quite a way that people are using it, and I will pull back and say, ChatGPT did become very popular, and it is known as a large language model. So using words, and anything that does involve words, it is understandable that that actually went it was such a powerful tool, and it was released only a short time ago, and it’s understandable that the base of people using it is those language options, such as content creation.

So we’ve had one more person to respond, so we can get 70% of the audience please, just one more person to respond. Do our There we go. Excellent. Thank you for that last what a great response there. So that was excellent. I love it.

Yes, that is one of the reasons why language and content creation initially became so popular. But it is improving, and there are other tools out there that we can tap into.

So any do you have any commentary about artificial intelligence enhancing human judgement there?

Cameron Anderson

Cam, I was actually looking and thinking that the people that have answered not yet using AI, I think that’s you’re in the right place today, because we’ve kind of broken it down into kind of some easy, easy steps to take so that you can take some action kind of straight away. But Heather, you asked me, what was the sorry,

Heather Smith

Do you have any thoughts on artificial intelligence enhancing human judgment?

Cameron Anderson

Yeah, well, I think at the end of the day, we need to own the outputs of of that. So there’s all sorts of things around kind of bias and how the models, particularly LLM models, are trained. So at the end of the day, the kind we need to exercise our professional judgement, our professional scepticism, and also that the human skills that we have around kind of judgement in terms of what we’re actually taking away from the outputs and then presenting in front of our customers.

So I think that, and we’ll kind of talk a little bit more about that sort of as we go through, but I think that’s important, that you’re always exercising your professional judgement as you would if you’re kind of reviewing to use. You know what you said earlier around if you’re working with an intern or a graduate, you would always kind of review their work and exercise your own judgement across that.

Heather Smith

Yeah, absolutely. And I know, when I was first employed, I spent ages working on a monthly budget and put it up on the screen to present to the all the important people. And the head of the important people walked in and within 15 seconds, said, there are errors in that report. So but what it needed was someone to come through and review it before me as a young intern, put it in such a public, humiliating place.

But let’s move on and talk about the pillars of successful AI adoption. And these are things to think about in terms of, I embrace these, and I can then sort of move move forward and successfully adopt AI.

So start where the validation is easy. Focus on sort of smaller bite size tasks that are not necessarily too important. And you can see, you can see the results immediately.

One of the things is, you can I find, and I’m going to personally take this, I’m quite straightforward and blunt. If I’m writing to someone, if I’m dealing with someone, I want this from you. This is when I want it. Give it to me, and I know the recipient doesn’t necessarily want that. What they want is kindness. They and they want professionalism. So I sometimes will take whatever I’ve written and put it through the process of, can you make this kind can you make a professional? And can you put some small talk in there, like asking them what the weather’s like, etc. So I’m not good at that, but I can filter it through that process, so I have almost enhanced and turned around relationships that I have by using that in email communication.

I don’t necessarily use it in social media, but social media is another great place if you are struggling. And you can help you put words out there, and it can also help you check sort of basic things that you need to check out, get to just get you moving along so you can do things a bit faster.

One of the great things is brainstorming. So as you could brainstorm, perhaps with an intern, you can brainstorm with some of these solutions.

One of the great things, and I think pretty much everyone on the call was sitting here from Australia, is we’re taught critical thinking at school, and this is something that Gen AI doesn’t have. So we can overlay what we’re doing with our critical thinking skills to ensure that we’re getting the best out of it.

Did you have anything that you wanted to say about starting with the validation?

Cameron Anderson

not on the on the that part, but I do want to double click in on the governance piece and just kind of, well, actually, on the whole slide around successful adoption.

And this is not, this is a business objective or a strategic move or change. It’s not something that rests with the IT manager or the, you know, the young person, the young person, or the tech wizard within within a practice or within a firm, because we know that there is a readiness gap between juniors and seniors today, which probably isn’t, isn’t shocking, but it’s making sure that we’re taking kind of the whole team on that journey.

And when we look at the kind of third box on the site, really stands out to me is around building capability and not just tools, so not just kind of ad hoc things, but actually changing how we go about doing the work that we deliver within our firms and to our to our customers, so making sure that we’re building that sustainable fluency.

So what I mean by that is considering the AI in conjunction with existing human capabilities. So as Heather said, critical thinking, judgement, problem solving, decision making, all human skills that we need. And of course, we then need to overlay that with ethics and integrity and an adaptive mindset, all of those. And of course, a technical the technical expertise, which we’re all very well known for as practitioners.

So I think, Heather, I don’t know if you’ve got anything more to add to that. That’s kind of my take on and my sort of advice on the building capability and providing guardrails or kind of guidance through governance.

And one other thing I will mention there is creating a culture of transparency and sharing. And so my team actually take time, time out of their days or of their weeks to invest in this so they can go away learn about something that is aligned to their role or aligned to the objectives of our business, and then we come back and actually share about that. So we’re raising the knowledge across our entire team by kind of dividing and conquering.

Heather Smith

Yeah, absolutely.

Now, did you have a poll that you wanted to run on that while I continue looking? 

Do we like interactive polling? Excellent. So the poll here is, how many times are you and your team each investing and developing AI capability, looking at that over a monthly basis. Now you can’t answer zero because you’re here, so you can’t answer

Cameron Anderson

this counts. This counts.

Heather Smith

What I did want to talk about while the poll is going is one of the things when you have an engagement when we talk about governance, when you have an engagement letter informing clients that you are potentially using AI, the reality is, AI has been sitting in Gmail and Microsoft Outlook for years. So the reality is you probably have been using it in some to some extent, activating it. So all of your small business, your clients, small business data is accessed by AI, sort of a totally different level. So staying aware of where that’s happening, getting their permission, if you’re integrating tools across their data, and how the data is being trained as things to be aware of and things to be involved, to be incorporated into your engagement letter.

Now the other thing that Cam talked about is building capability. What’s, what’s? One of the things that’s going to happen is you’re going to come across tools, and they’re going to make you so much more efficient. And might be happening now. It may be happening in the future, and it may be it will make you so much more efficient.

And one of the things Cam did touch on was growing sustainability. And I would like to really, really emphasise that I know when, um. Online. Let’s call it cloud accounting. I don’t think we use that term anymore, but when cloud accounting first emerged, I overnight tripled my client base. It was so much more efficient than what I had been doing. Tripled my client base. But then when it came to sort of particular sort of high compliance months. It was extremely difficult for me, and it was very, very challenging. And I for many of the months of the year it was easy, then for some of the months it was quite hard.

So I would encourage you to when we were thinking about all of the capabilities and the efficiencies that you’re you, you may absorb and enjoy through adopting AI, grow purposely and potentially, grow your small fish larger, or your clients into larger and offer them different, different services options there, rather than constantly taking on more clients, not that that’s wrong, but just think about how you are sustainably going to grow if that happens, if you increase, if you double or triple in the next two years, how are you going to manage and service that I did want to touch on, that we’ll

Cameron Anderson

talk a little bit about vision and strategy as we go through and kind of give some guidance on how to, kind of derive that in line with your own business objectives as well.

Heather Smith

Yeah, fantastic. How’s the poll going?

Cameron Anderson

have 42% are spending one hour or less, I’m guessing, a month. 30% on two hours and then, and that kind of goes, Peters off from there with 13% investing zero. But as we know, they’re here today, so they Yeah, an hour today.

Heather Smith

Let’s get two more people to respond to more. Let’s round it up to 50. Here we go, two more people, one more, one more person to respond. Please respond, and then we be nice and tight.

Cameron Anderson

50. Oh, that’s gonna play. Oh, bingo.

Heather Smith

Thank you. I think, Oh, look at that. We overshot the mark. Fantastic.

So moving along. Thank you so much. Everyone who responded to that poll, we’re looking at what does good look like. What does good look like here? So I think that we pulled out that something like 93% of organisations are using AI, but only 59% actually have formal governance in place. So, so this is something that we’re we should be sort of concerned about what.

What are your thoughts on this here?

Cameron Anderson

Cam, yeah, I think like, what does good look like to me, potentially will look different to what what good looks like for your business. Heather and we’re working in a space now which is not new, but we’re working with remote teams, micro practices, large firms, who are all kind of accessing data from different locations, different internet connections, all of that kind of stuff, which all requires kind of a different approach.

And so I think it’s probably good time talk about the dreaded QMS. I guess the QMS is fantastic for ensuring practice quality, but talking about how we can embed some of that governance stuff and through through our QMS. Heather, yeah, absolutely.

Heather Smith

I was on a call this morning with a lady who took over when someone sadly passed away, and she said they were not able to access any of the software, so they had to rebuild the financials from start, and it took them eight months to get everything back in order from the point of that the bookkeeper passing away. So while that’s quite dark, it’s really important to have these solutions in place to people have really understood understanding of what’s happening in the business, and so it can sort of operate with other people coming in there, and everyone has a clear idea of how the operations are running and what, what the guardrails are in place,

Cameron Anderson

and just on that, that vision over heavy process, I think what, what I’m trying to Say there is a clear vision is going to provide far more kind of direction to a team where we need to take our teams on this journey. This is a change management piece than what a policy is going to provide.

So we do need policy, but at the end of the day, as a leader, setting a clear vision is going to allow your team to. Kind of see their part and how that led us up into the objectives of the business.

So I think it’s really important to make sure that we do have a vision in line with our business objectives, and we are sharing that and referring back to that regularly as a team.

Heather Smith

Yeah, absolutely. And I think a clear long term vision, like a clear, stable long term vision. I was speaking a panel last week, and one of the ladies mentioned that her staff member quit because her vision changed on a monthly basis, and the staff members just couldn’t cope with it. So that’s something to be mindful of. You may be growing that fast, but your your staff need to either be open to that, or you need to provide something that’s palatable for everyone.

Cameron Anderson

A good place to start with that is kind of zooming out and thinking, Where do I see the professional the industry going in the next, kind of three years, and then where, where do I what do I need to change, or where does my practice need to sit to align with kind of that, to keep up with all the things, changing compliance, changing our customer expectations, and also changing people that we will be working with as we work more and more with, with AI and agents, but also kind of the younger generation coming through, and kind of how they like To work may be different, but making sure that we’re aligning that around a common vision in terms of where we see the industry going.

Heather Smith

Yeah, absolutely, absolutely. So this is one of the slides that I’m really excited about, AI fluency across practice operations. And I’m keen I’m going to sort of talk about the four blocks, but keen for us, as we do, if you have any insights to share with us, do share them in the chat, that would be fantastic.

So in terms of client communications, as well as and we have talked about it, it can assist you in drafting conversations. It can assist you in having meeting summaries. And I know someone has mentioned that they actually use otter AI for recording meetings. It is fantastic at recording meetings, so for everyone listening in and not here present today, it’s great for recording meetings because it then gives you a summary of what happened, and you can dive in and actually ask it question, ask the transcription questions. So another great way to constantly be embracing lifelong learning.

Now, one of the things that I really like about it’s sort of in client communications, is you can set it up to role play with it, to have challenging conversations. So you can say you are the client, and this is your perspective. I’m the I’m the bookkeeper, and actually have those conversations which, which can be helpful, helpful to move you along on that conversation.

Marketing we did mention briefly, but, but cam,

Can you describe how you made this beautiful slide deck using AI?

Yeah.

Cameron Anderson

So some of you, if you’re got attention to detail, may have noticed the made with gamma in the bottom right hand there. So Gamma is an AI product that allows you to create beautiful slide decks based off prompt inputs. So Heather and I got together, put together the outline for this webinar, or the brief, as we call it, and then loaded that into Gamma and said, I want 10 slides, and it’s put this together, and we’d made some some tweaks, but largely, it’s really hit the spot.

So I think a lot of bookkeepers and accountants do present regularly, whether it’s BNI or to a local scouts club, something like that. So that’s a fantastic way to kind of try your hand in a low-risk environment, really assisting you in delivering a Polish presentation. So recommend Gamma. I know you use it as well.

Heather Smith

Heather, yeah, and I think Cam and I are quite thrifty and still on the free versions. Every month it replaces the accountants. Every month it replenishes with, I don’t know, little chips or coins. And it keeps working for me. So for me, I use it as a really good brainstorming, a creative brainstorming, to sort of push me on the way. And I love the way. It sort of splits out things and goes, Okay, maybe, maybe represent this communication this way.

So I’m always seeing people, Oh, someone said that they’re not seeing the slides. Wonder why they’re not seeing the slides.

I always find there’s always presentations opportunities happening, and it can look really good if you use something like that. But let’s move to talk about reporting and insights and. How you can use AI for reporting and insights.

Now, someone did mention that they’re very timid. They reluctant to put client data into various things. And I do hear that there are some solutions. I’m being attacked by a fly. It’s Brisbane, it’s Australia. Some solutions are very strict about what guardrails that they have in place, and that’s something to sort of monitor and look at in their declaration, their terms, and sees terms and conditions, what they are in place, and whether you have the option to not let the AI be trained on the data. So that is something to be mindful of.

But one of the other things I will let you talk in a second, Cam, one of the other things that I do recommend you do, which is talked about, I’ve talked about it few times, and there’s another guy called Kaya Patel who talks about it is, before you go, you can, before you go to a meeting, do a quick search, like a quick standard search on the client and the macro, micro economic environments of that particular industry, can give you some insight and awareness of what you perhaps some of the conversations you need to have with them before you even arrive and help you prepare for that meeting.

For example, it could do a quick you could you have it all standardised through a GPT, but it could search for any potential grants that are available, any potential opportunities that are available, and what are the macro, micro things happening in the industry? Because a lot of that information is available to a certain extent, not necessarily on you, the bespoke micro businesses, but some of the other businesses.

Anything else you would like to comment on in the reports and insights there?

Yeah, I’m just

Cameron Anderson

going to jump in there really quickly. And I know certainly when I was when I was in practice, that a P and just if we think of P and L review, really basic task in theory, but it can be 45 minutes to an hour to do that, just to do the financial analysis, but where we’re getting really powerful insights.

Now to your point, Heather is adding the context to this. So to that. Sorry. So here’s a P and L review of Bob the Builder. How does, how does Bob the builders business compared to other builders within, you know, the same area, or the same geography, the same size, and we’re getting that level of insight that we can then start to kind of benchmark and have those really valuable conversations with our clients.

And I love Heather, that example that you used around, what are the funding opportunities available for this business? I think that’s really cool. And having that conversation with your client, your fund writing is or grant writing is an art in itself, but absolutely an area that we can be stepping into as bookkeepers in this, in the space.

Heather Smith

And as bookkeepers, we don’t necessarily need to do the whole fund writing, grant application process, but what we need is a little black book with names of great people that we can refer to.

I’ve walked into a client business. I’ve said, Did you know that your, your, this fund, this, this grant, would be suitable for you next meeting, they got $50,000 in their bank balance from that fund. So it was surfacing some information that took me two minutes to share with them, and then giving them in the name of someone in the network and that, you know, that’s brilliant for our clients to be able to do that.

Cameron Anderson

The other one that we’ve got on there is data quality, and I’m actually going to talk about this in the context of expert a little bit further down the track. But of course, the foundation of any good advice or kind of reporting an insider output is definitely sits with having that high quality data available to us, so validating the data and kind of making sure that we’re adhering to our professional standards there.

Heather Smith

Oh, absolutely. And that’s so important like to any data in decisions, any any decisions that are made or an inaccurate data are just fallible, and they will get you into a really bad position, and they’ll get and because everything’s moving so fast at the moment, they’ll get you into that position faster than before, at least before we we were kind of slow in getting To that point, but ensuring that our data is accurate and ensuring that we’re we’re actively bringing in the cash when it should be coming in is really helpful for our clients.

And let’s move to talk about practice or operations and streamlining routine admin tasks. So. Scheduling capability, which is a really scheduling capacity. Yeah, it’s really critical as businesses grow.

Now, what were the insights that you’d like to share here?

Cameron Anderson

Cam, I think in terms of capacity, it’s actually something that is not done very well by any firm, to be honest, that I’ve come across, but we’re we’re working with partners and customers, is to actually automate some of that scheduling.

So if we can build as much kind of information about a task or a job in expert, for example, then we can start to assess the capacity across our team. We’re getting really smart with integrating payroll. So we know if Heather is away sick or Heather’s away on annual leave next week, we can then start to look at, well, what’s that going to how’s that going to impact our capacity, and where are we going to be able to reallocate that so interacting with in kind of natural language.

So for example, I’ve got a 10 hour payroll fix up job. Who can I give it to? It’s going to let me know. Well, Heather’s payroll. Heather is not taking on that payroll. Heather is a payroll specialist, but she’s at capacity. How about we allocate it to Anne or Susie? Yeah, the ability to interrogate that data in in in live time is fantastic.

Heather Smith

And payroll is really complex, and I encourage anyone who specialises in it to increase their rates, because it is a really, really complex area that is forever changing. So that’s my championing the payroll specialist.

Let’s move to talking about building AI capability and the journey that you’re going on. So thank you so much for joining us here today and letting us sort of unpack this conversation with you.

And I think that at the moment, there’s so much talk in the atmosphere about it that it is tapping into those conversations and sort of pulling at those threads and doing a lot of your own remaining curious, doing your own experimentation, and checking things out for yourself, and staying in the forum conversations and tapping into the people Who are actually talking about it, and I know from the some of the people who’ve signed up to join the session today are absolutely Australia’s leaders in this space. So so we’re very grateful and appreciative that you’ve joined today, but you probably might like 100 times more than than myself for one, but I think it’s constantly looking at what is out there.

I encourage you to continue to join in with the ABN knowledge sharing educational workshops that they run, and to tap into your communities and to your networking groups and experiment.

Anything else you’d like to add there?

Cameron Anderson

I think, just the if I was to kind of summarise this around assessing developing and deepening, is that this is actually a loop, so it keeps you’ve got to stop, step back, assess your current state. Where are the skill gaps? How we’re going to continue to deepen so I think evolving that is part of your kind of psyche, or way, way that you go about business, I think is going to be beneficial in terms of meeting that vision long term.

Heather Smith

Yeah, absolutely, absolutely. So I encourage you also, I find YouTube really useful. YouTube and some of the podcasts out there, if you search for particular things can be, can be quite useful, but it does end up being a bit of a rabbit hole sometimes, so protect your time as well.

But let me move on, and let’s talk about prompting that works, and they’ve used the anagram the rtri framework, which is the role task requirements and instructions be the AI doesn’t know what you’re thinking, and it doesn’t know the nuance of what you’re thinking. So you may ask it something that you think is really quite straightforward, but it doesn’t know where you’re going with that.

I remember the story about an Asian man who went to London and he couldn’t understand he was walking around, and he couldn’t understand why all the walls in London had these big signs that said, Post No Bills. And he was Why would people want to mail duck bills to other people? And why is there so much signs protesting the mailing of duck bills to people? And I think it kind of sums up the nuance of the actual situation.

So. Role define what you want the AI’s perspective to be, and the expertise level from the task. So do you want it? I frequently will ask for it to give me a definition of something at a 15 year old’s level. So that’s towards the adult range, but still going to be simple enough for me to understand. And that will be, is this an explanation? Is this going to be done at an advanced level or at a beginner’s level? And if it comes out not at the right level, you can just go back in and readjust. So I do think that that’s really important. And hands up how many conference sessions we’ve been to, and we wanted them at advanced level, and they came at a beginner’s level. And we’re like so it will listen to you, it will listen to you, and it will deliver the level that you’re wanting, clearly articulate what you want the AI to accomplish. What are you trying to get it to do? And sometimes I will just verbally, just go do it on a voice recorder and just speak to it and tell it what I want it to do, and sort of frame it up. So it may not be a simple sentence, it may be quite complex or may be quite involved, but I’m asking it to do the requirements.

Specify the format. Where is it going to be used? Is it in an email? Is it in a blog post? Is it to another person, like an employee or social media. How long do you want it? I’m always telling it to be concise. Otherwise it runs off. It just the response is too long, really for me, I’m not here for war and peace. I’m here for something concise to absorb.

What tone Do you want? As I said before, I like it to be professional and kind and the quality of the expectation. So I’m not trying to explain it to Einstein. I just sort of want it to sort of basic, so Joe Public will understand it at that sort of standard level, and then furthermore, give it any other bespoke, contextualised, new nuanced details.

So I frequently will write something up that I need to talk to someone about, but I’ll always sort of put in and he has, I would highlight, he’s a plumbing business. It’s a plumbing and so it may contextualise it more around him being in a plumbing business. And I find sort of understanding those nuances and pulling them all together can really help.

If you’re using a solution like chat GPT, you can put things like your tone and like the length of output you want in the settings. You can define it in the settings so you’re not repeating yourself all the time. And then, if it comes out unsuitable, go in and re refine it. And if you go in and don’t get the response you want, go back work with it. But also take it away, work on it and go, Okay, what you gave me, I’m giving you feedback. This is what I turned it into. This is a feedback loop, and I’m educating you. This is the final outcome that I wanted, so it learns your style and your voice and improves like an intern over time.

Course, you don’t want an intern to be a mirror image of yourself, but Cam’s A Yes, I do. Yes I do. But you kind of want to move them along in terms of structurally, how they’re to do things. Yes.

So any questions about that, anything that you want to add to that, Cam?

Cameron Anderson

No, I think I’ve This is credited straight to the cans, child account, Australia, New Zealand. Ai, fluency handbook. And I think this is a really concise way of kind of looking at it, the role, task requirements and instructions. And that’s how you should be able to hopefully get the best out of it. And it’s and it’s not just, you know, our minds go straight to chat GPT. It’s anything generative. Ai, yeah,

Heather Smith

absolutely, absolutely fantastic. And, and someone’s commented, you can tell chat GPT not to agree with you, which is fabulous. You can actually say to argue with you, and that’s kind of touches back on that role playing that I talked about. You know, you can say, okay, you can work through a decision where some people won’t agree with you on particular things, and you can try and thrash out the discussion that needs to be had.

Add there. And we’ve all had those situations that that clients are not agreeing with what we’re telling them.

So with that, let’s move on to talk about the importance of responsible and ethical AI, use now. Cam, what are your thoughts on this

Cameron Anderson

general high level, I think we we just need to remember, be mindful that Gen a Gen AI, lacks the human skills to make ethical decisions, and it can be subject to bias. So my advice there is, don’t engage with it passively and really own the outputs. And you’ll see that it’s written around output ownership. So we need to validate the accuracy, ensure that it’s fit for for your context.

Often a particularly gener Gen AI will want to present confidently with plausible sounding responses, which is known as hallucination. You’ve probably heard that before, but ultimately, we need to be to be able to exercise our professional judgement, and in all outputs from Gen AI, the one that’s that I would like to throw to you as a question, Heather is question is the principle number three. So disclosing AI use appropriately to clients, and there’s a bit of a debate in some of the forums online around do you tell the client? Don’t tell a client. I’d love to hear what your thoughts are.

Heather Smith

I would absolutely be telling the client within the engagement letter or as you’re evolving with them, but essentially in the engagement letter and in the conversations. But I would be explaining, I would sort of be further explaining that, as I mentioned before, AI is sitting in your emails, in your email responses. It’s also sitting in your Word documents. Every time you write something, it’s trying to finish it off with off for you now, in terms of use of it, it would be saying, I have assessed the guardrails and I have the highest security in place. And of course, if you’re not at that level, then, then you would highlight that.

So I almost think it’s impossible for you to not be using AI if you accessing the internet and any software at any stage, it is the level of you using it, and it’s kind of staying informed, and it’s staying informed and sharing and being really, really transparent about that, doing it, keeping your professional association should be based able to sort of help you in terms of understanding what solutions are suitable and approved and Okay. Things happen all the time, as we know, and it’s more of a thing of how they the solutions again, respond to whatever happens. Yeah?

So yes, transparently, share that with clients. I don’t, I don’t see as there being any other option. You wouldn’t want, you wouldn’t want your data, your information, shared with people without your authority,

Cameron Anderson

and from a practice efficiency perspective, or just general kind of business practice, I would be concerned as a customer if my practitioner wasn’t starting to, at least starting to look at how they could incorporate some of this into their own workflows and processes.

Heather Smith

Yeah, absolutely, absolutely, yeah, absolutely. I went to my nail technician the other day, and they’ve got, now they’ve got this big sign that, like my lovely nails, that they’re using AI in the heather, so it directs the heat directly to my nails, so supposedly dry it faster. How good.

Now I’m not choosing the nail technician for the AI, however, she feels that advertising that is beneficial to me. So I do think that when I saw a dentist the other day, and they had a big sign up saying they use the latest technology, and it’s like, I perhaps would prefer a dentist that’s using the latest technology rather than sort of a dentist that that’s not so it is. It is an asset to you. It is something to be proud about and just do it smart and in an educated and professional way.

Yeah, anything else there?

Cameron Anderson

Oh, I think that’s, that’s good, Heather, I think that’s really valuable. Particularly, I wanted to double click on that discussion around, do we tell clients? Do we not tell clients? What are our actual obligations? So I think, yeah, absolutely.

Heather Smith

Now someone has popped in the chat that they find chat GPT continually provides incorrect responses. Totally hear you ask for sources, and there are a number. Solutions in terms of tax legislation that are now online that people are really raving about. So that’s something that you could again look to. I would jump in, jump in the expert Facebook group and ask that question, and I’m sure we can put a list of them in there, but ask for sources. So this is my question, give me legislation resources or something like that to refer to.

Cameron Anderson

And I wanted to talk about expert here, Heather and some of the quick win use cases. So looking at what are the things that take kind of high human effort, but are easy to validate, and we can leverage the AI directly within the expert platform today, so I know that there’s a number of current expert customers on the call. So special, welcome and thank you.

But if you actually look at this from one to three, it’s actually kind of Heather, the evolution of the Xero product, and we’ve always been at the forefront of technology. But if we look at the kind of data quality checks piece, and that’s kind of anomaly detection identifying issues at the data level, that’s kind of based on the sort of traditional AI algorithms, and then we’ve moved evolved into custom experts, so now we can actually Start to query large data sets across our entire client base using natural language.

So we’ve also got the maintaining kind of review gates, or what we like to call human in the loop. So if there’s something that’s particularly risky or needs validated, we can allocate that to a certain person within a workflow so that we can ensure that we’re giving the best advice there.

And then the third piece is the intelligence reporting. So this is really the insights reporting that we spoke about around the on the practice operations slide. So we’re looking at both client and practice level reports. The example I gave was who has capacity to do a 10 hour payroll job, but we’re actually starting to really speed up and accelerate those kind of reporting cycles, particularly around workflow and capacity in a firm, but also from the ability to provide client advice. And it’s contextualised client advice, because we are using the prompting framework rtri to actually interact with those outputs.

So to use your example, Heather, this is a this is a builder, or this is a nail tech. Or what should the wage margin be for nail tech? For example, we can start to evaluate that and really provide solid advice to our customers. So I thought this when I laid the slide out, I thought it was really interesting that it does show the evolution of AI within our own product.

Heather Smith

Yeah, absolutely. And it takes people on that journey of really feeling confident about the data and what they can do with the data. So it’s very, very exciting. Micro and small bookkeeping practices now have the ability to do so much more with their clients, and bigger clients, because they can really dig deep and get this sort of access and this intel, fantastic.

So we’ve had the three quick wins use cases from expert, and now we want to talk about your adoption road map, sort of starting small monitoring and learning and sort of scaling with capability.

And I’m very much a person who thinks, ah, if I want to learn something, I put aside 15 minutes a day, every single day to go over something, to learn it, and then I will the next day, I might go back to understand that first thing, or I will go forward to learn something else about that particular topic. And I’ll do it with sort of a book or a YouTube video or something like that, or a podcast, and just push myself to to understand 15 minutes. So it’s it’s over a cup of tea, it’s not too daunting, because it’s just 15 minutes, and as soon as I’m into it, it’s like 40 minutes, and then it’s 30 minutes. So it’s a really good way for me, sort of bite sized way of learning something.

We see a lot of people in the industry who do it through experimentation, through putting side time aside and learning, tracking how you are using it. Is it being effective? I will typically pay for anything that’s going to save me time. Time and assess it as I sort of go along with that process. Don’t necessarily adopt that philosophy, but that’s kind of the way I would do it, but monitor the impact it’s having on you. And sometimes it’s funny because you think it’s going to do one thing, but it does a whole lot of other things that you weren’t expecting, which are really beneficial, and talk about it and share those insights with other people, with other people in the community, in the expert Facebook community, and within the Australian bookkeeping network community, because they’ll bounce off things.

And I know that, like I facilitate a community and so much learning happens from people just accidentally having a conversation about something and something else bouncing off there. Oh, I could do that with that. Hadn’t thought about that, that sort of that brainstorming from real people. And and you will build, and you will adopt, and you will sort of grow with the AI understanding and sort of your own fluency.

And one of the sort of the challenges that the bookkeeping industry faces is that that frequently you’re in the micro, small, medium sized firms, and sometimes the people talking to you are at different sizes. So it’s sort of really being mindful about whether what they’re talking about is adaptive for you, and that’s why the communities are so great, and so the Australian bookkeepers network is so great.

Anything that you’d like to adopt?

Cameron Anderson

Just on this one, I think, is mapping out your key processes and identifying where there’s any kind of potential repetition and or bottlenecks, and looking at how can you kind of chunk those off. So a lot of people kind of scratch their head of where to start, and we’re looking at identifying two or three of those. And if we can, can make a difference of five to 10 minutes on each of those that then that adds up, particularly if you’ve got a team.

So it’s all those things that kind of always go on the back burner, but you do take the time to to to do that process, sit down and share with the team the lessons across, make sure that we’re monitoring and continuing to iterate that. And then back to our original point at the very start, is make sure that we’re scaling that, that capability across the practice.

Heather Smith

Yeah, absolutely, absolutely. And so it looks like you’ve put up a another poll, poll question. There is that correct?

Cameron Anderson

I have indeed. Would you like a further discussion on AI integration for your business? So there is an opportunity to meet with the expert team and discuss how we could look at integrating AI further into your business, but our advice from today’s session is really to start your journey if you if you haven’t and you have already by being here today. So thank you for joining us.

Practice, learn the rtri framework and consider what if you’re an expert customer, what are the quick wins that you can get from leveraging the platform today? If you’re not an expert customer, highly recommend that you reach out, indicate through the poll there that you’d like to hear more, and someone from our team will get in touch.

And just finally, I wanted to mention that we do have a setup Academy series kicking off next week. Heather know if you know about that one, but that is a four part guided setup series for accountants and bookkeepers to configure and optimise their Xero account and get it set up from scratch from day one. So Cassandra Scott and our customer success team will be leading those sessions. There’s two sessions next week, and then two sessions the week after that.

Heather Smith

Fantastic, fantastic. And it’s such a great time of the year to be looking at that, because it kind of, then you’ll have the festive period to sort of have have sort of some deep musings on the beach. And I have been told that the most sign ups for software services happen in January. And I suspect there are people on the beach and they’ve had a thought about they’ve got lots of Turkey in their belly, and they have having things about how their business is going to run in 2026 so with that, what is Cam?

What is one thing you’d like our listeners to take away from the session today?

Cameron Anderson

My one piece of advice would be, make sure that you are developing capability in line with human skills. So on our very first slide around the AI human partnership. Make sure that we’re looking at, how can we augment AI without human kind of, our human skills, our human judgement, to really provide the best output for both our businesses and the processes for how we run our businesses, but also in the service delivery to our customers. Absolutely.

Heather Smith

Absolutely, thank you so much for that Cam. And look my I’ll share. And I can see people saying that sometimes the comments from ChatGPT are incorrect. Tell it just that you would do an intern. Tell it, give it feedback. It needs that feedback loop to learn. And yeah, sometimes you just need to shut it down and start again. Sometimes you do the day, but I would, I would remember there’s a feedback loop, and it’s a learning feedback loop. Tell it what it should have been, and circle back and keep, keep trying.

It is extraordinarily confident, and it is funny how we believe people who sound confident. So we know in the big wide world that that’s not always the case. So thank you, Cam.

Cameron Anderson

Thanks so much, Heather. Take care.