I hired my first AI assistant, here's what I learned and mistakes you can avoid

Trent McLaren: On the dot, and we are here. I was just saying this to Nat before: strategically, I've actually tried to place some really great sessions all the way through the different points of the day. And this session in particular had about the fourth most registrations. So you were all very excited to hear how Nat has hired her first AI assistant in her business. So, Nat, how are you? Welcome. Thank you for being here.

Natalie Lennon: I am well, Trent. I am very excited to show you what I've been doing with my AI assistant, Viktor.

Trent McLaren: I love it, it's very good. We're excited to hear all about it.

Trent McLaren: If you've got questions, dump them into the chat. I can see a few chat questions already coming through. And yeah, Tarun, that's a great first question: does hire mean pay a subscription fee? Look, things cost money, but I'm sure we'll cover what that means in a sec.

Trent McLaren: The more fun part, though, is what Nat is about to tell us, which is that she hired her first AI assistant. Nat, you're going to talk us through what you learned and what mistakes you can avoid. I'm excited.

Natalie Lennon: Yes, I am. Let me just get the slides up. There we go. Now we're cooking. And yes, hire means pay a subscription fee. My cheapest employee.

Trent McLaren: Love that.

Natalie Lennon: Before we get into what we're talking about, just a little bit about me. If you know me, you know I don't really like talking about myself, so I'm not going to go into the full detail of all the things I've been up to, because I've been very, very busy. But basically, I'm a chartered accountant and tax agent with over 20 years' experience working with and in small, medium and large businesses. I love working smarter, not harder, so AI is really my jam. And my latest project, which is launching next week, actually I just dropped that, that's a huge secret, we haven't publicly announced when we're launching, is Automation Playground, which is a place for accountants and bookkeepers to learn AI in a practical way with real-life workflows. This is not a sales pitch, but I thought I would include it.

Natalie Lennon: So, the next 45 minutes. We're accountants, we like checklists. Well, I do anyway. So here's the checklist. Pain points: what's eating into my week. The hours I got back, by the numbers. What Viktor is and why I hired it. Setup: how to actually onboard it. Use case one, Xero, multiple client files. Use case two, Karbon: where jobs and deadlines actually sit. Mistakes, so you can skip them. Keeping it real, and where to start. Because when we're learning about AI, one of my biggest pain points, and the reason behind wanting to help train accountants and bookkeepers on AI, is that a lot of people go, oh, I'm doing all this cool stuff with AI. And it's like, yeah, but how are you actually doing that? Show me how. Every time I ask someone to actually walk me practically through how, they either can't do it or it's too convoluted. I need to see it. So that is what we're going to show today.

Natalie Lennon: So first thing, what was eating my week? I'm sure we can all relate to this. Drowning in email: my inbox, plus the team's when they're away. Meeting follow-ups. Vinyl makes meetings absolutely fantastic, but the follow-ups, I still have to manually go into Vinyl, pull out the transcript and email it off. We were trying to set up a Zap, but because I use Outlook, it doesn't play nice with Zapier, so that was a bit manual. And now I've been able to get Viktor to help me automate that. Chasing the team, checking Karbon and seeing where jobs are at and when they're due. Checklists: TPARs, pain in our butts. Xero health checks, a great service to offer clients, but they do take a lot of time. And workpapers, the number one bane of accountants' and bookkeepers' lives. They take a lot of time. And also group clients, so inter-entity reconciliations. They're another pain point that we just love. They're just so quick and easy.

Natalie Lennon: So these are the fixes, the things Viktor has helped me solve. A daily brief, which I'm going to show you in a minute, of the emails that actually need me, with the replies drafted in my Outlook before I even log on in the morning. Game changer. Meeting follow-ups: what Viktor does is he, or she, checks my calendar in the morning and gives me my daily brief of what's on for the day. Then it notes the meetings I've got. Then it remembers that five minutes after the meeting the Vinyl transcript is going to come into my inbox, and it takes that and uses a template I've given it to draft an email follow-up and save it in my Outlook. All I need to do is review and hit send. Now, that's been an absolute game changer for me.

Natalie Lennon: If you've got back-to-back meetings and you're constantly having to get the transcript, draft the email, check the email and send it off, it takes a lot of time. Even though it's a lot quicker than before we had amazing apps like Vinyl, it's still a couple of minutes each one, logging in, clicking, doing all the things. Every morning at 8am I get a Karbon practice review. It goes into Karbon, pulls out all the work by status and team member, and gives us a brief in Slack of what's happening. Checklists: checklists for TPARs and Xero health checks and workpapers. Viktor does those now. We don't have to do them manually, which I'm going to show you in a minute. And group clients, so the inter-entity reconciliations are now a lot simpler.

Natalie Lennon: I'm not going to go into detail on this. I'm not a salesperson, I'm not selling you anything, but we love numbers, right? So on average, six hours back in my week just from automating a few little things like emails, client meetings and the Karbon weekly review. I think with AI we get so caught up in wanting to do these big, amazing things, automate my BAS, automate this workflow. But if we just start with these little quick wins, they all add up, and I think they're underrated. All right. So I've given you a bit of an overview, a bit of a teaser, and now I'm going to actually explain what Viktor is. My best employee never calls in sick, never asks for a pay rise, and doesn't judge me when I message him after 10pm.

Natalie Lennon: It's basically an AI agent that lives in Slack or Teams. So if you don't have Slack or Teams, you can't use Viktor. But one of the reasons I love it is that it's not another app I have to log into. It's in my Slack, where I'm already interacting with the team. How we met: like any love story, on Instagram. I saw it on Instagram and I was a bit sceptical, because at the time I had the team on ChatGPT. I'd moved off ChatGPT onto Claude but hadn't put the team in there yet, so I was kind of dating both at the same time. And then I thought, let's check out Viktor. When I checked it out, I was like, wow, this is a game changer. And every accountant or bookkeeper I've told about it, even though, disclaimer, it can be a bit more expensive than the other apps, has said to me, Natalie, it is absolutely our favourite.

Natalie Lennon: Look, we know a lot of AI is moving very quickly. In fact, Trent and I were having a chat before about Claude, and it's been a little while since I've used Claude for certain things, so perhaps I've got to go back and have a look at that. So obviously, always do your homework and test out these apps and make sure they work for you, because just because it's right for me doesn't necessarily mean it will be right for you. The main reason I started with Viktor was that I felt ChatGPT is very siloed. You give people their own access, even though it's part of the team plan. You can set up your custom GPTs and your projects, which I think are now GPT plugins, and you can share them, but you can't work collaboratively together and see what each other is doing.

Natalie Lennon: With Viktor in Slack or Teams, you're doing it together. I can show them the prompts and what's happening in a team environment, and that for me was pretty cool. Not to mention the integration with Xero is better than any of the other apps natively, and I'll show you how in a minute. So like I said, the team can see exactly what I'm doing. So what exactly is an AI agent compared to a general chatbot? Like I said before, it lives in Slack or Teams, no additional login. It connects to all your tools: your email, your calendar, Xero, Karbon and a lot more. It works with your whole team, like I said, so you can have it in shared channels. The memory is very, very good. It remembers everything, which I know ChatGPT and Claude have now caught up on a lot better with their memory.

Natalie Lennon: It actually does the work: reports, spreadsheets, reminders. It's even proactive. It's in our daily huddle channel, for example, and our team channel. Just before I went overseas recently, I reminded everybody in our daily huddle that for our monthly payroll clients this month, the 15th was going to be on a Saturday, so don't forget to do the payroll on the Friday. Viktor jumped in on the Friday and went, hey everybody, just a reminder, Nat's on leave, don't forget she told everyone to do the payroll today. And we were like, what? So, my proactive employee. We were like, that's really cool. He can also give you thumbs up and fire emojis in the team chat. That's not the main reason, but it is pretty cool.

Natalie Lennon: So, chatbot versus AI employee. This is basically comparing the chatbot in Claude and ChatGPT. I know they can do some AI agent things now, but when you're comparing it, you've got, like I said before, your own app versus a team environment. With ChatGPT or Claude, everyone has to connect everything themselves in their own login, whereas with Viktor, as the administrator, which I'll show you shortly, I go in, I do all the connections, and then I decide the permissions. For example, my email is in there, and I give my practice manager permission to my email, nobody else, so she can use Viktor to query my emails. So it's more collaborative, and the team learns, like I said. It's got more integrations with Xero. It can actually write to Xero, which I know you can do with Claude and ChatGPT with specific connectors, but that's relatively new. A lot of the Xero connectors before were read-only, whereas with Viktor you can actually write into Xero.

Natalie Lennon: Which for me was an absolute game changer. Someone said in the chat, I always thought you used Claude and renamed your AI Viktor. No, that's the actual name of it. And you can actually rename Viktor as well. Sarah Lawrence and I have a shared Slack channel for our podcast and for Automation Playground, and we've both got our Viktors in there. We had to rename them so we knew whose Viktor was whose, and her business name is Hot Toast, so we renamed hers Hot Viktor. Now, I think that could be an HR issue. But anyway, we thought it was very, very funny. And I think we've got another friend who's named his Vicky, Vicky something. But anyway, I digress. That's a little bit of fun.

Natalie Lennon: All right. This is not a sales pitch, but I will give you my Viktor link in the chat, because if you do sign up you get $50 off for the first month, and I also get some credits. So it is a bit like a pyramid scheme. But the reason I push it out is not to earn any credits. The reason I push it out is because I just absolutely love it. I'm not sure if this is too small on the screen, but basically you log in and connect your Slack or your Teams, and then you go into the integrations page. If you've ever used something like Zapier, it's got a very similar interface. I like to think of it as AI for normies. You don't have to be very technical. You go in, you connect everything up, and the way you chat with it is more conversational than with the other AI tools.

Natalie Lennon: It connects to a third-party app called Pipedream, which is an integration layer for AI agents. Obviously do your own research in terms of security. But my opinion was, well, if Xero allows them to connect, then I'm comfortable that it's safe. Like all of the AI at the moment, none of them connect to Xero Payroll, which is a Xero limitation, so that is one of its limitations. But yeah, you connect it up in here. And then the next part of that is the permissions. As you can see here, which I love, you can give it different access for Xero. I have mine fully set to ask for authorisation for everything, because I don't want it changing anything in Xero without me hitting the button. So you go in and you can update all the access.

Natalie Lennon: You can also see over on the right-hand side that you can give certain members of your team access to certain Xero files. You connect the Xero files one by one. That's probably the only thing that takes a long time, and when you're connecting them you need to rename them, because otherwise they show up as team accounts. You want to name them so Viktor knows which one's which. And then you can give the certain permissions as you're going through there. So the next step of onboarding your new hire is simply to send it a DM and say, hey Viktor, what can you do for me, given I've connected all these tools? What are some of the examples? If you're starting out with AI, I always say, if in doubt, ask the AI. The AI will teach you how to do something. I won't go through all these, but I believe these sessions will be uploaded, so you can go back in and take a screenshot.

Natalie Lennon: There's a whole long list there of all the different things. Then the next step would be to set up some channels. Channels in Slack or Teams for Viktor are kind of like your projects or custom GPTs in the other apps. If you've got a larger team, you probably don't want to set up the channels, and you can just give the prompts to the team to use in their direct messages with Viktor. But I've got a small team, so I like doing it in the chat, so that if there are any issues and we need to change the instructions or anything, we can do it all together. We just have the channel muted so people don't get interrupted all the time. This one for us is a TPAR channel where we have the TPAR instructions, which I'll go into in a minute.

Natalie Lennon: Like I said before, I was going to do this demo live and actually show you the screens, but the reason I haven't is that I realised it's a lot easier to pre-record this stuff and blank out client names than it is to do a live demo and try to hide the information. So that's why I've put screenshots in where I can, so please bear with me. I'm also happy to answer any questions after the session if you want to connect with me on LinkedIn. So like I said, AI for normies. Plain English, you don't need magic specific words. Send it a DM: find the latest email, draft a reply in my Outlook. Sometimes you don't even have to tell it how you want it to reply, it can just read it and work it out. If you want a specific response, you obviously have to tell it.

Natalie Lennon: So someone said, I've found that it costs a lot of money talking back and forth to Viktor, so I'll talk to ChatGPT first and then go into Viktor. Yeah, that's a good call. Potentially. I don't notice that as much anymore because I spent a lot of time training Viktor before, and when I was originally training Viktor I was referring a lot of people, so I had a lot of credits. Whereas now I guess Viktor knows what I'm like. A bit like an employee: you train your employees, they know what you're like, and then they don't need as much help from you. If that makes sense. So like I said, you don't have to go, hey Viktor, in its own channel, TPAR for example. You can also have recurring tasks. You can go into the scheduled page, which is where the integrations are, similar to what I showed you before, and set up scheduled tasks like you would in ChatGPT or Claude.

Natalie Lennon: Or you can simply ask Viktor to schedule it. So: at Viktor, every Monday morning at 8pm, send me a report of what's overdue in Karbon. The way I like to think about it is, if you're asking any AI anything more than once, set it up on a schedule. It'll save you so much time. All right, so first use case: Xero. Like I said before, most AI apps only connect to one Xero file at a time. Now, I believe Claude, ChatGPT and Copilot can now connect to more than one Xero file. Xero hasn't announced it yet, but it is in there. But it is very limited in certain things it can do. Whereas with Viktor, I think I've got 50 files connected. So basically any time we need something from Xero, a report or whatever, we just query Viktor and Viktor pulls it out for us.

Natalie Lennon: And the inter-entity stuff is crazy. The other day I had four entities with inter-entity loans. They've also got inter-entity management fees and director fees and a whole lot of stuff going back and forth. That would have taken me a long, long time to do manually. I know there are other apps that do it, but I got Viktor to do it and he gave me a report. I think he did it in about five minutes, and he did a report in Excel with all the variances. I showed the team, and I was like, this is what we would have dreamed about when we were first starting out in accounting. And for me, one of the biggest game changers with Viktor has been the multi-entity Xero connection. So like I said before, the inter-entity loan accounts, balance sheets.

Natalie Lennon: I also use it for Xero health checks. We have a service where we offer a Xero health check to clients. It's a great lead gen. I think it's $200, off the top of my head, and it would probably take me a couple of hours to do before. It was a Word document, and I did it manually. I gave the template to Viktor and said, okay, recreate this, and made a channel for it: whenever I ask you for a Xero health check for a client, this is what I want it to look like, and make it prettier and branded and all that sort of stuff. So now it does those health checks, which is fantastic. Oops, I've just skipped across here. What's happening? There we go.

Natalie Lennon: My quarterly meetings: I use it to go, at Viktor, I'm about to have this quarterly meeting. Can you go into Xero and summarise their position? Let me know anything I should be flagging in this meeting. That really helps me. Obviously I do it manually as well, but then I get that to help me go, okay, these are the things I may have missed before the meeting. Sorry, I'm getting sidetracked by the questions. I'll leave those for the end, if that's okay. All right, so what we get back from Xero. Like I said before, we get a really pretty report. We look like a superstar. One of them even picked up deleted bank transactions. That's not something I would generally look at when I'm doing a Xero health check, but I thought that's pretty cool. So sometimes AI picks up stuff we would never even pick up, and it just makes us look amazing.

Natalie Lennon: A supplier with advisor access to the file, which I thought was a bit random. TPARs. TPARs have been amazing as well. We set it up so that it goes into Xero, cross-checks all the transactions in the materials report and reviews the attachments in those transactions. I'm not sure if any other AI can do that. I'd love to hear if you've been able to do that with another AI. Maybe I'm late to the party on that, but it read the attachments. Then it checked the ABNs on the attachments against what's in Xero and the ABR, and then gave us a report. And the team was like, wow, that has just saved hours. Then we checked it, and we said, okay, Viktor, update those ABNs in the Xero file. So that cut our TPAR time down substantially. I was pretty impressed, I must say.

Natalie Lennon: And then the end-of-year workpapers are another huge one. For our workpapers, up until recently we were using Excel, and it was something I had played around with AI for months and months trying to get working. It does finally work now in ChatGPT and Claude, but it's not as pretty. What we used to do is upload the GL summary and the GL detail and ask it to put them into one Excel file, with the GL summary on the first tab hyperlinked to the GL detail in the different account tabs. And before, it couldn't do it. Whereas now we just say, at Viktor, prepare the end-of-year workpapers. It's just the draft, right? It's the starting point. But now we just get Viktor to do that. We don't have to download the reports, we don't have to manipulate them, we don't have to do any other work.

Natalie Lennon: So these are the starting points, I guess, for probably any AI, whether it's Viktor, ChatGPT or Claude: look at the things that are taking a lot of time, and then go, okay, can AI do this? Then test it, and then share it with your team. Because we're all time poor, right? We're all up against it. I think I just jumped ahead. No, Karbon. So use case two, Karbon. We use Karbon for our practice management. Every week I get Viktor to give me a weekly timesheet review on a Monday morning. This was something I used to do manually, or not do, or get my practice manager to do. Checking timesheets. No, sorry, sorry, Trent. Checking timesheets is no good. We don't like doing timesheets. And now Viktor does it, and I get visibility on Monday morning. It's the best. I absolutely love it.

Natalie Lennon: You probably can do that with ChatGPT or Claude. The one thing I couldn't get as good from the Karbon MCP was the workflow statuses. It was a bit more limited in what it could do with workflow statuses. And from memory, when I had it push from Claude into Slack, it didn't notify me properly and the format was not good. Whereas Viktor gives me a really good timesheet review. I've had another accountant say they've asked Viktor to be quite savage: okay, this team member this week looks like they were bludging, they only did this, this and this, cross-check what they did here, their productivity's dropped. So you can get these tools to really help you drill down on timesheet data, which you probably wouldn't really do on a regular basis before.

Natalie Lennon: Timesheet review, workflow statuses. The other thing I've set up on Viktor for Karbon is the practice review. It goes into Karbon. I thought I had a screenshot. Sorry, I changed these just before we logged on. Oh yeah, there it is. So we've got the timesheet review and the workflow statuses. It goes into Karbon, pulls out the workflow statuses, tells you what's overdue, and tells me by team member and work type, or tells the team. So it saves them going into Karbon and cross-checking all their work. It gives them a report we can all see, and I guess it helps us all be accountable for what's in Karbon and saves us having to go and drill down.

Natalie Lennon: And then the other one is our audits of Karbon clients. I get it to go in and catch inactive clients we haven't spoken to in a period of time, and it gives us a list, because in Karbon we put ZZ at the front when they leave or they're inactive, so they don't come up in a search. That was something my practice manager was doing manually before, whereas now it's all done automatically by Viktor. And I think you could probably do that with the Karbon MCP, but I'm not 100% sure. Actually, something else we're doing with Karbon and Viktor, which I didn't mention, and which the MCP does as well, but there are a few different nuances to it: we can use the MCP in Claude to set up work in Karbon, but you can't set it up from a template.

Natalie Lennon: Whereas with Viktor, you can set up work from a template. You can give it a whole long list and go, hey Viktor, here's a list of clients we want to do end-of-year payroll for, for 2026. Can you set up the work from the template and assign it to the team members? Done. That stuff would have taken my practice manager a long time before, but now it's automatic. So, a summary of what we've talked about: balance sheet, workpapers, completed work plus budgets, branded Xero health checks. I also use it to help me with super accruals and payment reconciliations in Xero. That's another good one. TPARs, like I said. Meeting transcripts. I think I've jumped around some of this stuff, but you can see I use it every day, all day, every minute of the day.

Natalie Lennon: So here are some other examples. And someone asked in the chat, can we trial this or is it a yearly sub? No, it's a monthly sub. If you use my link, which I'll share at the end, you get $50 credit, so you can use it for free to start and test it all out. So by all means, test it out, and test it against the other apps, and let me know how you go. Another thing I've done: we used to have a Typeform on our website to collect our new client leads, and I thought, I don't want to pay for Typeform anymore. I wonder if Viktor can do this. So Viktor went in and recreated the form, and also automated the whole thing. They fill out the form, we get a notification in Slack, the contact goes into Karbon, the contact goes into Mailchimp.

Natalie Lennon: Viktor also cross-checks the ABN with the ABR and their website, and then sends an email to my practice manager that says, this person filled in our form, I've checked their ABN, the business name matches, they're not registered for GST, or they are, and here's a little bit about them from their website. So there are little things like this that are quite simple to set up but can be big wins. We were paying, I don't know, 40 bucks a month for Typeform. The daily brief goes through my calendar. Like a crazy person, I like to colour-code my calendar by category, and I was doing it manually. I tried to automate it with Zapier and Power Automate, and it did some weird thing where it cancelled all my meetings, and it was not a fun time. So then I thought, okay, let me see if Viktor can do this. Now every Monday morning he goes through my calendar and updates all the colour coding for me, which my little OCD brain just loves.

Natalie Lennon: It also goes through my emails, drafts follow-ups for emails I've sent that haven't had a response, and pings me in Slack to let me know. Every Thursday I get it to research different blog post topics for me. It'll draft LinkedIn posts. It'll do a weekly news watch to give me different news articles that might be of interest, things I want to talk or write about. HR letters, not letters, contracts. I might put in a previous contract from a previous employee and go, here are the employee's details, do that up for me. Finding a file that someone shared in Slack or in email, because I've got multiple businesses and podcasts and all the things. So it acts like my little assistant that helps me organise my life. Oh yes, here we go. These are the slides I was adjusting earlier, and I've put them in the wrong spot. So we already went through that.

Natalie Lennon: Mistakes. How are we going for time? Okay, we're going all right. We're nearly there. Mistakes. All right, so what I got wrong. When we're testing out stuff with the team in the team environment, we don't want to just hit go first. I set up this Karbon checker and it was wrong, and the team was like, Nat, this is not working. And then I look bad. So do test them first in your DMs with Viktor before rolling them out into the channels. Always spot-check the output. AI is amazing, but as we know, it hallucinates and makes mistakes. We've got to check everything, okay? Don't just take the figures as gospel. Do a few spot checks and make sure it's right. Treat it like a new employee. It needs to learn. It's not going to be running on the first day.

Natalie Lennon: Don't ask it any technical tax questions. Leave that for your LawCyborg or your Tax Nuggets. Don't use it for any legislation-type issues. What I would do differently: probably start with a couple of tools and one channel, and expand once the team's comfortable. Like all AI, don't go hell for leather in the beginning. Like I said, don't give it the keys to send. Just make sure it drafts, and you've got to authorise everything. And then once you're comfortable with certain things, you can let it go a bit rogue. And then the credits. Like I said before, you've got to watch the credits. I think someone just asked, as a guide, what kind of credits do you use each month? I'm not entirely sure, I'd have to have a look at that.

Natalie Lennon: Okay, so where to start? What to hand over? Pulling reports from Xero, give that a go. Checklists, health checks, TPAR reconciliations. I know it's not August, so there are no TPARs to do, but maybe you've got some old ones. ABN and GST checks against the ABR. Where jobs are at in Karbon, by person, by deadline. Drafting emails and summaries. Anything you do the same way every week, like I said. And keep human: tax technical judgment, final review, anything going to a client, payroll and employee data, and the client relationship. As always, AI is just to complement us, not to replace us. So your first week: pick one. What time is it? It's 5:30. One boring, repeatable job, something you can check in five minutes. Connect only the tools it needs. Ask it one request, check it, correct it. Once it's right, then give it its own team channel.

Natalie Lennon: Viktor also made these slides and the screenshots. I said, Viktor, I'm preparing a presentation for accountants on AI and Viktor. Can you please mock up some screenshots of things we've done so we can put them in the slides? I went to sit down and do the slides myself, and I went, what am I doing? Why don't I use AI for this? Alrighty, so thank you very much. I think we're good, Trent. That was 5:35. You said 35 minutes. I may have talked a bit fast, but we're on the dot.

Trent McLaren: You crushed it. There's questions. There's questions. So you're not off the hook yet.

Natalie Lennon: No, I know, and I'm just going to find this Viktor link too.

Trent McLaren: Lots of people are engaging and asking questions. When I read some of the questions, and this is what you and I have been chatting about as well, there are tools that do versions of these things. I think the important part is that you find the thing that works for you. You and the crew have found Viktor, and that's working really well for you guys, and that's awesome. It doesn't mean you can't do it in Copilot or Gemini or Claude, or whatever it may be. And these things are changing so quickly that it's up to you whether you're going to hop around: oh, I can do this here, now I'm going to go and do that. But I think you're going to get a lot more benefit out of the depth of staying in something. It's not too different from when everyone made a decision to pick a ledger or a practice management system. They all go in waves, right? It's good, it's great, it could be better, and you stick with it, and eventually you've built all your habits and workflows around those things.

Trent McLaren: Everything in AI is very much like that, except it's all changing so quickly. There's a Q&A question from Tom. Tom's question was: when managing your inbox, can Viktor open attachments and work out where to send them based on a set of rules, even splitting PDFs to do so, multi-entity clients with one accounts inbox, etc.?

Natalie Lennon: All of the above. All of the above.

Trent McLaren: Good. There you go, Tom. That was the easiest answer ever. I'll scroll back up. Did we get to Tarun's question? Can it look at an activity statement and reconcile it to the accrual balance sheet GST account, where there's GST in debtors and creditors, and the opening balance was subject to bad bookkeeping? Can Viktor do all of that?

Natalie Lennon: It probably could. It's on my list to automate the BAS, but anyone who's tried to automate the BAS knows how difficult it is. And because of the payroll portion, you can only get it to do half. So you've got to choose which is the easiest way. I've even tried going, okay, here's our GST workpaper template, and tried to get it to do it, but it just doesn't. I haven't found any AI to do it yet without a lot of errors. But I'm open to it if anybody works it out. I have had a few people reach out with some forms of that for Automation Playground. I'm trying not to plug Automation Playground, but this is where all the stuff's coming from at the moment: people popping up with tools and things they're building. But yeah, the BAS piece is a real beast.

Trent McLaren: Nat, you've given us 40 minutes of your time. You're allowed to talk about Automation Playground if you'd like to.

Natalie Lennon: And I don't want to. I don't like being a salesperson.

Trent McLaren: Yeah, I know, but I am a salesperson. I know the only way to do that is to talk about it. Do you want to tell us what it is?

Natalie Lennon: Well, I'm just about to share a video on it as well. Basically, Automation Playground is built for accountants and bookkeepers by accountants in practice. We found there's so much AI learning online, but apart from the AI summit, there aren't a lot of people actually showing practically what they're doing, like I said before. So we've built an online platform with a library, which has been a bit of a mission in itself, because AI changes and we've had to scrap a whole lot of the videos already. So a library on AI basics, and then we're showing things like this, what we're doing in our firms. There's going to be a community connected to that, so a safe place to really share what we're doing and learn from others. We're getting other accountants and bookkeepers to collaborate and show us what they're doing, along with tools and resources.

Trent McLaren: Yeah, yeah, please do. Put it in the chat.

Natalie Lennon: Yeah, it's a monthly subscription. If you want to find out about the pricing and everything, that will be happening next week. You can scan the QR code, but I'll put the link in here as well. We're very excited to launch it. Oh, and we're also planning on doing live sessions. We're still working out what those look like. We want to get everyone into the platform first and work out where everybody's at, what they want to learn and what works best for them. Then in mid-January, once everyone's back from holidays, we're going to kick off with some live sessions and live training.

Trent McLaren: Good. Well, thank you for answering those questions. I'm going to throw a few more questions at you because your audience demands it.

Natalie Lennon: Let me stop sharing.

Trent McLaren: There's a question here around cost and credit usage and all that stuff. We had a great session on this earlier today, everyone, if you're trying to figure out token usage and credits, and you're going to need to at some point. Matt's hilarious question is: so is this your best and cheapest employee? Is that what you're saying?

Natalie Lennon: Well, yeah, we pay about 100 bucks a month. But the caveat to that is that it could be a bit more expensive, because if you refer people, you get credits back, right? So I've referred a lot and gotten some credits back. But my attitude is, even if I wasn't referring and getting any credits, even if I was paying 500 bucks a month, I'm getting huge value. This stuff is time-saving. Sometimes we've got to not be tight about it, not be typical accountants.

Trent McLaren: Do you have a rough idea of how many credits you use on some of those tasks?

Natalie Lennon: Let me have a look, and I'll do a social post, because I think that's a good one. I don't have any idea, because I just kind of use it.

Trent McLaren: No, that's right. It's just to try to give everyone an idea of where to budget.

Natalie Lennon: Yeah, I say Viktor's AI for normies, so I don't really know. Trent probably knows exactly how many. Do you know how many credits you use in Claude each month?

Trent McLaren: Mate, I've got a Stream Deck right here that tells me I'm at 43% of my usage, and it runs out on Monday night next week. So I am lasered in on how much I use. I'm at 43% in two days. I've blitzed through that already. I know I spend about $300 to $400 a month, but totally worth it. Any other questions? If not, I'm going to ask everyone to start hitting that react button and showing Natalie the love she deserves, because that was awesome. Very good, very practical. You hit my practical brief, which is to inform people of good things they absolutely need to know.

Natalie Lennon: Oh, thanks, guys. Thanks, guys, and thank you for turning up at 5pm on a Wednesday afternoon.

Trent McLaren: Jarrod had a question on... Hey, I told you people turn out for the good. A question on information and data security setup and the ability to lock down. Is it all locked down to start with? Did you have to gradually let those permissions relax a little bit as you go?

Natalie Lennon: Look, the Xero ones I've left on, because the last thing we want is the AI doing something it shouldn't. There are certain things, for example drafting emails. Before, it used to say, here's the draft, are you okay for me to draft it in your Outlook? And I've said, I don't need to authorise that anymore. When I ask you to draft an email, just draft it, and I'll review it in Outlook. And you can either update the settings within the Viktor interface, or you can just ask him in the chat to do it.

Trent McLaren: Cool, good. Thank you. Like it. Sam, are you worried that we could become too reliant on Viktor? If the price changes or something happens, we might be stuck doing work that we can't complete without hiring new employees. You can answer that question, but I have opinions on that as well.

Natalie Lennon: I think yes, but it's the same as hiring a person, isn't it? You hire a person, you train them, they leave, and then you...

Trent McLaren: What's the alternative? It's the same as every bloody product you're using. I think you're all seeing this play out with all your accounting software, to be honest. Just as a rule of thumb, everything you pay for is most likely going to increase in fees forevermore, every year or two. So the other side of that coin is that you should also be increasing your fees every year or two to align with those things. I think that's the thing people miss, right?

Natalie Lennon: Yes. There's another question here I just saw. Stephen says, does it draft in Outlook or Karbon? Outlook. It can't draft an email in Karbon. That's one of its limitations. It would be great for it to be able to do that. And I think I saw a comment there about Karbon's AI. I think once Karbon's AI comes out, which we're eagerly awaiting, that'll be another game changer, because they've really been investing heavily in AI from what I've seen, but I don't know when. They've been talking about it for a long time now. I think it's in beta, but it hasn't come out yet, so I'm not sure.

Trent McLaren: Well, since June, but June's a long time in AI years at the moment. I've seen bits and pieces. It looks great. A lot of people are enjoying that. Tarun, your question: do you find that Viktor mimics your email voice the more you use it?

Natalie Lennon: Yeah, so it's trained. When we started working together, it was similar to when you go into Claude and build your brain and do all that stuff, but you don't actually have to do that with Viktor. He just kind of learns. You can tell him at the start and give him an overview, but yeah, he does learn. The other day I was preparing videos, I think I was doing a video on Viktor, and I thought, oh, I'm going to ask him what he thinks of me and how I work. And it was like, yeah, you're very busy, you like stuff sharp and to the point, quite casual, not professional. So it was really cool to ask him what he thought of me.

Trent McLaren: Love that. Stephen, your question was: if you're not using Slack or Teams, is this going to be a hard product to use?

Natalie Lennon: Yeah, you can't use it. It needs to live in Slack or Teams. I am actually testing out Ignition's new AI coworker, Parks, which doesn't live in Slack or Teams, and I'm working quite closely with them to make it as good as Viktor.

Trent McLaren: Yeah, I did wonder where that was going to fit into your scheme of things. Yeah, interesting.

Natalie Lennon: Yeah, yeah, yeah. Because at the moment Ignition is the only thing I can't connect. Well, actually I think I managed to connect it the other day, but before that the MCP wouldn't even connect. So that was my only missing piece of the puzzle.

Trent McLaren: Awesome. Hey, we're going to start to wrap. If anybody's got more questions, or they want to reach out to you, Nat, what's the best way for them to get on to you?

Natalie Lennon: Probably LinkedIn. Search Natalie Lennon on LinkedIn, you'll find me. Send me a DM if you've got any more questions and I will try to respond. If you're interested in seeing more of Viktor, as I said, I'm recording some Viktor videos for Automation Playground, so we're going to be sharing quite a few in there. Would love to see you in there. Yeah, thanks everybody. Thanks for making me feel good this afternoon. Someone's asked for a code. Hold on, I did share it before. Let me just share it one more time for you before we close it off. Actually, I can't find it right now. Just send me a DM and I'll give you the link. Anyway, that's all right. We will sort that out. Thanks, Trent. Thanks, everybody. Have a good night. Bye.

Trent McLaren: Thanks, Nat. We appreciate you. Appreciate it. Thanks, everyone.