Transcript 0:00 [upbeat music] The Start Your Own Business podcast. Hear inspiring startup stories and get reliable expert advice on how to start your business and get off to the best possible start. 0:14 Brought to you by Startup Donut, helping small businesses succeed. Hi there, thanks for joining us. You're listening to the Start Your Own Business podcast. My name's Fiona and I'm your host. 0:31 AI, or artificial intelligence, is increasingly becoming part of our everyday life. Its rise to prominence has become more rapid in recent years, and resistance seems futile. 0:40 Although concerns have been raised over the implications of AI, it's already being widely used, providing significant benefits to users, many small business owners among them. 0:49 Potentially, AI can provide significant savings in terms of effort, time, and cost to new and established small businesses, enabling you to be more efficient, successful, and profitable. 0:59 In this episode, we speak to AI expert Phil Hobden, head of business development at Sage, the multinational software provider. 1:07 Phil will be sharing his considerable insights into the fascinating world of AI and offering tips on how AI can be used to save time, effort, and money when you're starting or running your own business. 1:18 This series of the Start Your Own Business podcast is brought to you by Startup Donut in association with Sage. Before we meet Phil, let's find out more about Sage. [upbeat music] To be honest, work had been a slog. 1:32 I'm a designer, but some nights I was chasing invoices like my full-time job. Now Sage Copilot chases my invoices for me. I just need to click a button and it's done, and I get paid one week faster. 1:44 Sage Copilot is built with over 40 years of experience helping real businesses like Sophia at Solver Design. That's more than artificial intelligence, that's authentic intelligence. Search Sage Copilot to learn more. 1:56 Working with 21 customers, each round payments came in seven days faster with Sage Copilot. T's and C's apply. [upbeat music] So it's now time to meet our guest today, Phil Hobden. Hi there, Phil. Thanks for joining us. 2:09 Hey. Hello. How are you? I'm very well, thank you. So to get to the kind of nitty-gritty of things, in basic terms, what is artificial intelligence? Oh [laughs]. 2:21 You start, you start with a fantastically c- a fantastically large question. 2:27 [laughs] Ultimately, it is a, it is a computer brain that can do a series of tasks quicker, faster, and in theory, better than you can. 2:39 So the, the best way to kind of put this in a, in an example, right, if I wanted to, to, to read a 35-page report- Mm-hmm... it's gonna take me, let's say it's gonna take me an hour, right? 2:51 Just broadly it's gonna take me an hour. And out of that, I want to write four or five bullet points of what the key takeaways are so I can then present them back. Yeah. 2:59 So you're probably looking at a good hour and a half's worth of work. Now, if you put that into a, a large language model, an LLM, or something like ChatGPT as it's broadly known, right? 3:08 ChatGPT will do that in about half a second, and that's artificial intelligence. It's a bazillion brains working on your task. 3:20 R- you know that old thing that they used to say, like, you know, if you put a load of monkeys in a room and gave them typewriters, in 350 years they'd write Shakespeare. 3:29 That's artificial intelligence, but the monkeys are really smart- Yeah... and they can do it in seconds rather than decades or years. Yeah. So how is that different to automation? 3:40 You know, lots of, you know, packages will have automated pro- you know, processing. What's the difference between the two? 3:45 Yeah, that's a really good question, and I think for me it's automation you, you kind of have to tell it all the steps of what to do. Mm. 3:52 So to, so if you think of like old style automation, if, if this then that, um- Yeah... which is kind of like a good basic way of doing it. 4:01 So if this happens, then you do this, and then you do that, and then you do this, and it'll go through each step. Yeah. The automation will go through each step of that process. 4:09 With artificial intelligence, you just kind of let it know the outcome you want to get and your start point, and it will probably do all, you can make it do all that for you. So w- with LLMs, that's pretty simple, right? 4:21 I want to do X or I want to do Y. But when you get into kind of the world of generative AI, which, which we'll talk about later, which is like the more kind of automated AI, 4:32 that will automate whole processes and whole tasks. So you know, for example, if you're a, you're a, a business, you know, I get... I, I complete a job. Well, on that completion of job, it automatically sends an invoice. 4:44 When that gets paid, um, I'll automatically reconcile it against my bank account. If it doesn't get paid, I'll, um, I'll then tell it that it needs to go out and, or it'll then know to go out and send a chase letter. 4:56 If that chase letter isn't paid, it then does X, Y, and Z. So it will take that whole thing and do it for you, and it will just run in the background at all times. 5:05 So there are subtle differ- I think automation you almost have to create it, which whereas with, with AI, it will create it for you as long, and here's the, the huge caveat on this, as long as you tell it the right thing to do. 5:17 At the moment, the AI is only as, to a point, only as smart as, as the instructions you give it. 5:23 And as I found out yesterday, very real-life example, I'm, I was trying to get it to resize a cover of a book for me, and I think by the end of an hour and a half of telling the AI what to do and, and, and seeing it create the most wonderful, weird combinations of what I was trying to do, I was like, "I could have actually done this myself quicker." 5:42 Um, so there is a little bit of that where sometimes, you know, you can, you can ask it to do something, um, and it won't be as efficient as you. But it's all about the prompting. Yeah. I was gonna say that. 5:56 It is about, you know, learning to give it the kind of instructions and the directions that you want it to take, and that can be a bit of a learning process in itself. It can. 6:08 I mean, it's like, it's like anything, right? 6:09 Like a good manager or a good boss in a business will have a clear strategy, um, and guidelines and structure on how to execute that strategy and what the marks of what good and success looks like, right? 6:21 And kind of in a way, that's exactly AI. You need to give it really clear instructions of what you want it to do, and it will execute that. The danger- Mm... is it will sometimes, it will sometimes, 6:34 it will sometimes make stuff up. And this is- [laughs] This is the problem with AI, right? Sometimes it will, it will make stuff up. 6:42 And so sometimes you'll look at AI and you'll go, certainly with like, um, generative AI, um, like large language models, you'll look at it and you'll go, "Well, actually, that doesn't look quite right to me." 6:54 Um- Yeah... and that's the danger, right? Like, AI can... It's getting smarter every day. Like, the old law of microprocessors was X amount of years, the processing speed would double, right? Yeah. 7:07 Um, I can't remember whose law it is, but, so, and that's been pretty true for the history of computers. Every X amount of years, the processing speed doubles. 7:15 Well, if you look at AI where it was six months ago, it's 10, 50-fold where it is now. And the speed of that is getting even faster and faster and faster and faster and faster, right? 7:27 So, so you know, it's quite scary how, how quick it's moving, but it- the more it- the faster it moves, the better it gets, the better the use cases are. Yeah. And what I mean by that is, you know, people are, um, 7:43 people are gonna move away from using AI just to create funny images of their, their partners on, on the back of a shark in the ocean. [laughs] Speaking from experience. As fun as that is. As fun as... 7:56 And, and it is fun, right? Like- [laughs]... like, you know, um, make me, make me one of the crew of Oceans 11, um, alongside the crew, uh, with Brad Pitt, and you can put yourself there and go, "Hey, look, cool." 8:08 Um, all that's fun, but, you know, there are serious applications for AI. Yeah. 8:13 In terms of, in terms of small business and, and kind of, you know, the ability to automate, create, backup and improve what you do is huge. 8:27 Like, genuinely, like gr- like, I think AI is one of the biggest revolutions in technology since the internet. 8:35 Um, and we all know, uh, yeah, ultimately the internet started off as a relatively silly tool where, you know, you, you, it wasn't anything particularly serious, 8:47 and then it grew into something that, that basically controls the world. And if the internet got turned off tomorrow, the world would potentially cease to, cease to function as it is now, right? 8:58 Um, I think AI is, is that level of, of, of potential transformation. We're going through the, kind of like the, the child phase of AI at the moment. Yeah. 9:06 Albeit a super smart child that can undo all the child gates and make his way out the house. Yes. 9:13 But we're not at the Terminator stage yet, so- [laughs] So we're, we're a little bit far away from someone coming down the streets, um, like big Austrian guy telling you he'll be back and stuff. 9:22 Um, and I'm hoping we won't get to that. Exactly. I mean, I was thinking, you know, just about every kind of tool or app that you use these days will tell you it's powered by AI. 9:32 Um, so it seems that it's, it's quite common. Is that a kind of fair sort of summary of things, that AI is in most of the sort of tools and devices we're using these days? In one form or another. AI has been in stuff for 9:47 five, six, seven, eight, nine years, potentially even longer, right? In one form or another. AI has now become the buzzword. 9:54 Like, I work in the accountancy tech market, um, and cloud was the buzzword for a very long period of time, right? Like, oh, you must be on cloud. Cloud's the new way forward. 10:04 The truth is only about 30% of accountancy firms, 20, 30% of accountancy firms are digital, are cloud native, and so even though that has been a buzzword for many, many decades, it's still not the, the norm. 10:18 Um, AI is a buzzword. AI is a sales tool now, like cloud was a sales tool. Everything has AI in it. But everything, everything has AI in it, but generally that AI is just a version of ChatGPT or a large language model. 10:36 It's, you know, we have AI. Well, what do you mean by that? Oh, well, you know, we've got an AI chatbot. Mm. You know, that's, it, it's kind of like the base level of AI. 10:46 So, so it, it, it's that arms race where people know that they need AI, or, sorry, they're told that they need AI, but so they then go out and everyone puts AI in everything, AI. 10:57 Um, everyone goes out there and, um, buys these products expecting them to do the world. But actually, really, the functionality is really basic. And I think, 11:07 uh, whilst I love technology and I love AI, there's a fundamental flaw with people's thinking. The fundamental flaw is AI is only as good as the data that you have, right? 11:20 Going back to the LLM point, if you feed it rubbish, it will quite happily produce you rubbish. Yeah. So if you're using AI in accounting software, for example, fantastic. 11:31 You know, we at Sage have some amazing practical uses for AI for small business in our accounting software, and some great stuff coming soon as well. But if your underlying data, your bookkeeping isn't good- Yeah... 11:46 the data will, the AI will only take you so far. Yeah. 11:50 So like, like the foundations of a house, they have to be strong before you can build the bathroom, like, before you can build the bedrooms, before you can build the roof. 12:00 If the foundations aren't strong, then, you know, you will never get the full benefit out of it. 12:06 And that is the mistake that businesses make, is they, they chase the hype- Rather than chasing, rather than realizing that actually, you know... Do you know what the best analogy for me is? And I, I love this analogy. 12:21 You know, you go to a car park, you go to McDonald's on a n- 10 o'clock on a Friday night, right? So you've gone out for a beer. You've gone, "Oh, fancy a, a cheeky McDonald's on the way back." 12:31 And you go there, and in the car park are tho- all those teenagers with their souped-up Fiat Pandas and Volkswagen Polos that have these amazing spoilers on and these massive exhausts and side skirts on the car. 12:46 But it's still only got a 1.2 liter engine that's never really been looked after, that's probably running low on oil and, you know- Yeah... if you, if you put it in a race, it's not gonna go anywhere. 12:54 It looks great, but fundamentally underneath it, it's a rubbish car with a rubbish engine that doesn't get looked after 'cause all the money gets spent on the window stickers and the mods and everything else. 13:06 That's the risk with AI, is that you're effectively putting a spoiler on your business when actually it, it needs oil. So what... That said, what benefits can AI bring if it's implemented properly in your business? 13:21 Benefits are huge, right? Like, like from a, from a financial perspective, it can automate a load of the processes. It can take away a lot of that, a lot of that, a load of processes. From a marketing perspective, right? 13:32 Just i- just a base tool that every business needs, whether you're a plumber or a media company or a podcaster, right? Marketing with ChatGPT is amazing. Like, if you teach it how you talk, and that's really- Yeah... 13:48 important, by the way. Yeah. You know, if you feed it a load of stuff that you've done before or the voice you want it to have, it will then produce content in your voice. 13:56 Caveat, there are little things that AI does that gives away that it's AI. So you have to make sure- Yeah... you train that out of it. 14:03 And then you have to, you know, put a human element on and go from read it and don't just put it out there. 14:09 But if you wanted to write LinkedIn posts or social posts about your business, you know, you can kinda say, you can kinda go to AI and go, "Hey, look, I'm a plumber. 14:19 What are the five biggest risks to properties in the summer?" Mm. And it'll come up with a list. "Right. Now create me an article for LinkedIn of around 500 words focused on that in my tone." And it'll do it. 14:35 "Now create me an image about this post basic..." Now, I don't wanna put marketeers out of business, but if you're a plumber, 14:44 you're not gonna be paying thousands of pounds to a marketing company to produce this stuff for you, right? But actually, with an LLM subscription, £20 a month, with ChatGPT or, or, or, or Copilot or whoever, 14:56 you've got a marketeer effectively that can create that content for you. Interesting tip. Probably most of those marketeers are doing the same thing that you're doing [laughs] and then polishing it- Mm... 15:06 up a bit better, right? So just from a marketing perspective, that's fantastic. From a sales perspective, you can do data and, and kinda analytics on stuff. 15:15 So, you know, if I'm a, if I'm a podcaster, you know, or say, say, I don't know, say I'm a... Whatever it is, you can use it to use to get analytics about the industry that you're doing- Yeah... and, and pull that data. 15:27 Um, and so on and so on and so on. So there's practical implications to the way you work and how you market and promote and, and do your business. 15:35 There's practical implications to the, the finance and, and all of that side of your business if you've got the right technology, the right software that, that has that bed embedded in. 15:45 You know, there's, there's, oh, there's so much positives about AI and, and kind of how it can transform a business. Mm. It, it's how... It's one of those kind of... 15:55 It's kind of one of those limitless tools that's down to how you want to use it. Now, that's a struggle, right? My dad was a farmer. 16:05 If I said to my dad, "Hey, you should be using AI," my dad would be like, "What's a computer?" Um, so you know, there is that hurdle to get over. 16:14 But if you start to le- And there's loads of great places you can learn about this stuff. Like, there's so many good YouTube videos and so many good practical places. 16:23 Like, if you're on LinkedIn, there's LinkedIn Learning that has loads of great stuff about AI, and you can start to learn the basics on what to do. Yeah. There's a bazillion, I mean, literally a bazillion videos on, on, 16:35 on YouTube talking about LLMs and how they work. Yeah. Um, so you can train yourself to, to, to work with that, and then you can take some of that heavy lifting for you. Yeah. AI is not gonna replace your job. 16:53 There's that great saying, I can't remember who said it, but yeah, AI will replace... AI will not replace jobs. It will replace the jobs of people that are not using AI. Yeah. Yeah. 17:01 So if you, if you're, if you're using, utilizing AI and utilizing it in your business and your company, you stand a... You know, that's a good protection method for you as a small business, as a... And like, 17:17 as you can tell, I'm passionate. I really love technology. Um, I would urge everyone to be out there just trying these things and just having a play. 17:24 And e- even if it starts with creating a photograph of your mate on the back of a shark, whatever that is, right, whatever that starting point is. 17:33 Go out there and play with it because you'll quickly learn and see how you can adapt it to your business and what you do. Yeah. Yeah, absolutely. Just to... I think it might just be worth rewinding back. 17:44 You've talked a couple of, um, terms there that listeners might not be familiar with. So you've talked about LLM. Can you just explain for listeners who might not be aware of the kind of jargon, what that is? Yeah. 17:56 So lang- large language model. So that is your ChatGPTs, your, your Geminis, your Copilot. And what that does is it takes... It doesn't necessarily create its own content. It takes content that's been fed into it, 18:10 and then- Will then recreate that content for you. Now, what you can do with LLMs and ask it to, to, to... You, you can say, "Hey, here's a website. 18:22 Can you go here and, and kind of make a, a version of this content for me?" But every time you feed something or create something, it then builds in the background of this, this la- large language model. Yeah. 18:33 Um, they're getting smarter, and smarter, and smarter. At the f- at the beginning, you know, it was, they were trained. It was a bit, bit odd at the beginning. 18:40 They were trained by, like, companies feeding it loads of data. Yeah. So, like, feeding it books and, and copyrighted material, and there's a big hoo-ha about kind of- Yeah, yeah... all of that going in. 18:51 But, but essentially, a large language model is your ChatGPTs, it's your Copilots, it's your Gemini. It's where you go in and you make a prompt and you say, "Could you do this?" 19:00 And based on everything it's got, all of its experience, what you've told it, what, what it's seen from the internet, what other people have searched, it will then create content based on what you want, based on- Mm... 19:10 what other people have told it and shown it in the past. So then moving on from that, another term that we- you've talked about is generative AI. How does that work and how is that different? 19:21 It's kind of, uh, generally, it's, it's kind of a, a evolution of, of LLM, right? So, it, it's, it's kind of LLM plus. 19:31 So generative means it produces something new and not just kind of predicts a, a label or find patterns. 19:37 Like, so it'll use kind of advanced models, so often large language models or, or other stuff that's been trained on huge volumes of data. 19:46 Um, and you know, so it's kind of, kind of large language models and generative AI are kind of the same thing, but it's just kind of a little bit more kind of, um, elevated version of it. Yeah. Pivot up version. 20:01 And, and, and these... Yeah. And these can produce text, images, audio, video, code. Like, you, you asked me about practical implications. Here's a fantastic practical implication, right? 20:10 I am rubbish at Excel, which is ironic- Yeah... 'cause I work for accountants, but I'm rubbish at Excel. Um, so if I wanna do something on Excel, I can, I never know the code, the way to do it. 20:21 If you go into, uh, uh, LLM and go, or ChatGPT and go, "I want to do X, Y, and Z on a spreadsheet. Can you give me the formula?" Yeah. 20:30 It'll give you the exact formula that you need to then take that and then put that in there. I mean, that is an absolute game changer. Um, and, and that for me, you know, it's that kind of stuff that these kind of, um, 20:46 large language modules, generative AI, that they do so well. Those little tasks that you could, could spend hours doing. Um, there's a really... 20:53 I love that and it's, like, such a great practical application of that technology. Yeah, yeah, yeah. So what would you say the main advantages of them are? Oh. Time-saving is huge. Mm. Um, creativity is huge, right? Um, 21:11 research and data analytics is huge. So, again, a practical example. Yesterday, I was trying to create a list of data from a variety of sources, um, for award winners in our industry. 21:23 I wanted to kind of pull probably 20 webpages together into one document, um, so I can then do some further analytics. So I just pointed it in the pa- in the, in the... I said, "Hey, here's 10 links. 21:35 Please take, scrape all the data from that, put it in an Excel spreadsheet in this format for me." Yeah. Probably an hour and a half, two hours worth of work for me, five minutes on a, on, on ChatGPT. 21:48 So that kind of analytical work, um, if you're a student and you're doing an analytical produ- project, and you've got to pull from a variety of different sources of data and then analyze that data, you can use an LLM, a, you know, a ChatGPT to do that. 22:03 Caveat is you, you know, does, is that your own work? Is it someone else's work? Who knows, right? That's not, that's not my decision to make. But, but ultimately, I think if you're using all the tools at, 22:17 in play to be able to then create a final product, you're still instructing that tool what to do, so therefore it is your work, would be my law lu... 22:27 But no one please use that in a court of law or in a tribunal- [laughs] Yes... for A-level exams, right? Yes. Having, having- Like, there is a very fine line on that. Yes, absolutely. 22:36 Having, having a recent graduate in my household, I know all about the kind of things now that are going on in universities in terms of filtering out, you know, generated content. It's... 22:44 And, and that's a whole, that's a whole p- piece of, piece of work now, right? Where the com- Yeah... there are actual companies creating tools to filter out AI and check if something is AI. 22:57 I think there's a difference between using AI to form your data set and using AI to write your end result. I think that's the, that to me is the line of it. 23:07 And, and that's also the danger when people use, use ChatGPT to do social posts. I can look through LinkedIn and I can tell you- You can smell the milage... you can. You can. [laughs] And I, I tried it once. 23:20 I tried training a, a... And this is, I think this comes under the danger of, um, AI to a point. I tried training a tool to respond to LinkedIn comments for me. Right. 23:33 And it took three responses before someone called it out and saying, "That's not you. You're using AI." Yeah. And I was like, "Wow." 23:43 And I trained it really well, and I'd kind of, I'd looked at it and I was like, "Broadly, I, I think that's okay." I don't think that's too bad. Yeah, yeah. Yeah. But someone picked it up within... 23:53 Probably the reason they picked it up is I'm dyslexic, so there probably wasn't a spelling mistake in it. It was probably something as simple as that. Um, but they picked up it was AI and I was like, okay. 24:02 And I, I deliberately did the experiment with three people that I thought, well, I think five people I was targeting that knew me quite well to see if they would pick up on it. Yeah. And they did. So danger of AI is it- 24:15 It, it can, if people know the signs, it can be, it can take away credibility pretty quick And I think it's worth bearing in mind, it's like, you know, when you get a new version of software, there's always kind of snags and things, and that, I think that's still true for some kinds of AI p- you know, you, it, on paper it's great, but always check, you know, like you say, the language. 24:38 Does it, does it sound like you? Does it feel like you? Is it, is it authentically from your business? 24:43 And that's where I think you can start to use it to generate, but always kind of make sure that there's a person behind it too. Yeah. The, you always have to have safeguards, right? 24:54 Like, always, with any technology, always have safeguards behind it. And ultimately, the, the end safeguard is you. If it's your business, are you happy for this to go out? Yeah. If you're not, that's fine. 25:08 If you are, that's fine as well. But be your own safeguard to your, your, your content and what you're putting out. Um, 25:15 and then, you know, be that final bit, you know, finger in the air and go, "Actually, does this feel, taste, smell right?" And if it does, that's great. Like, AI will, will fundamentally improve what you do. 25:28 It, it is a rabbit hole, right? Like, I can... I mean, I've got stuck on, I've lost hours to AI just to having, like, trying to get it to, to do something. Or, or having fun with it and, like, going, "Ooh." 25:39 'Cause I, I am that kind of person that when the new software comes out for my phone, I, I sit there for an hour and a half look, looking at all the new bits and going, "Ooh. Ooh, there's a new slider there," right? 25:49 Like- [laughs]... don't be that person. It's a tool. [laughs] Yeah. It's like, like, like if you're a plumber, I'll just use that example practically, right? 25:56 If you're a plumber, you have a set of tools in your tool bag, and you know what tool to use for the right situation. AI is a tool. AI does not solve your business. It will help your business. 26:06 It will support your business. Financially, you can do some fantastic stuff. If you look at kind of, um, 26:12 um, financial modeling tools, forecasting tools, like s- financial software, like the ability for that to be able to predict where your business is going and, and kind of if you carry on at this trend, you're gonna be 26:27 X amount up on last year, which means your tax liability will be Y, which means you're, uh, currently looking at your tax b- tax savings, you're not saving enough to cover that. That kind of stuff is smart, right? 26:38 That's the kind of stuff that's gonna save the backside of businesses where... 'Cause I always say, no matter how good you are in a business, um, or good at what you do, you go into business to do what you do. 26:50 So if you're a plumber, you're probably not gonna be amazing at the, the, doing the accounting part of your business. There are some people out there that will be, but you'll be an amazing plumber. 27:00 I had a film media business, and I was rubbish at the financial side, to the point where I actually had to pay an accountant more money to correct the mistakes I made than if I'd just given- Yeah... 27:12 it to him in the first place. Yeah. Yeah. Um, and I thank that accountant to this day, 'cause they probably still saved me a fortune. Yeah, yeah, absolutely. 27:21 So in terms of, um, other possible downsides, and you've already sort of alluded to this, is about the quality of the data from which the AI is drawing. 27:30 Um, is there anything else that people need to be mindful of when they're using tools? Um, I think it's that over-reliance, right? 27:37 Like, that thought that it will do everything for you, and you then start to, you then start to kind of give away responsibility a little bit and, and, like, treat AI like it's a part of your business or it's a person in your business, right? 27:51 You have to be careful to, you have to be careful to not allow it to be your business, which is l- like, you know, it's, it's a tool. Like I said, always treat these things as a tool. 28:05 If you treat them as a tool, you never rely on one, one tool for every job, right? It's... See, I think that's the biggest, always the biggest danger with, with, with AI, is not relying on it just 28:18 to do everything for you. Like, we've got, um- Yeah... in, in, in our, in Sage, uh, some of the, our A- our AI, AI products will do a load of really cool stuff for an accountant, right? 28:30 Some of our Gen, uh, some of our, um, agentic AI products, like, will do some amazing stuff for, for people. But it will show you every step of the way how it's doing it, what it's doing- Yeah... 28:40 why it's doing it, so you can then have trust in the process. And that, to me, is the best form of AI, that enables you to see the workings of how it works. Yeah. So you can then trust it. 28:54 Because if it's just doing stuff for you, there's a real danger of, you know, how's it actually doing it? How do I know that that's correct? And some people might go, "Well, I don't really care. 29:05 I don't, I, like, I trust it." 29:07 Like, my view is, and I've worked with technology a long time, I trust technology implicitly until it goes wrong, and that's, and it, and it will at some point, you know, something will always not be quite right. 29:21 Like, it might be one in a million times, but it's, it, it's that one in a million times, if you don't catch it, you know, what's the impact? 29:28 So, so I, I, I love AI that kind of supports you by showing out, showing how it works, what it does, how it- Yeah, yeah. Lifting the curtain. Yeah, 100%, right? 29:38 Like, I am the guy that goes to a magic show and wants to know... Like, I'm less interested- [laughs]... in the trick and more interested in working out how it's done. Yeah, yeah. Yeah. No, it makes sense. Um, 29:48 I was gonna say, I probably should explain, 'cause we were talking about terminology, I should probably explain agentic AI, 'cause it's a term I've used a couple of times. So, um, think of agentic AI as, like, agents. 30:01 Yeah. So they can take actions, make decisions, and pursue goals on their own, rather than just responding to a single prompt. So it's about kind of- Autonomy and an, an initiative. 30:14 So, like, it'll be goal dri- uh, goal-driven, it'll use different tools, it'll keep a memory of what it's done in the past. It- you can iterate it based on feedback or new information. 30:26 So a, a, a Generative AI model might write you an email if you ask it. An Agentic AI might notice that an email needs to be sent. 30:33 It'll draft it, it'll look up the recipient's detail and schedule it without you having to do every step of it. So, like a g- uh, I, I was trying to find a good definition for this earlier, and it's, it... I love this. 30:45 It says, "Essentially the difference is between a smart typewriter and a personal assistant that gets things done." Yeah. So think of an LLM as a smart typewriter. 30:54 Think of, uh, Agentic AI as, as an assistant that gets stuff done. Yeah. That's the, that's the magic of AI. 31:02 That's where, that's the next step of AI, and that's where o- a lot of these time savings and a lot of these benefits- Will come. Yeah... can be done. Yeah, yeah. That's the really... 31:10 That's kind of like, you know, like that's the, the, the, the young adult of the, the child at the moment that is kind of ChatGPT. 31:18 Agentic AI is gonna start doing some really interesting stuff, and these agents that automate these tasks for you. Like I said at the beginning, that, that ability to say, "The job's complete. 31:29 Right, I need to raise an invoice. I need to send the invoice. I need to check if it's been paid. If it hasn't been paid, I need to chase it. If it hasn't been received after that, I then need to refer it on." Yeah. 31:39 And all of that, that does it for you. Yeah. That is cool. Yeah, yeah. So putting this into kind of practical terms for our audience, how would someone starting their own business, how would you advise them to 31:55 kind of use AI to help them in their startup process? Um, I... So I would go onto ChatGPT and say, "H- I want to start a business as a social media consultant. 32:08 Um, what are the steps I need to take to start that business?" And it will give you all of those tools, um, that you need to... I'm, so I, I want to, I want to start a business as a social media consultant. 32:26 What do I need to do? Let's do a live experiment. I love it when we do, you do, you try something and see, and see if it works. And this i- then there you'll see the power and the danger of AI, right? [laughs] Um, 32:37 so what's come back on this, it goes, "Define your niche and offer. So choose your audience. Do you wanna work with small local businesses, e-commerces, coaches, restaurants? Build credibility faster. 32:47 Optimize your social media. This is your shop window." I'm doing a sh- you know, a, a, a short version of this. "Set up your business basics and register your business. Get a simple service agreement contract. 32:57 Choose tools for scheduling. Um, set up a simple invoice system. Find your first client, so use your personal network. Offer a free audit in exchange for a call, then upsell your services. Deliver and scale. 33:08 So tre- track metrics. Create reusable templates, reports, proposals, and content. Pro tip: don't be on every platform from day one. Master one to two that suits your audience. 33:18 If you want, I can give you a 90-day launch plan so you can go from zero to your first paying client easily." So I'm gonna put, "Yes. Give me that." And now it's gonna create me 33:28 a 90-day plan on how to launch my business week by week. So phase one, phase two, phase three. Each phase has a goal, and by each week it has three activities. 33:41 So for example, if I look at phase two, client hunting, week eight, um, ask early clients for testimonials and permission to share results. Bef- post a before/after case study to show your work in action. So what? 33:54 So that took us a minute for- 10 seconds. Yeah, yeah... two, two minute, two minutes from you saying, "What do you need to do?" To me saying, "I wanna be a social media consultant." I wrote it. Yeah. 34:03 I put it into, um, my LLM that I use, and it spurted out a basic kind of five-step plan. Yeah. 34:12 And then it's given me a much more detailed 12-week kind of action points, and it's even kind of said like, you know, um, "If you can m- if you want, I can make you a one-page version of this plan." 34:25 And I'm gonna put, "Yes. Do that." So now I have a plan on a page. Yeah. So now I can take that plan on a page to a bank and say, "Hey, I wanna borrow..." Or a grant maybe. "I wanna borrow 10 grand grant cash." 34:37 In fact, maybe, maybe I could do that and go, "Are there any grants in the UK that could support me?" And now it's gonna kind of do that research for me. It's searching. 34:53 It's, it's kind of, it, it's come up with a tailored grant system. So, um, it's kind of said different places I can look. Oh, look, "Entrepreneur and Social Enterprise Support. The Prince's Trust program. 35:04 If you're between 18 and 30, you may qualify for 5K in grants, plus mentoring and coaching. New Enterprise Allowance. 35:11 For those currently unemployed, scheme can pay a weekly allowance to support you while you start your business." Um, and then, you know, social and community focus support. 35:20 Um, UN Limited focus for, uh, Foundation for Social Enterprises. Offers between 2.5 thousand and 20,000 along with mentorship network support. You see, see what I'm saying here? Like, all of this- Yeah... 35:32 I, actually I think I might go and start a job as a social media consultant off the back of this [laughs]... it looks like it's quite fun. 35:38 But all of this is, like something I've done and I don't have to pay anything for that, by the way. That's- Yeah... you can do all of that without a subscription to ChatGPT or, or Copilot. That's just on the basic model. 35:50 Um, and I've gone from you asking me how it can be useful to building a plan, looking for grants, and kind of giving me a, a checklist. 36:02 And by the way, I can say, "Produce that checklist as a PDF," and it'll produce it as a PDF. So now if I wasn't to do that, what do you reckon? How much research would that have taken? Yeah, exactly. A week? 36:13 Or how much would you have had to have paid to a small business mentor to help you get to that point? Exactly, right? And now you can pay that- Yeah... small business mentor to help you elevate that. Yeah. 36:24 So you've got a basic idea. Um, I could maybe go, like what other social media consultants are in my area? What areas do they focus on? So I can look for the white space, you know? 36:33 So maybe I can start to, to hone down what I want my niche to be. Yeah. And all of that is from one idea, one conversation, one prompt into ChatGPT. 36:45 And by the way, like I said earlier, I'm dyslexic, I'm, I'm like, I sometimes write things wrong. Um, I put probably at least two or three spelling mistakes in that initial prompt, and it just ignored them, by the way. 36:57 It just kind of knew what I was trying to say. Yeah, yeah. So even if you, you have those neurodiverse issues, it's gonna still understand majoritively what you want. Yeah. Get to the crux of what you're asking. Yeah. 37:11 So like, there you go. Like, hopefully that's demonstrated the power of what AI can do just to start, and that's just initial idea, initial concept, initial c- hey, look. Is it perfect? No. 37:27 Is it, is it the most detailed? No. But you can then build on that. But like you said- Yeah, yeah... all of that is days, weeks worth of work done in seconds live without any... 37:39 I, I had no idea you were gonna ask me that. It feels like I'm like the magician on stage. I've said, "I'm just gonna show you a trick." 37:44 And you- I've said, "Pick me a, pick me a, a town," and you've gone Bristol, and I've gone, "Brilliant. Right. Bristol. Right. Magic trick." Off we go. Yep. Off we go. But, but it's that simple, right? And, and, and 37:55 this can be used by anyone with a computer. Yeah. You do not have to be a, a, a AI scientist. You do not have to be a genius. You do not have to be someone that doesn't have dyslexia or, or whatever, right? You can... 38:10 Anyone can use these products. They are- Yes... as basic an interface as you're ever gonna get. It looks like Google. If you can use Google, you can use a g- uh, like, like, lo- LLMs. 38:21 Agentic AI, a bit more complicated, right? That's kind of a bit more down the line. But to start with, anyone can use these products regardless of what your business is. 38:32 And I think the other point that you've just perfectly demonstrated is that actually you can then go, "Well, I would like you to create me a business plan," or, "I would like you to estimate my startup costs," or, "I want you to help me with branding and logos and coming up with an image for my business." 38:47 You know, actually, you can be really very flexible in terms of what you ask, and it might not be perfect, but it gives you that starting point that you can then move on from. Yeah. 38:57 And in fact, while you said that about starting costs, I actually put that in, and it's, it's come up with a whole... So bare bone start cost, no major kit upgrades. One-off, £400. Monthly, £100. 39:08 Professional polish starts at better branding, maybe kind of a one-off of 1,500 and monthly of £200 a month. Yeah. There you go. There we go. So again- Yeah, yeah... 39:17 like, and it's, it's that kind of thing to like I will, I will quite often be sat there at work and I'll go, "Oh, I've got an idea. Oh, I'm gonna stick it in ChatG-" In fact, great example. 39:27 I was like, I was writing these posts on LinkedIn linking movies to business. I, I've kind of, I, I like... It was a bit of a Mickey take out of people that do all these kind of life coaching posts. 39:38 I was like, "Right," like I'm gonna... My f- my thing's films. I'm gonna, I'm gonna... Life lessons from films. So I started doing that. And I was like, "I can write a book on this. This could be a really fun book." 39:50 So I went into ChatGPT and I said, "Give me 20 chapter headings around this idea," and it did. And I was like, "I'm gonna write that book." Yeah. I can see that. And over a couple of... Yeah. 40:00 And over a couple of, well, couple, about 10 trips to Newcastle and back and various other bits that I do, I started fleshing it out. 40:08 Um, and then I used ChatGPT to make sure that the grammar and spelling was correct and, and like the formatting was right. 40:15 And then I used Fiverr because it's a really easy way to do it, to get someone to format it into a, a book format- Yeah... 40:23 for me, and now it launches on the 9th of September, and I only started the project on the kind of beginning of Ju- J- end of June, beginning of July. And I used... And, and it was literally taking an idea, putting it... 40:36 And I, I even said to the p- I even said to ChatGPT, "Here's the concept. Come up with the name of the book." And it gave me four or five different names. I said, "Make sure none of these have been used before." 40:47 It couldn't find it. So we came up with the idea of, um, The Popcorn MBA. You know, business lessons, uh, like business lessons from films and, and TV. And I, and I'm like, "I love that title." 40:58 It's a really cool title where it's exactly- Yeah, yeah... it does what it says on the tin. And if you search, no one else has that. Yeah. 41:04 Again, that's an LLM taking an, a kernel of an idea into something more, helping me with chapter headings and ideas for different chapters, helping me f- like correct the content and then coming up with a cool title. 41:20 So I was gonna say, and then when you needed to, you used Fiverr, which is a sort of platform for finding kind of freelance designers and, you know, writers and blah, blah, blah, to get the human side of things to enable you to reach your end goal. 41:33 Yeah. And it's a perfect mix of me, technology- Yeah... human. And people. Right? Yeah. And people. And I think that for me is where... That for me is a perfect use case of AI. Like I didn't- Yeah... 41:43 say, "Write me this book," that's cheating. What I said was, "Help, come up with some cool ideas that I could use." And it was like, "Well, here's a chapter idea." And I was like, "Oh, I love that. That's really good." 41:53 Like, like, um, business, you know, quotes from films that you shouldn't say in a business meeting was one of the chapters. [laughs] Yeah. 42:01 Um, a- and I'm like, like, I, I think one of the quotes was like, you know, you, you... In the middle of a meeting you definitely shouldn't say things like, "Say hello to my little friend." Like, you know? [laughs] Yeah. 42:12 Scarface for, for those wondering. Yeah. Um, you know, so like, like it's stuff like that and you kind of go, "Brilliant. That's great," right? And then, like you said, at the end there's a human. 42:22 And then by the way out of that I've gone back to a, a, a software to then help me create- Um, a physical copy of the book. So I've got s- like 100 physical copies, so I'm then using that. 42:35 So I'm combining the two together, and it's like- Yeah... it's that bounce between human, technology, human, technology, and then I'll market it somehow, and I'll- Yeah, yeah... put it on LinkedIn. 42:48 And again, like, so much time has been saved across that and, and that to me is, you, if you're using... 42:55 I think we've, we've, hopefully we've demonstrated a couple of times on this that forget the terminology, forget the complexity, forget the, the BS around AI. Yeah. Like, it will do what you need it to do- Yeah... 43:10 as simple, as complex as you want to make it. Yeah. Um, but if you just start simple, go to- Yeah... go to, if you use Microsoft, go to Copilot, which is embedded in. 43:22 If you use Chat- if you're not, go to ChatGPT and just, just type something in it and, you know, see what it comes up with. Because eventually you'll come up with that use case that's perfect for what you want to do. 43:34 Yeah. Yeah. Talking of Copilot, um, Sage counting software has something called Copilot built into it. Um, I wondered if you could tell us a bit about that. 43:46 Yeah, so like i- it's a lot of what we've been talking about, right? So one of the things I love about the AI that we have is it's, it's a co- Copilot is a great term, right? Cop- it is exactly that. 44:00 It is there to be your support, to be your buddy, to, to, to kind of help you along the way. So like one of the great uses of, of AI in this space is the ability to spot errors. 44:13 So, you know, like we as humans, we type stuff into systems. Our clients sometimes come in and type stuff in. Actually, if you've got like a, someone there that can kind of go, "Actually, that doesn't look quite right. 44:26 Is this correct?" That's kind of what the Copilot does in a, in a roundabout kind of way. It sits next to you and works with you to make sure that what you do is, is, is better, the data's better, the quality is better. 44:43 It doesn't do all it for you, it doesn't kind of necessarily kind of work by... You know, it's there sitting alongside you as your copilot. Exactly like in a plane, right? 44:52 So it, although the s- it has a series of tasks and bits and pieces it does. What we are developing is that agentic AI piece. Yeah. The invoice chasing piece. 45:00 You know, like all of those really cool things that are then gonna enhance that again, but with that Copilot methodology in place where it's going to always show you, talk to you, ask you 45:13 if what it's doing is right, and if, if you need more of this. Um, I think what makes our AI stand out above others in this space that are doing similar things is it is real. 45:26 Like most of the other accounting software, it's in beta still, or, or it's just a version of an LLM. Ours is kind of that little bit more, it's kind of taking it to that next level. 45:36 It's out there, it's being used, it's being used on a day-to-day basis by accountants, by small businesses. 45:43 Um, and it's helping them just streamline what they do and be faster, be slightly better, make sure the data's correct, make sure the information's correct, run some little tasks for them. 45:55 Um, the, the agentic AI stuff that's coming, that's launching soon and, you know, it's gonna be launching for the rest of this year, I mean, that just takes all of that to the next level, right? 46:02 That, that ability to create agents and run processes on your behalf, um, that should make the life of small businesses a lot easier. And the invoice chasing piece is, is I think one of the most important ones. 46:16 Yeah, absolutely. Because small businesses, you know, they need money. It's, uh, it's that old cliche, it's the lifeblood of your business. If you run out of cash, your business is stuffed. Yeah. 46:26 So if it can help you stay on top of that and keep your business solvent- The most important thing, right? It is the actual mo- ch- chasing invoices, getting you paid. 46:36 Like you can have the best business in the world, but if your cash flow's rubbish, that business is not gonna last, and we've seen that year after year. Like there was always... 46:44 You, you look at the financial crisis of 2008, I go back a little bit. It's one of the reasons I have gray hair and gray beard and no hair, right? [laughs] Like, I was, I was in banking- Likewise... at that time. 46:54 It, it was brutal. Yeah. Um, but you look at all the great businesses that just, that didn't have, that were impacted by things that weren't their fault, but because the cash flow disappeared, the business disappeared. 47:06 And that to me, like if we can solve those problems with a combination of agentic AI tools within Sage, and maybe some of our other tools like our Futury tools, which are a fantastic forecast, where I f- first started in this industry, believe it or not. 47:22 Like a fantastic tool to predict your business and predict kind of your f- your, where your business is going and everything else. 47:28 That to me is an absolute win because, 'cause genuinely like, like the ability to go, "If we make this decision, what is the impact of that decision? 47:39 If I take on one more person, how much revenue does that person need to bring in?" 47:42 'Cause people look at it sometimes, that, that kind of naivety of business, you go, "I'm gonna hire someone at £30,000, and that's the cost." Yeah. Yeah, yeah. But it's not the cost. NI, pension, 47:55 all the other- Holiday, sick- Holiday-... maternity... sick, maternity. Yeah, yeah. All of it. Ramping as well, right? If I hire a salesperson- Yeah... for 30 grand- Yeah... 48:05 that salesperson is probably not fully effective for three to six months. Yeah. So there is a drain on that business for three to six months- Yeah... 48:14 before that person starts to like, starts to kind of make that impact and cover all- Yeah... those costs. In fact, in reality, that person should be covering four times their salary and costs, right? 48:27 So what point do they get to that? That's the, that, tho- those are the kind of tools that are essential for businesses. 48:35 And I think that's where kind of, we, we look at, we look out and we look at all the flash, sparkly stuff, um, but we forget that, that ultimately it's understanding your cashflow and u- understanding the oxygen of your business. 48:50 Yeah. And, and also when to get support as well. Like, I know that there is going to be... I know December is our quietest time. We're a little bit behind. 48:58 We're probably gonna be a little cashflow negative over a three-month period. Yeah. So maybe I then need to look at some kind of support for that period to make sure I'm not eating into my reserves. Yeah. You know? 49:10 And it's enabling you to be proactive if you've- Yeah... got that kind of- But-... that long sight. We saw it with COVID, right? 49:17 How many, how many businesses had their head firmly in the sand and were like, "It's gonna be a month. It's gonna be, like, two weeks, three weeks, four weeks. Okay, it's gonna be a month, two months, three months. 49:28 Okay, now we're in trouble," right? 49:29 Like, whereas I was working in a finance company at the time, um, another tech app called, um, that I was working with, and, like, those businesses that were reactive and at the, at the beginning came to us and said, "Okay, we know that our cash runway is three, is six months." 49:46 Yeah. "If this impacts us beyond three months, we're really gonna start to feel the pinch." Yep. "What do we need to do, and where can we access support?" 49:56 Those businesses are still trading and trading better than ever today. The ones that came when the money had gone and said, "Right, I need support," they didn't last. Yeah, yeah. 50:07 So anything, any of these tools, and like, you know, talking about Sage, for example, we've got these tools that can help you do this, and I think that to me is essential, you know? Yeah. 50:19 Bells and whistles are great, but going back to my first point at the beginning, all of these things rely on fundamentals in your business. Yeah. Your business relies on the fundamental that is cashflow. 50:29 Your AI and technology r- relies on the fundamentals that are data. Get either of them wrong 50:36 and, or, or not focus enough on them and focus on the shiny thing down the road, and, and this is why what I love about what we're doing at Sage, and not to make this all about Sage, but what I'm... 50:47 We're, we're making it part of it. So if you use Sage, our AI is part of that. 50:52 It uses the data, and the logic will be if you're using the software correctly, the data will be correct, so therefore you're gonna get some really fantastic results off the back end. 51:01 We've covered tons of stuff, um- I mean, so much, right? Chat, chatted, chatted for far longer than we, we set out to do, but that's great 'cause there's millions and millions of great tips in there. 51:11 I'm gonna ask you prob- the, probably the question all of our guests dread is, can you boil that down, all of that that we've covered, into top three tips for using AI as you start and run a small business? Yeah. 51:25 My number one tip is use AI. So go out there and play around with AI and see what it can do for your business. Um, second thing is, I guess, understand where you need AI and where you don't. 51:42 Because AI doesn't need to be in everything. It needs to be a tool in your toolkit that you take out for the right job. And three, it's, all of these things is only as good as the fundamentals you build it on. 51:56 So yes, AI is great. Yes, AI is gonna revolutionize the way we do business, and yes, AI is gonna enable you to do some really cool things. 52:04 But if the fundamentals underneath that are not strong, if your data's not strong, if your financial information's not strong, then you will never get the maximum results out, out of the- Yeah... tools you're using. 52:17 And you'll, you'll end up being that, that, that line where it's like, "I'm spending so much time, and I'm not getting the results I need." 52:23 Well, you're not getting the results you need because actually you've not spent enough time in the right place. Yeah. So focus on that first. Yeah. Um, then understand where you need it. 52:34 But ultimately, you know, have a play around with it. Don't be scared by AI. Don't be scared by these tools, 'cause actually you can do some really fun, cool stuff with it. And- Yeah... 52:45 just play around and, and, and understand how it works. Brilliant. Before I let you go, Phil, where can our listeners find you online? Uh, so I am, I'm definitely on LinkedIn. 52:55 So, uh, if you just search Phil Hobden on LinkedIn, uh, my, uh, headshot will come up. Um, it's got a blue background. It's quite, you know... 53:01 It's, it's probably the only headshot in my entire business career that I haven't hated, so- [laughs]... you know, that's a, that's a plus, hence why it's on everything. Um, so you can do that. 53:08 But, uh, more importantly, right, if you're a small business and you're, you're looking at what you do, go and have a look at Sage and the suite of tools we have, because we have a suite of tools for every part of your business that, that's really crucial, right? 53:21 So go and have a look at Sage and, and what we can do. And if you want a demo of the software, reach out, like, like, talk to me. 53:28 I can put you in contact with the right people, because we've got 40-plus years of pedigree in this industry. We've been supporting SMEs since the early '80s. Our, we're a UK company. All of our stuff is done in the UK. 53:40 I, I always, I joke that we've got, like, 300 developers sat in an office in Newcastle developing an AI product, and I, I'd like to think of it, I think we missed a trick by calling it Y-A. Um- [laughs] 53:53 I'm never gonna be allowed back to Newcastle with that joke. [laughs] But I think we missed a trick not calling it that. 53:58 But, you know, so, so, you know, I, I think that, you know, we've got that pedigree of supporting British businesses for 40 years. We understand the challenges. We understand what you need. 54:07 So, you know, start there, and, you know, I'm always around if anyone needs support or, or just wants to talk about AI, because it's a cool subject to talk about. Fantastic. Thank you so much for your time today. 54:16 It's been great fun chatting to you, and really insightful, so brilliant. Thanks for speaking to us today. Thank you very much. 54:22 [upbeat music] I know I say this every time, but I've loved my chat with Phil today, and I hope you have too. 54:31 Phil's demystified some of the terminology, and he's explained and showed us with his live demo just how easy you can get started with AI. 54:38 You can create business plans, startup costings, marketing plans, just about anything you'd care to ask for. And of course there are pitfalls. You need to make sure the data you're relying on is accurate. 54:48 Your results will only ever be as good as the data you're putting in. You also need to make sure that there's a human stopgap to ensure quality control and give the information your authentic voice. 54:58 And at the end of the day, if the results aren't quite right first time round, you can tweak your prompts. You can find Phil on LinkedIn. Simply search for Phil Hobden. 55:07 We'll share links to Phil and links to handy resources from across the donuts in the summary. You can also learn more about Sage and their AI tool, Copilot. 55:15 You can find the summary notes at www.startyourownbusinesspodcast.co.uk. If you enjoyed the show, please hit follow. It will make sure you're notified about new episodes as soon as they drop. 55:28 All that remains is to thank Phil once more for showing some time to talk to us today and to share his insights and tips. Until next time, all the best. [upbeat music]