How To Use AT Without The Hype - Ian Lockwood

Hayley Baxter: There's a lot of noise around AI. Most of it sounds impressive, but very little of it feels useful. And for most small businesses, the question isn't, should I be using AI? It's what problem would it actually solve? And that's exactly what this episode is about. We're joined by Ian Lockwood, who shares how he used AI in a very specific, very practical way, not to replace people. not to chase trends, but to take the pressure off a growing business and scale without adding additional cost. This is a conversation about using AI with intent, not excitement. Let's dig in. Ian, let's get straight into it. When you look back at your journey so far, what was happening in your business or your life that led you to your what one thing? Well, I mean, There's so much buzz around AI stuff over the last few years and thought it's time we tried to do something that was meaningful with it, not just write me this or write me that or ask questions, what have you. And we were coming into our peak season and typically the sales team on the phones will be creaking at that point. And we decided we want to try and take the load off a little bit. And it was an opportunity to try some new technology, see how well it worked. And if it did, then we expected it to make a reasonable impact on the workload of inbound calls. So we had a look at the old chatbots and we created AI chatbots for the websites so that inquiries and typically asked questions and the usual kinds of stuff that the sales team would deal with could be. a large degree automated. It was a learning process for sure. Took a little bit of time, something that has to keep refining as time goes on. But yeah, ultimately it did make a significant difference. Our business is something like 20 % up year on year. We're still selling the same kinds of stuff just in a higher volume, but the core volume stayed about the same. So we've had a decent impact, I think, on that. It means we've been able to scale without any real additional costs or overheads to the business. So take me through that. process or that mindset, because I know that a lot of people are kind of worried about AI or they don't really see how it's going to interlink with their business or they feel it's going to be such a huge behemoth of a project they don't want to get started. What was your, how did you assemble your, your elite team to get to that? Yeah. The elite team of me, it was, ⁓ yeah. not that elite on that basis, but yeah, it was something where There's a lot of choice in the market. You can imagine like with most tech stuff, you know, there's, there's a million companies doing this kind of thing. And my initial kind of diagnosis was find something that isn't inordinately complex or expensive because this really is just a test. So I did have to spend some time digging through the spec sheets and pricing pages of various different websites. Classic kind of, I need a tool to do this and now I've got. 5 million options and they're all priced in different ways. So I can't compare them directly and all that kind of thing. Settle on on two or three to try. Naturally, you tend to get free trials. So enormously useful. You've got to spend the time, but at least you don't have to spend the money and kind of work my way through it from that. So I was looking for something that was easy and low cost. And there's a very big spread in this world from just tell us what your website is and we'll create a chat bot. And you can't really do much other than that all the way through to here's a massively complex flow chart system where you can for every feasible scenario anybody might ask of anything and have a whole routine and looping system and what have you. So I didn't want that because it's just too time consuming. you know, we decided to do this quite late in the day relative to the impending busy season. So I managed to find one that was a sensible cost. And essentially, yeah, you tell it what your website is, it crawls the site. You can then add your own documents to it as well. You can add FAQ type stuff. So you can have very specific responses for some quite specific questions if you need to. And then, yeah, we got that set up pretty quick. Got the team playing with it, asked them for what, are the general questions, you know, you get on the phones most regularly, kept firing that at it. And it really. that particular one that we're using, which is called Chatbase. It really is very much like using AI tools. There's no kind of builder type stuff. It's all prompt based. So you literally type out what do you want it to do? What should it do in this scenario? And it gives you some templates for that. But yeah, that was kind of the process of getting to something that we could realistically use without it being horrendously complicated. the sales team welcome the change or was there any resistance to it? Well, that's a fair question. I mean, in the main, they welcomed it. in the sense that it should have lightened their workload. mean, it didn't lighten it to the point where they were sitting around flicking elastic bands at each other in boredom, but it reduced the length of the queues of people waiting to speak to them. ⁓ But to be fair, they probably wouldn't have really noticed that because they work there, they answer the phones when the calls come in. And if it's a constant stream of calls, that's it. They don't really look at the queue length too much, I don't think. But at the management level, we could see that that was happening. The other thing that it's kind of done is try to automate a quote process. So what tends to happen is that we've got a lot of standard size, standard, whatever off the shelf products, and that's typically what people buy. But there are a fair number of people who want to buy something that's specifically configured for them. And sizing is typically the issue there. So we get quite a lot of quotes, more than I expected until I saw this in action. A ⁓ lot of people wanting, ⁓ well, I need this size or I need this. whatever. And we're able to some extent to get the chat bots to do that, gather it up, pop it out as an email to the sales team. So they're then able to deal with that when they get around to it or they've got it all written out in front of them. hopefully, in most cases, the chat bot has coaxed out of the person the details we actually need to get them a quote and and so they they generally are able to work off of that and get back to this this person whoever needs quote without further interaction which is quite a time saver compared to having to get all of that over the phone and so yeah it's it's not bad it's not perfect either but I think I don't see any concerns from the sales team with having introduced the chat bot side of it you know it certainly hasn't reduced interactions to the point where We're looking at the team size thinking we could do without some of these people. It certainly hasn't hasn't got to that level. So give me an idea if you will do it. So how many have you got in the in the call center or the sales team at the moment and roughly how many calls are they kind of taking and have you got any ideas on how much or how little it reduced at all? Well, I'm just counting up. think we've got 10 or 11 now on the phones and we think it's probably had about a 20 % impact on calls. But I think the other aspect of that is, and this is much harder to quantify, I don't have any data on it, but you would hope it has to an extent reduced the more basic or feasibly time-wasting calls, because if it's straightforward and somebody's tried the chatbot, they will get the answer. So people who have product questions, a lot of the calls are ultimately, ⁓ it's on the website, mate. So if people do, and obviously not everybody's going to use the chat bot, but those that do use the chat bot, if it's that kind of question, they will generally get a very sensible answer and the chat bot's pretty damn helpful at pointing to the right products and stuff. So I think it's probably changed a little bit the nature of, or the amount of those types of calls that we get, which is great because we want the sales team to be really busy on the phones, but you want them to be busy actually handling sales calls, not the classic sort of, ⁓ do you do this in white? Yes. Why don't you just search for it on the website? But you know. That's the nature of the beast. But the one big thing that we have not yet tackled, which is really the single biggest thing, but more technically complex, is where's my stuff? Because that is your absolute classic call. I've ordered this. When's the delivery coming? The delivery was supposed to be today. Why isn't it? And the reason that that's more challenging is, as you would expect, there's a lot more... integration with third party systems to be able to pull that kind of stuff in and not mislead people. And that was our pause for a second, because you might want to just edit this in as a thing. But so, yeah, when we're training the AIs, the classic thing that it does is massively over promise. And if you don't think in advance to put the guardrails on there, it starts promising people stuff that could never feasibly happen. Like I noticed one time somebody had come on and said, ⁓ it wasn't delivered today. And this is eight o'clock at night, whatever wasn't delivered today. And the chatbot told him, ⁓ don't worry. I'll rebook that order for 10 AM tomorrow. Now the chatbot has no interface whatsoever with anything to do with deliveries or anything along those lines. Obviously, no human is there to interject and stop the chatbot from talking rubbish. So that person would have been somewhat disgruntled, suspect. That's easy to fix when you see it. You just put that into the prompting, don't do that. And it hasn't done it since, but it's one of those classic kind of unforeseen circumstances, which is much more prevalent, I think, with AI than other types of technology and what have you, because it's much more fuzzy in the way that things get generated and how it responds to stuff. And that, for me, would be one of the big tips to people is... You do really need to think of every feasible scenario that you possibly can and test it in order to not launch with something that might not be quite what you want it to be. And there will always be things like that. know, nobody can think of everything, but yeah, that's definitely one of the things we learned. I'm an obviously, I am not a techie person at all. So just share a little bit around what the process actually looks like. to implement it and the time period of that, like how long it took and the process. I found it was very easy to implement. I'm saying that from somebody that's, you know, worked with computers and used them since I was seven years old. But to me, it did appear to be very straightforward, even without, you know, any great deal of technology. So essentially you have to create a knowledge base and that knowledge base can be any combination of your website. It can be files that you upload, which could be ⁓ PDFs, text documents, HTML, and you can also directly enter information. So you have these kind of FAQ type things where you put the question in and the response in and that then trains the system. You can have it relearn on a regular basis. So obviously your website will get updated and things like that. You can have it go and fetch that data on a regular basis so it stays up to date and the configuration of it beyond a few basics of what color do you want the chat thing to appear, the widget to appear on the website and put a logo in it and what do want it to be called and that kind of stuff. It's really just down to a prompt. And I say they do give you like a template to begin with based on your use case. So you could be using this, like we're using it for kind of pre-sales customer service, not really customer service, suppose, because not customers yet, but you get the idea, know, pre-sales type stuff. You can use it for more direct support, which is after sales type support. depends what you train it on. And then the prompting to set the kind of boundaries for it are based on that. You choose your model as well. So you can have chat GPT of various flavors, Gemini, Claude, whatever you think. Hopefully they don't have Grok as an option. That could go quite unfortunately. And then you write out your prompt and it's pretty long. You know, it's basically a list of do this, don't do this, when this happened, you know, whatever things you need it to do, tell it what kind of tone of voice to use. basically give it context. you are a helpful sales agent for whatever company and a little bit of background, you know, just give it some context. And then you have a playground. So you just go and, you know, like any, if you were just using ChatGPT, if you're familiar with that or Gemini or whatever, it's exactly the same. It's just a chat window. You bang stuff away, away it goes. And it logs it all. So you can go back and look at the logs and review what's gone on. Now, the bit that gets more complex is if you want the chatbot to be able to do stuff. not just essentially regurgitate information that you've fed it. And for us, there's a couple of things in that. One is if somebody wants to know whether something's in stock or not, I need it to go and check that because although the website tells you if stuff is in stock or not. It obviously is very dynamic and the agent isn't retraining itself every hour. And you're talking about, in our case, many thousands of pages. So it's not feasible to do that. So it's got to have like a live link. Now that is where you get into proper techie development stuff. And I needed our development team to give me something that I could interface with and all that kind of thing. And then the other thing is to create a quote. So it does have a very basic quote functionality, but it really is just a contact form, name, email, whatever. And that's no good. We need details. So I then had to create something else which would be able to do that and send an email to the sales team because this particular platform I've chosen for some reason doesn't have this function. Some do, some don't. As I say, there's a vast, vast array of them. So that's where I hooked it up to Zapier. If you're familiar with that, that's the kind of system you can use to integrate other things and perform functions on specific triggers. So essentially I can get the chatbot stuff to talk to Zapier and it sends the contents of the chat. with the contact details of the person gathered from the chat, extracts the email address, extracts the phone number, formats it into a little email and pings it off to the sales team. And that's all kind of done by Zapier once the chat has been triggered. So you basically set a trigger point. And again, that's down to prompting. So you have to tell it, ⁓ if you're getting a quote, don't bother generating a quote until you've got an email address, a name and a phone number. because fundamentally it's a pointless if we don't have that. So, you you can set those rules and but you do then need some extra stuff to make it happen in the background kind of thing. So that was more advanced. But if you really did just want something that was going to answer people's questions and not really do anything else, it's really easy. love that because I think for most small business owners that are going to take on this project themselves because you have to be directly involved. This is going to be the face of the business. The biggest thing as business owners we're trying to make it do is to ease the sales process, isn't it? We don't want people sat on hold waiting. We don't want our most valuable staff answering questions that can just basically be regurgitated off the website. I think, and I've got you complete run through, I'm wondering when you pitched it to the business here and you are a master of the pitch, let's be honest, how did you pitch this to the business? What problem did you? Sixox, I think one of the things with AI that I often find is it feels like a fantastic tool, but I'm trying to work out what's the problem I've got or what's the outcome that I want before I start using this tool? Yeah, that's a good question. I mean it was an easy sell really to the business because this is not an expensive thing. This is, think it's costing us about 150 quid a month, but that's across five websites handling quite a lot of chats. So if you're a one website relatively low traffic kind of site, you're talking about probably 30 quid a month. So it's not an expensive thing to do. So on that basis, an easy sell. And obviously we're trying to improve the customer experience, ultimately generate more business by being able to handle more stuff in any given period of time, really, and hopefully improve the user experience on the websites to help people make choices and stuff, which ultimately leads to sales. And if they didn't get that information, they might be stuck and never actually commit to buying something. So all of those things. There's no real downside beyond the AI fantasizing things that aren't real. And obviously that is the big risk with this stuff. So you do need some pretty hefty guardrails in terms of the prompting type of stuff to stop it. the example I gave a moment ago about promising people stuff will be redelivered when it doesn't even have the information, let alone an interface to make that happen. That's your risk. So it's a classic case of computing garbage in, garbage out. If you're not careful with what you feed it, you can't really expect You can't be disappointed when it comes out with rubbish if you didn't give it good quality in the first place. But yeah, it wasn't a difficult sell because it was relatively quick and easy, very low cost. And as long as you've got the resource to be checking those chat logs quite a lot in the early days and nipping problems in the bud, then it's really a no brainer in that scenario. what was the... I suppose I've just got one question. Just does it come in at it? from my own point of view as a business owner, if I was implementing this, being the non-techie person, you sold it really well, that it's low cost, straightforward to a point you need guardrails. So sorry, I've got two questions. What's the time period? Like, could you do it over a, is it like a day, a week? Is it a couple of months of testing? Like, what's that time period like? What is the one biggest challenge of implementing it? Sure. So the time it takes to get this set up is it could be almost minutes. Depends how much information you have to prepare to feed it in the first place. If you've got everything in a format already, and if you have a website and all your information is on the website, then all you do is tell it the address and it'll crawl it. That's it. And it will train in relatively quick. Now, if you've got a big website, it will take some time to crawl, maybe an hour or two, if it's a big site, and then training, you know, some minutes on top of that. So that's very quick. And if you're uploading documents, obviously as quick as you can select them and upload them. So in that sense, not very long. then you want to have a play with the prompting and the testing, obviously tweaking things like the visuals of it and what have you. If that comes together quite quick, really is something that you start in the morning and you've got a working chat bot by the end of the working day. It really is very quick. It will take you longer if you realise as you go through the process, ⁓ I want to feed it this information, but I don't have that in the right format or something along those lines. In that sense, it's like a lot of sort of creative processes, people building websites, putting brochures together. The time is not so much in the build, the technical or the design elements of that. It's the collating of the information that is actually going into it. That is the real time circle and that really pushes projects out. This would probably be similar in that respect. The biggest challenge that I came across and it's not perfect yet by any means, but in getting it all set up in the first place, the biggest challenge for us was the allowed size of a knowledge base. All these platforms have a limit in megabytes of how much stuff you can feed them in the knowledge base. And We've got websites that are many thousands of pages and some of these things will let you set rules. Like if you don't need all the pages, only certain ones, you can tell it not to others. You've got no choice. We crawl your website. That's it. And if they hit the limit, you can't really do much about it. That's very frustrating. But yeah, our big problem was we've got lots and lots of products and that's lots and lots of data. And I did struggle in some cases where it looked like it was going to be a great solution, but the knowledge base just did not allow me to put everything in there. It's all or nothing. there's no point having a chatbot that only knows about half your products. So ⁓ yeah, that was one of the challenges. And I ended up having to work with our developers to essentially extract the product information and format it in a way that reduced the size because web pages are very big in the main. They've got lots of stuff in them. And most of that stuff is not actually the core information on the web page. So I had to kind of get an export of the data. and I had to format it in a certain way. One of the big things that I learned going through that process is kind of how AI handles that sort of data. So all AI stuff has what they call a token limit. And it means that things get chunked up because it can't process an enormous document in one go. That means that you need to provide any sort of substantial information in such a way that the AI will be able to still understand it. when it gets to the end of it, even if it's split into however many chunks, possibly many hundreds, and it still understands what the last bit of information relates to, even though it will forgotten what the first bit was effectively. So classic example, if you have a spreadsheet of all your product data, you will have column headings and let's say there's 8,000 rows in it, you upload it. There's no way it can process 8,000 rows of that stuff in one go. So the first chunk is fine because the column headings are there. Every chunk after that loses the context of what the column headings are. So then it's guessing at what each cell contains does not work. Your solution to that is to take the column headings and insert them at the start of each cell. it is, so for example, if it's like product name is the heading of the column, you've got to take that. And for the, the cells that actually have the product names and you've got to insert product name, colon, and then whatever the content of the cell is so that you're literally telling it every single time product name, product. And then you've got to bear in mind, you've got to have linking context. So if I'm telling you what the price is, you need to know it's the price of what product. So you start having to tie all these different bits together to make sure that the system understands everything you're telling it with the right context and not just disassociating stuff. And then it all gets very confusing. So that was a ⁓ big learning thing for me that, yeah, you've got to do it that way. Otherwise it's just not going to work for you with large volumes of data. And yet the other thing was how do I file size wise shrink down my information? to fit in the knowledge base, but still do the job. you've got to bear in mind, AI stuff does understand semantics in terms of formatting. So if you make something a heading and then you follow it with some text, it will make the logical assumption that block of text relates to the heading that is above it. So you don't just want plain text all the way through because then there's no semantic kind of guidance of, well, that relates to that because that's the heading of that. So you then have to, you use whatever. So they generally accept what's called markdown, which is a much more lightweight way of formatting text rather than using HTML tags. You know, it makes stuff a lot smaller text file wise. So that was the other thing that we had to do with it, which is all rather techie and boring. But those are just specific examples of a more general challenge that you might have with this stuff. And that's just as true. If you go to chat GPT and you want it to take a look at some data and you've got a massive export of data of whatever it is and you want it to do some analysis for you. You will have the same problem if you've got column headings and thousands of rows in it. So whatever you do with AI, when you're working with large amounts of information, worth bearing that in mind. It's going to get chunked up. What happens when that, when, it does that, will it lose the context? And that might, you know, you might get much better results if you try and address that at the start. think you might just have saved somebody three or four days worth of research and work within. what that statement you've just said there. So for me, the kind of the first step of this is almost like a no brainer because the chat bots at the moment are such a low price from what you're telling me. It's kind of look, we were purely going to save time by having this on there. The second step is to create your knowledge base. So to get that knowledge base in there, Haley and myself tend to cheat and we use often these transcripts from these calls to say, look, this is Phil, this is Haley. This is our tone. want you to write in this tone and use this tone going forward. And I'd suggest that to sales teams as well. Record a sales call, get it transcribed and say, this is how we like to treat customers. You've then got set your chat bot to learn regularly and decide how quickly you want it to update itself to look over there. We then go into the configuration or the prompt questions, working through the templates and actually building the chat. Then we've got choose the model we like, which is GPT or Gemini or where we're And then we're really into the right out your prompt, the tone of voice, the context. And then if I take it to the next level, which I think is where, you know, the real nerds are going to absolutely enjoy it is we've got a chat bot that's working. Can we now make it interact with things like stock lists? Can we make it interact with the rest of the website templates? And of course we're going to need Zapier. And I think that's probably where our small business owner says, do you know what? I really like this chat bot. I'm now going to bring in a third party. to start helping me out with integrations because the system works. But I think those first five or six steps, I think every small business can go through. So for yourself Ian, a little bit of a shameless plug for you in whichever way you would like to, is there anything you would like to promote or is there an ideal client that you'd love to get in touch with you? And if so, how would people get in touch with you or how would they find out more? Good question. I don't particularly have anything to promote at this moment in time. I've been a bit busy, so yeah, I'm kind of happy where I am. But I am considering, and you and I have spoken about this, putting together something which is kind of like an on-demand support service on digital marketing for small businesses. understanding that agencies charge a lot of money and you're paying for a lot of overhead with agencies. Many small businesses can't or wouldn't want to spend that kind of money and get tied into annual contracts at however many pounds a month, usually four figures. But those businesses still need some help. And usually because budgets are tight, they're trying to do stuff themselves, but not necessarily having the expertise to know what the right answer is to certain questions or how to get certain things set up. And maybe what we've been talking about with ChatVox is an example of that. in some respects, hopefully at some point in the near future, I will have something together where businesses can pay a monthly fee or join a group of some sort and kind of have like a bit like the service you provide in terms of tech support ticketing, but not for tech support for digital marketing. And then it's a low cost, but you know, trusted expert, hopefully kind of to lean on when you've got some questions. I'm not proposing that I'm going to be doing all these things for people and spending hours working on their stuff, because that would just be an agency again, that's very expensive. But in terms of I'm trying to do this, am I doing it the right way? Or I want to do this, what's the best choice for this platform or software? Or I've seen this in Google Analytics, what's going on or whatever, know, those kinds of things. That's what I want to try and support people with. And hopefully I'll have that up and running soon. So. If want to hear about it, your best bet is probably to either find me on LinkedIn and if you just search Ian Lockwood, look for my face and you might find me on there. Ian Lockwood Digital Consultancy is the name of the business. Ianlockwood.net is the website and you'll find my contact details on there as well. So one way or another, you should be able to kind of find me. If we connect one way or another, then yeah, I will let the people I'm connected to know when I've got a service that might be interesting. Well, thank you very much, Ian, for joining us. It's been a really, really interesting conversation and definitely a one thing that we've not had before. So I think people will find it really useful. So thank you very much. quite welcome. I really enjoyed chatting to Ian. He is an absolutely amazing guy. He's so clever. He knows his knowledge is absolutely unbelievable. And I've got so many little takeaways. But what was your big takeaway, Phil? I completely agree that Ian is miles ahead in my mind of the curve. think the biggest takeaway for me is how rapidly AI has moved, especially when it comes to chat bots and putting things on websites and the ease of which it was to implement it for Ian. Now, as you say, Ian's an incredibly intelligent guy. So I think it would probably take me longer than it has done with Ian. sounds like with Ian, it was a matter of kind of hours to get something up and running. Having said that, I think for half a day to be able to get a chatbot that can react to frequently asked questions and can absorb all of that information about your business. And we're probably similar in, we probably asked the chatbot to train us almost, you know, what, what are its responses? I want it to act as the best version of me. So how would you answer this question? So yeah, so simplicity, cost and ease of access, I think were my big three takeaways. How about yourself, Hayley? I think the cost piece of it. really surprised me at how accessible it was, the cost effectiveness of it. It was such a, I don't want to say cheap, because it's not, it's not, it's always an investment, isn't it? But compared to what I thought it perhaps would be, it's incredibly accessible for a lot of small business owners. I think the big takeaway for me was how complex something so simple can get and the different versions or the different layers. So you can start with a chatbot that's like very basic off the shelf. It just does this easy to set up. But then if you've got a really complex business, you can add all those layers in and keep going with it. So it almost can grow with you as your business grows. So if you were relatively early on in your business life, it's something you can invest in now and then layer all of those different steps in as your business grows and it gets more complicated. I thought that was That was really interesting. Not that I've ever explored the world of chat box before, but the other thing that I thought was really interesting was how even it did get to a point with somebody like Ian, who obviously knows so much about that world and so much about the internet and the marketing side of the internet. But it did get to a point where it was still challenging to him about how he connects all of the, was it the deliveries that was the piece that was the challenge, wasn't it? How he got that delivery piece into it. I think the big takeaway from that was that the... ⁓ Off the shelf chatbot has developed massively in how simple it is and how much it's just like a, you know, next, next finished kind of job. However, then integrating it with other systems, because I mean, Ian was talking about when he gave it a lot of data that he had to be very careful that he had to put precise titles within the data so that it didn't get confused. And I think the integrating with third party systems, because they're speaking in a different language is then going to be another piece again. However. Once we get to that, mean, the kind of, you know, the old cup of coffee, it's a chat bot, it's cheaper than a cup of coffee every morning. I think the integration side of it is, how much time will I save? How much more useful that is this? And I mean, if you can integrate it into your stock system and your delivery system, because I think the world is moving to the same. kind of thinking that I've got that I almost prefer to buy things off Amazon sometimes because I don't have to make that phone call. I don't have to go into a shop. So we're easing that customer journey by getting that chat bot really tooled up for want of a better word. Yeah, I completely agree. And you actually said it in the conversation about how you prefer to be able just to message in the chatbot. And I find it really frustrating. I'm exactly the same. And I find it really frustrating now if you get a chatbot that isn't very sophisticated and it can't answer you. It makes me very cross. So I think it's a great way for the world to be going. I also think what was interesting was the impact it had on the workforce and on the sales. So it's allowed that business to grow significantly. In theory, give better customer service and not adding a huge amount of additional cost to do that. Yeah. And probably make employees days a bit more interesting because now they're answering the out of the ordinary rather than just the repeatable questions. So yes, it's probably got a really, before Skyline takes over and takes all of our jobs, it's got a very beneficial impact for. everybody involved. Yeah, yeah, it definitely has. And I think it was, it's just a great example of how we're using AI in everyday life now. I think the only huge caveat that I would place on it is that I've often seen AI hallucinate and give us the answers that we want to hear. think Ian bringing that up, how many times has it hallucinated, you know, how tight do we need to make the guardrails? Because depending on your business, one really bad answer could cost you a huge amount and I don't know from a legality perspective how easy it is just to say anything this chat bot says might actually be rubbish so you can't sue us. Well that was another great takeaway wasn't it of the whole conversation was those little snippets that would trip you up if you were doing it and you weren't aware of that. So when will we see a firm chat bot in the world? I don't think that we sell a simple enough product or get enough inquiries at the moment to dedicate that time. Even if it is a couple of days, there's probably more beneficial marketing and sales activity that I could do with those couple of days than design a chat pod. And it's likely that I would get too entrenched in it and really excited about it and start adding too much to the picture. How about yourself, Hayley? It's really interesting actually, because I've been thinking since we recorded about how could we utilize something like that at Corba? But I don't know, I don't think we're quite at that point because I think about the questions that we get asked that you could utilize that kind of software for, but I'm not really sure how beneficial it would be because some of it can get very technical. So then it goes back to that reliance point and how you were talking, you know, the hallucinations and all of that. Would you really want to be putting that out there? I don't know. I think I'd feel a bit nervous at the moment. I think what it could do for both of us is when somebody comes onto the page say, have you got any questions? There any quick questions you'd like to be answered? One thing that I'd like to do more of is kind of displaying our prices on our website. So if there was a chat bot, it could potentially give people the chance to go through that because nobody likes to enter all of the details or not nobody. I don't like to enter all of my details before getting a prize. I like to know a ballpark figure and then let's work on refining it. So it could potentially be used for that. I'd love to hear from our army of listeners if they're using chatbots, if they're finding this useful, if this is something they'll implement. And if you have joined today, I really hope that you found what one thing AI useful and found that it was helping to move your business forward. And if you did, please like, subscribe and share this episode with another business owner just like you.