146. Advanced AI Tactics With Varant Bomoushakian

This episode of Professional Builder Secrets is brought to you by Apparatus Contractor Services. Apparatus provides full scope, end to end QuickBooks, Precision Construction accounting and CFO services for residential builders. If you feel like you can't completely trust your weekly and monthly financial reports or you feel like you're flying blind, Apparatus Route will rebuild and operate your QuickBooks accounting system and give you precision construction accounting capability you can trust. Learn more@apparatus team.com or on the APB Rewards Partner Directory.

Play with AI. There's no point in learning, learning, learning, learning if you're not going to apply it.

Welcome to Professional Builders Secrets podcast for building company owners wanting to grow safely and securely. I'm your host, Will Blunt, and today I'm joined by APB's head of operations, Vrant Boomashakian Varant. Welcome to the show.

Thank you for having me, Will. And you did a great job pronouncing my name.

I had to pause because I'd basically been practicing a different pronunciation all day in my head. And then I asked you what it was before we got on and apparently I was wrong.

So you nailed it. You nailed it. Well done.

Very good. Well, awesome to have you on Voran. I must admit, every time you and I talk about AI, I guess we call it offline, even though it's online. When we're outside of this environment, I'm kind of like a kid in a candy store. I listen to the things you're doing. Often I'll come into one of our leadership meetings and I'll think, yeah, I'm pretty confident with where I'm at with AI. And then you'll start talking about the stuff you're doing and I'm like, okay, pull myself back into line. Because you're really always keeping on the ahead of the curve and testing things out, finding out new opportunities and applying them. So I'm super excited that we've got the opportunity to now start bringing that to the builder audience as well. On a broader level.

Excited to share anything, anything that's going to positive come out for our community. I'm keen to get it across and yeah, let's do this.

Awesome. For anyone that doesn't know all the members out there, Vrant's the one that shares tips about AI in our private members Facebook group as well each week. So if you're a member, watch out for those. But for everyone else out there, let's dive into AI in the building industry. Vrant, what makes you uniquely positioned to talk about this topic?

Yeah, look, so I'm a bit of a tech nerd myself and as APB's AI internal AI expert, I'm always having to put myself in the forefront of anything AI related, whether it's already been developed or it's coming out. So there's a couple of different ways I do it. And the journey really started in November 2022. That week that ChatGPT was launch, there was a lot of fanfare on the media. That was the first time that that tool was released to the public so anyone with an email address could register and my phone was going off the hook. A lot of my friends in tech were messaging me saying, Varant, you've got to jump on this ChatGPT bandwagon because you're going to love it. So I did, I tried the first time and I couldn't get through because I think every Tom, Dick and Harry were trying to log onto the website in the same week. But sure enough, a day or two later my registration was accepted. Started playing around with ChatGPT and I haven't stopped since then and certainly my, my scope's only expanded and I guess the way I try and keep on top of it because it's, it kind of reminds me of when the Internet was first launched. You know, things were developing at such a rapid pace. There were developments almost on a daily, weekly level that if you weren't reading, reading the newsletters and, and, and staying on top of all the new newness of what was going on, you'd get left behind pretty quickly. So over the last two years, I guess I've subscribed to two dozen or so AI newsletters. I've whittled that down now down to six. So newsletters such as AI Advantage, the Rundown AI, Reuben Hasid's newsletters, they provide some great weekly, weekly tips on what's coming out, what's new. I attend regular AI community meets. I attend a Mastermind bimonthly with some other expert marketing professionals who are just pushing AI to its limits. Anytime there's a conference locally in my town, I try and put my name down for it because seeing what others are doing sort of expands your mind as to what is possible with AI and what kind of solutions you can bolt and plug in together and of course, unashamedly YouTube. It is an amazing resource for seeing how people are using Chat GPT and not just ChatGPT, a whole range of other AI tools on how they're utilizing it, how they're connecting it together, how they're integrate, integrating it into their systems. I get a lot of ideas from YouTube and also how to's, how to set something up because some of it can be quite technical. And last but not least, I'm. I'm a very curious person by nature. So if I see a tool that's really cool or see a tool that can make a difference, I'm generally clicking up, signing up, trying to get the free account, just to test it out and see what it can do. And I've done, gee, hundreds of those tests in the last two years for different AI tools that have been released. So that keeps me as sharp as possible, I guess.

Yeah, that's amazing. I think my biggest takeaway from what you're talking about then is how fast it's moving. It feels as if you don't look at something new or stay on top of a newsletter or a podcast or something like that, that a couple of weeks go by and you're already a bit behind the pack.

Yeah, absolutely. And even if you're using just one application. So for example, ChatGPT is the one that everyone knows and loves, and that's probably the most mainstream application. Even Chat GPT, they have some regular releases that quietly slip under the radar if you're not paying attention. For example, their memory feature or their custom instructions features that have come in or only about a month and a half, two months ago, their desktop feature that came out. And if you're not paying attention to it, you could be happily using the Chatbot as you've always been for the last year and a half, two years, without knowing that those extra enhancements have been added in, which could actually improve your workflow quite a bit or make things that were a little bit tricky before much, much easier.

That that desktop example that we were just chatting about off air has totally.

Revolutionized my days using using the desktop version. And at the moment I think it's only available on Macs. Okay, it's going to come out on Windows, down the track. But the Mac version of ChatGPT allows you to interact with it via voice just the same way that you can on your mobile, which, I mean, if you're wr email, it's quicker to dictate what's on your mind, doesn't even have to be perfect or professionally spoken. You could just sort of sound it out and it will generate an email for you. And if you get a little bit more technical or advanced, you can actually create a GPT with previous emails that you've written so that it can write in your tone and your language and style. So the users at the end of the day, the end user won't know that you've used GPT to create it because it won't start off with that pesky, I hope your day is going well, or I hope this email finds you well, or anything that sort of flags that it's been AI generated.

I love that. Yeah, you told me about the desktop app a little while back and I was like, yeah, sounds pretty cool. But I was a bit slow to use it. And then as soon as I did, I was like, wow, I should have listened to Fran.

You're right.

But absolutely amazing. I mean, literally, like you get off a meeting with someone, you need to send them a follow up email. That email could take you 15, 20 minutes to write if you're doing it manually. Otherwise you can just dictate straight into ChatGPT, one minute, two minutes max, and it pulls out the perfect email that is probably written better than you could have done yourself.

It's just mind bending what GPT can do. And there's so many use cases once you start using voice as well. So, for example, I was doing a team planning workshop with myself and sometimes when you've got a blank screen open with that cursor blinking at you, you've got that writer's block. How do I start? How do I. What are the first, what's the first sentence I'm going to kick off with? But if you're going to have to explain to someone what your plan and strategy is, it comes out much easier. So I remember about a month ago, a month and a half ago, I just turned on the voice feature on the desktop app and just started spewing out what, what, what sort of plans I had for the team and what sort of goals I want to set for individuals and what are the outcomes and challenges that we're going to try and overcome. And after about six minutes of speaking, you know, press the little tick button on ChatGPT and it just took all my raw voice notes transcription and then generated a beautiful summary for it, which I could then, you know, enter in a few prompts to get it into a, into a format that I can extract and replace my solo individual blinker on my, on my white page with a bit of text that I can then modify and tune to get it to the perfect state.

Yeah, I don't know if you found the same. It seems like voice or transcripts tend to get really good outcomes from ChatGPT. For example, I noticed when we upload a transcript of a podcast episode or an interview or something like that to write articles, the output's much better than if you just, you know, try to prompt chat chatgpt with a few questions or, you know, points to cover.

Yeah, I found that as well. And I'm not exactly sure as to what that specific reason is, but it could be a lot to do with. When you're writing something out, you're trying to get it as succinct as possible. It doesn't incorporate your tone and your personality into it, but if you're just having a conversation with it, the AI is going to sort of understand and learn what your tonality is. You're probably going to be giving it a bit more context, because if you're writing a long prompt, by the end of the prompt, you're like, I just want to keep it as short as possible so I can press Enter and see what the. What the outcome is. When you gob on, you know, you, you could, you could say a lot to AI. And it does a great job at filtering it out. So I think it's giving it more context and because it gets more context and more verbiage, it's able to produce a better outcome. So that's my theory anyway.

But that's a great observation I've found myself after. Sometimes you'll do a really detailed prompt, but then if you want to follow up with multiple prompts, you find yourself getting a little bit lazy. Like, I just want to get to the next step. Whereas when you verbalize it, you can. I. I found it. It cleans up all of the language anyway. You can say anything and get a lot more across in a short without worrying about needing to write it succinctly or any of that kind of stuff.

Yeah, and you don't really have to edit it. I remember when I first started using the voice feature, I would get the transcription finished, then I would edit my transcription to take out grammatical errors or take out chunks that I didn't want AI to be looking at. But as time moved on, I got a little bit lazy with voice as well. And I thought, bugger it, I'm just going to press Enter and see what happens. And the results have been just as good, if not better, because it's got that extra nuance, it's got that extra, extra context that it can lean into. So, yeah, I definitely encourage folks to give the voice version a go, even if it's on their mobile phone. So if you haven't got a Mac machine, put GPT on your phone, make sure it's a paid account and you can have a conversation with it. And the results are quite incredible. You'll be surprised at how clean it can transcribe your voice. It doesn't have to be for writing emails. It could be for, for anything that needs to be into a written output eventually.

Or if you don't have a Mac, just get a Mac.

Yeah, I didn't want to start that, but sure, let's do that.

I promise Apple aren't sponsoring this episode. Okay, so back to the topic at hand. What is the current state of AI in the building industry specifically? We've spoken a bit more generally about it.

Yeah, sure. So I think AI is becoming increasingly integrated across all industries, including the home building sector. And as time goes on and as AI develops more and more and more and gets integrated more into software, I think we're going to be seeing it's going to be synonymous with our, with our existence down the track. So it's, at the moment, I think in the building industry, it's in its early stages of adoption. Despite its potential, you know, AI's maturity still at that early level, it's only been out for, it's only been out to the mainstream public for less than two years. So. So you've got to give tech companies a bit of time to catch up and really sharpen their tool sets on how they've implemented AI into their tech stacks. But it was an interesting survey that I came across a couple of weeks ago and I had to actually look for it just before this podcast, just so I can refresh myself on the statistics. But a survey at the end of 2023 was done in the US by a group called the Shin Builder Partnerships that revealed, and they, they interviewed only, they surveyed only residential builders and they found that 35 of residential builders across the U.S. and this data is only for the U.S. of course, we're using AI and interestingly, about 93% of them were using AI to, for their market for marketing purposes. And you know, that's probably one of the most common use cases for using GPT, creating blogs, creating articles, social media posts, etc. And there were some fringe cases of AI being used for things such as land acquisitions, of finding the right plot of land that ticks a few different criteria and price points. They were using it for predictive maintenance, for a bit of customer service and of course project management. So I think as time goes on and software makers become more mature with their products, I think AI is going to be integrated into every tool that we use. I mean, it's starting to appear in our phones quite a bit with Facebook meta AI, which is actually quite incredible in itself. If you haven't tried that, you should definitely give it a go. And yeah, it's it. I was about to sneeze then, but. Sorry. Yeah. So over, over, over a period of time, I think it's going to be more and more. And I'm sure that number at 35, that was towards the end of last year, I'm sure that number will grow quite significantly. And as software providers that are getting more specific with their use cases, especially in the building industry, things like project management, estimating, design and planning, risk assessments, all those kind of tools will have AI inte into it as time goes on and it'll only improve with time.

Yeah, I definitely think that number would have increased in the last six to six to 12 months. Just looking at our Facebook members group, the private members group we have for APB members, the chatter in there about ChatGPT and when you chat to the builders in our community at events and things like that, it's pretty, it's pretty well penetrated into the market. Just the use of ChatGPT specifically.

Yeah, yeah. And I mean, ChatGPT is only one AI tool. There are exactly, yeah. So many other AI tools and so many other language models as well. So one that I've been playing around with is Anthropics Claude, their, their large language model is probably a little bit better than chat GPT4O at the moment. I'm seeing that it's producing better outcomes and I saw some statistics on the newsletter the other day where they benchmark certain things that the LLMs are compared, compared to each other with. And Claude's. Claude's large language model has just slightly outperformed ChatGPT. So just watch that space.

All right, let's take a step back. What's a LLM and what's CLAUDE like? What does it do?

LLM is basically stands for large language model and it's what ChatGPT is. It's a large language model. It's using natural language to be able to query all the data that it's consumed. So, I mean, people. And I used to think this as well, I used to think, oh, AI is just this black box. And some, in some cases it is a black box. You don't know what's going on inside of it and how it's all put together. But essentially ChatGPT is a curation of a vast amount of data from various sources, whether it be scraping all the sites YouTube, Twitter, all the social media platforms that are out there written, written books on specific areas. So I know they're training teams, did a lot of work pre launch to teach it all these different contexts, whether it be medical or, you know, in our case, construction, building, they tried to feed it as much data as possible. So that's why it's called Large Language model, because there's a lot of data in it and essentially it's, you know, the ChatGPT interface or the CLAUDE interface gives you the ability to query all the knowledge that it has and give it instructions on how to put bring that knowledge together to give you the outcome that you're looking for via prompt.

Thank you. Okay, so before we move on to more different tools in AI, let's talk a little bit more about ChatGPT, because everyone out there I think, is probably using it to some extent and there might be opportunities to go a little bit deeper. What are some of the most advanced ChatGPT tactics you've seen implemented?

Yeah, so I've seen quite a bit of stuff over the last year and a half, especially when I've attended some of these masterminds and followed some of these influencers that are just at the cutting edge of all these AI tools and what they can do. So I'll drop a couple of the examples that I've come across that impressed me anyway. I mean, previously I said the highest use for builders in the US at the end of 23 was for marketing. And I can almost bet my bottom dollar that if people are using ChatGPT for marketing, most people are using at a very basic level, for example, generate a blog about xyz. But in one of some of the masterminds that I've attended, I've actually seen some really clever marketers take, take, use GPT and take it all the way back to fundamentals and foundations. So to create marketing materials and they start with, you know, identifying pain points and describing their avatars, describing the audiences, delving into the desires of the customers to really prime GPT really well. So that GPT understands what kind of audience that's going to be writing for what kind of target market it's going to be using language for what kind of pain points it's going to be addressing. So that the final outcomes of these, final outcomes of these flows are creating some beautiful website copy or landing page copy or blog articles that are in the tone and voice of the company that's generating it. And you almost look at the first draft and think a human, an expert Copywriter had written it. Now all these marketing guys don't use, have never used the output from GPT verbatim. They still add their touch and flair to it. But watching them, watching them take it to the fundamentals, marketing fundamentals, and use GPT to bring it all the way up to creating some really consistent results was truly impressive. Other use cases where I've seen GPT use quite a bit outside of the user interface. So everyone knows the user interface. It just looks like a chat. You go in there, you type your questions and it's going to spit out a result. But some really advanced use cases I've seen ChatGPT used is actually using it via the API. So using tools such as Zapier or Make.com that are integration tools or middleware tools, I've seen people connect it up into various different pieces of software, predominantly CRMs. So for example we use HubSpot. Salesforce is another big one where if they want to drive a certain action and get a response or an output back from ChatGPT, they would use these middleware tools and connect to ChatGPT using API. So one example which, which I thought was pretty cool for companies that are doing outbound calls using their with a sales team, sometimes these SalesPeople would have 50 or 60 companies to call in a day. Cold call. And I don't know about you, but when I used to do cold calling a long, long time ago, it was better to do research on the company that you're calling so you've got a bit of context or what you're going to talk about. So I saw a clever fellow online a little while ago that basically said, right, we're going to serve all this information directly to the salesperson without them having to lift a finger and do a single piece of research. So from the CRM they had automated a workflow that would connect to ChatGPT using Zapier that would go out into the web using a prompt query and search for specific criteria on a customer and then bring that data back from ChatGPT and embed it within the customer record or the prospect's record in the CRM so that every time the salesperson finishes one call, they've already got the next customer pre researched with some talking points and some interesting news things, news clips that have happened in the last couple of weeks just to give them some talking points. So that's an example of using ChatGPT using APIs and integrated tools.

That's awesome.

Yeah, Chat GPT itself and so do, so do Claude have a great analysis and reporting functions within it as well. So we were playing around with this a number of months ago where we were loading up huge spreadsheets full of data and getting Chat GPT to basically analyze that data and give us trends and insights to what was going on within their information. So we tried it out on our sourcing data as well and some of the insights that, that IT generated were quite insightful and spot on and even some it picked up on some areas that we hadn't thought of before as well. So that's a great tool. If you haven't used the data analysis portion of Chat GPT or other large language models, I definitely recommend that you jump on YouTube, check out how others are doing it and just, just start off by uploading a spreadsheet into a spreadsheet full of data into ChatGPT and seeing what it, what it comes out with and then having a question on.

Yeah, quick question on that. I've always been frightened of doing that to a certain extent because I've seen ChatGPT make stuff up before. Like you might, it might, it might pull 10 pieces of data from something you upload and eight of them are accurate and two of them are made up, you know what I mean? So how do you go about fact checking what it's doing?

No, great question. AI is known to hallucinate and with any output that AI produces, I would never advise anyone to use it verbatim because AI can make mistakes, not because it's making a mistake, because it's wrong and it doesn't understand it. It probably hasn't got the context right or it hasn't got the column, column reading right. So I have done quite a few tests where I've loaded up pieces of data and the way to verify it is obviously spot check it yourself, make sure that you're going through it, you understand what that data is, you're not just feeding it blindly. And you would know what the calculations or the output or the outward output would be for a particular section and that way you can verify that it's understanding, understanding your data well and giving you the outputs correctly. And once you verify, you know, five or six data points that are matching up with, with what your results are, then you have a better trust of the output that's generating. But if you start seeing a mistake or two in the first five examples that you, you're checking, then I would scrap it and start again and maybe add a little bit more context into the column headers sometimes that, that even does the trick, adding a bit more context into the column headers before you load it up so that it knows what it's reading gets a better context for it. So yeah, definitely don't use it verbatim but once you do stress test it and make sure that it's producing some consistent results, you know, multiple times, you know, five out of five times, then you can trust the data a little bit more as you start using it. Cool. Yeah. And there's quite a few other points as well if you want me to go through them. I can.

It's up to you, man. I'm interested. It's. Yeah, it's really valuable information so. So dive in.

Yeah, I've got two or three other examples, ones that I've actually used myself and have saved me not just time but a lot of money. And one is, I mean, you're not a lawyer. I'm not a lawyer. I never pretend, pretend to understand the legal speaker that well.

Good.

Yeah, I've seen a lot of legal documents in my time, but I still can't speak legalese. So Chat GPT has been incredible for, you know, because at the end of the day it's a document. It's a document, it's got a lot of text. You can load that up into GPT, ask GPT to assume the role of a solicitor or a lawyer and give it instructions to be able to explain it to you in a simpler format, whether it be, you know, describe it back to me in as a 10 year old or simplify it and you can ask, you can start asking questions about what does this cause mean? What does that cause mean? And then bringing it a little bit further, you can actually get it to write new clauses as well in the same language and tone as the contract that you've uploaded. And I've done that a couple of times over the last two years where, you know, something I might have had to sign, there was a little clause that I want to change but if I just try to write it out myself, you know, any lawyer half worth their salt would go through it and go, ah, a non lawyer wrote this. So ChatGPT has been incredible for creating those little, those little clauses that can just fit right into the document and, and sound like it was written by the same person that wrote the whole thing out. So you could save yourself a few legal dollars doing that. But of course, always get legal advice. I'm not, I'm definitely not advising that you replace your lawyers with ChatGPT but just to understand, understand some contracts and clauses a little bit. Clearly. It's definitely a great tool. Yeah.

I'm sure it's much more cost effective to get a lawyer to check a document rather than to write it from scratch. I don't know that for sure, but you'd hope so.

Yeah, I'd assume so as well. I'd assume so as well. And last but not least, and this is something that I've had to do not long ago, is you can actually create presentations and scripts all using GPT. So you can use your raw content or your, your voice to provide it context and the topic guidelines. And with some clever prompting or some specific prompting, with some tight rules put in place, you can actually get it to produce your first draft scripts for a video presentation or even a slide deck that you might be producing to give you some guidance as to what kind of imagery to use, how to transition from one side to another, and how to link the bridges between. Between topics. It's been really useful for that as well. So, yeah, there's a lot more stuff, there's a lot to do, but we'll be, we'll be talking for hours.

That's all right. We can chat about it offline. Sticking to ChatGPT for another moment, all the builders out there that are thinking, hey, there's so much I could do here. What are some practical steps that they could take to improve their use of ChatGPT? And obviously they'd be at different stages of their journey, but what, like, how could they apply some of the things you've been talking about?

Yeah, absolutely. I think the first place to start is definitely remembering to set clear objectives. You know, what do you want out of the tool, what kind of output do you want out of the tool? And then working back from that, you know, really defining the goals of the exact outcome they want. And the first area that I think people should brush up on their skills is how to write prompts. Because at the end of the day, when you're speaking to ChatGPT or any large language model, it doesn't know who you are, it doesn't know who your business is, it doesn't know the context of the question that you're asking. So learning how to write a great prompt will really take you far down the track of utilizing AI and getting better outcomes and to learn how to prompt better. I mean, there are plenty of resources on YouTube on how to prompt better, and it'll cover things like assigning GPT a role, you know, giving it a role, for example, act as a marketing expert. Or a social media generator or something like that, a very specific expertise role and then giving it context, specific outcomes of what you wanted to produce, giving it some rules and boundaries so that it doesn't just give you a whole bunch of guff. And learning how to do multi shot prompting, which is asking, asking GPT to do multiple tasks within the same prompt. I think that's a, that's, that's the first place that I would start because any other AI tool that, that would have a prompt window where you'd have to go in and give it instructions. If you don't know how to prompt well, you're going to get average results in all the different AI tools that you'd use. So I'd definitely say start learning how to prompt well there. If you haven't, if you haven't learned how to prompt well, there are Quite a few GPTs that are out there and GPTs are basically little apps that are within chat GPT that you can use where someone else has already written a prompt for you that can produce a consistent outcome. There's quite a few GPTs out there that will improve your prompt. So you can put in your basic prompts. It might be a paragraph or so, and then it'll expand once you run it, it'll expand it out and add some more boundaries and guidelines and context to make that prompt a little bit better and also give you a bit of a framework to add and edit that prompt. So definitely start with improving your prompt skills. The other areas to sort of speed up on AI and get better at it is invest in learning. You know, the last year and a half I've been consuming hours and hours and hours of content and that's the only way that I've been able to get my head around what's possible in AI and how you can use AI. So subscribing to a couple of newsletters, listening to a few podcasts, and I'm talking regularly here, not just one off. I've listened to a podcast now, an expert. No, no, no. AI is changing all the time. You've got to stay on top of it and listen to it regularly. And yeah, just being immersed in what's, you know, what the new developments are will keep you ahead of the game. And last but not least is play. With AI, there's no point in learning, learning, learning, learning if you're not going to apply it. So using AI, even at the basic level of having it on your phone, speaking to it daily, asking it questions, asking it, to summarize things for you as you get into the practice of using it more and more and more, you'll get a bit more confident using other tools or a bit more confident using. Using tools that have AI integrated with it, because you. You kind of have an understanding of the type of things that AI can bring to the table. So it'll give you a bit more of an understanding of its capacity, its limitations, how far you can push it, the type of questions that you should and shouldn't ask it and. Yeah, so it's just practice makes perfect, right?

Yeah. Just to touch on that learning thing for any Aussies out there that are listening to this. I know next month from the launch of this podcast, Fran will be speaking at the Professional Builders Summit in Sydney as well on these. On this exact topic, actually, and providing some practical examples. So that'd be a perfect opportunity to learn more. Also getting in as a member because you're also working on an action plan or an improvement on an action plan, which is going to look at some more advanced AI topics and expand on what we've already got for members in the back end as well. So that's some. Some other ways to condense a lot of learning into a short period of time, because you don't need to listen to the 100 podcasts that France doing.

Yeah, looking forward to it.

I wanted to touch on the prompting process a little bit more. One thing, and I'm not sure if this is helpful for other people, but the way I look at it, I remember when I had my own business and I would need to brief a contractor on doing a task of some sort, and the learning process of getting really poor output from contractors, until eventually I started to get good output from contractors, was understanding how to brief them in a way that gave them the information they needed and all the context they needed so they could understand what I actually wanted and go away and implement that task. And the more detail you give them and more context and all of that kind of stuff, the better output you get. And it's the same with ChatGPT. And that's why I think the voice side of things is really, really helpful, because it's like describing to a new employee or a contractor or something exactly what you want in as much detail as possible, so there's no guesswork. And then once you've done that, it just. It just understands what to do and just gives you what you need.

Yeah, absolutely. Context is everything in this case, you know, AI is much, I wouldn't say more intelligent than us, but it Definitely knows much more than us because it's consumed so much content. And you know, one good analogy I use when describing it to people that don't really know what AI is or they're hesitant in playing around with these tools is think of AI as your best friend who you don't owe any favors to, but who actually doesn't know you well at all. And if you want to ask your best friend who doesn't really know, who doesn't really know you well, a favor, you've got to give them a lot of context for them to be able to pick up your request and give you an outcome that you're going to be happy with. So there's the same way how you described, the more context and more information and guidance you give your new, new employee or your contractor, the better the outcome is that they're going to give you. AI is exactly the same.

Yeah. On a basic level, if you asked ChatGPT to write write me a blog article, okay, it could be on any topic and it just starts, it just starts writing. But if you said write me a blog article about educating a new home buyer about the benefits of going with a custom home builder, okay, now we're getting a little bit more specific and obviously you can go, you know, as far as you want from there, but I think that's, that's kind of a good example of why context is important.

Yes, absolutely. Nailed it.

Cool. All right, so Moving on from ChatGPT, what other advanced AI tools and tactics should builders be aware of? You touched on a couple with Claude.

Yeah, so, I mean, there's, there's, there's a range of AI tools that are incredibly versatile and they can be applied to various aspects of the business. So if we're talking about builders in specific, you know, a building company is very similar in terms of structure to any other business. They'll have a marketing department, they'll have a sales department. It might all be the same one person wearing different hats, but it's still a department nevertheless. And you could use a whole range of different AI tools outside of ChatGPT to, to get some efficiencies in those different departments. So, for example, on the project management side, there are some great project management tools that are out there that are integrating AI into, into them that can query your tasks. You know, create, create tickets, automatically query the knowledge bases. One one that we're using at the moment we just turned on the AI future is ClickUp, where, you know, you can ask for a summary of what's happened in a series of tasks that might be assigned to a person. And it'll give you a nice short, some nice short summary without you having to go in there and read every single task and ticket to see where they're at. You've also got, on the sales side, you've got a lot of sales follow up tools, whether they be standalone tools that might have automated email sequencing or SMS sequencing that would be able to personalize the message based on what the customer has looked at, what kind of blogs they've read, what kind of interaction they've had with you. And I know a lot of the bigger CRMs are already integrating these AI tools within them so you wouldn't have to use a standalone tool. And we're in 2024 right now and I think the last four years has really shaped how we communicate as humans thanks to Covid and the isolation that we've had had to experience for a year or two where everyone moved onto platforms such as Zoom, Google meetings, teams. And I remember in the first sort of six months of COVID you know, the number of Zoom meetings just went through the roof. And you cover a lot and because you're remote, you're covering a lot of topics and you're furious, you're writing notes and you're jumping from meeting to meeting and then all of a sudden at the end of the day, you've got a diary full of notes with extremely messy handwriting. And then you've got to sit there for an hour or two transcribing it back into your CRM or creating a task for yourself to follow up on. So there is some amazing and efficient and helpful tools out there, transcription tools that can aid in this. And we're certainly using one in our company called Fathom AI. I've played around with a few others, such as Fireflies, Firefly or Otter AI, which basically join your digital meetings as a note taker. And then at the end of the call, it gives you a transcript for the entire meeting. It gives you a summary of all the points that were covered and the topics that were covered. The better tools also put subtitles to break down the different topics and then finally gives you what outcomes or what, what tasks are going to be generated and who they're assigned to at the end of the call as well. And some of the better tools that have great integrations connecting to your CRM as well. So you would, you would do a meeting, it would transcribe, take all the notes for you, give you all the summary actions in bullet points and Then push it into your CRM. It's all there without you having to type a single thing. That's been a massive time saver. If you have, if you're not using those tools at the moment, definitely jump into one. The 20 or $25 a month subscription fee for a tool like that pays for itself time and time over with the amount of time that you save. And also the. It also saves you from letting things slip through the cracks because it's capturing everything. Sometimes when you take notes, you might miss. You might miss something.

I've noticed that getting a lot better as well. When we first implemented it was a bit inconsistent with the action items and the notes sometimes in terms of what it would recommend to do. And it's just got progressively better over time too.

Yeah, and that's large, large language models or any of these AI, AI tools, they get better as you give it more data. So as more and more people use it and as they provide feedback as well, the models can be improved to provide better outcome. You're right. I've noticed that it's gotten much, much better the first couple of times I use them like ah, the results are not that good and I used to actually have to copy and paste the transcript put into GPT and prompt it to give me the exact extracts that I want. But now it's, I'm finding that I don't have to do that at all these days because these tools are just getting better and better. Some other sort of areas to look at for builders as well is chatbots. Everyone has a website. Everyone was a website. And if you don't have a website you should get one because that's the, that's the shop front of your business in the digital world and that's how customers are going to find you. You know, chatbots have been around for a long, long time and previously they've either been manned by humans. So someone would come into your website, start a chat and then there'd be a human in the office that would be responding. That's generally how chatbots on these websites work. And, and as time went on these got smarter and smarter and they were adding, you know, some, some of these chatbots would have workflows where if you ask a certain question with a few keywords, it would take you down a different path. But now we've got AI chatbots and they've been, they've been quite prevalent in the last year and a half and we've, we've just recently implemented one on the apb Website that basically, as you install it, it scans your entire site, learns about your business, learns about your products and services. You know, you can then give it some more web pages to, to learn and train itself on. So, for example, our blogs and articles, our short snippets on the action plans, so that any customer or prospect coming to our website can ask a question, and if it's been trained on that question, it'll respond almost instantly. I mean, there's a second delay, but almost instantly there was a. There's a response. And as time goes on, we're training it with better questions and better answers when we see that it hasn't answered a question quite well. And at the moment, I would say, you know, five out of every 10 conversation is being answered through this trained AI chatbot. But as time goes on, that number will incrementally go up to 6 out of 10, 7 out of 10, 8 out of 10. And you think about speed to response. You know, that's one thing that we cover in our action plans. When someone, when a prospect calls or a lead inquiry comes in, you want to be able to action that as quickly as possible and give them a response as quickly as possible to build that credibility and trust. And I think these, these, these little tools being implemented across the business can help with that and give your prospects and customers quicker answers and a better service. So that's definitely an area that I'd be looking into. And a lot of these chatbots are quite easy to install, and, you know, they're plug and play if you use.

Can you just touch on the testing process used for the chatbot to kind of make sure that you were mitigating the risk of giving out wrong information?

Yeah, so there was a lot of testing done before we went live, and the testing was literally opening up, opening up the chatbot, making sure the site was trained and asking it as many questions as we possibly could. And luckily, we have a history of all our chats that have come through the apb APB website. So I had a huge sample set to go through and pick out questions that looked a little bit funky or looked a little bit different to see how the chatbot responded. And it did really well. And I think the developers of the chatbot that we're using have done a great job in keeping it very constrained so it doesn't hallucinate and create new, new content. And I think that was the important thing that we were looking for. We didn't want a chatbot just losing its mind and spewing out garbage that was not relevant to our industry or not relevant to our customers. But these guys had limited it quite well so that it only uses the data that we trained on. And if you didn't know an answer, it would actually tell you I'm sorry, I don't know. I don't know that answer. Can you ask a different question or ask it. Ask a question in a better way. And if I can't you, if I can't help you. Furthermore, you can always. The chatbot would suggest to the, to the user type in talk to a human or talk to a person and that chatbot will notify someone on our team that will jump on and be able to answer those questions human to human.

What's the name of that chatbot? I'm sure that's the question on everyone's lips right now.

Yeah, definitely. The one that we're using at the moment is called myaskai and we chose that specifically because it had a very, very nice integration with the CRM that we're using using. So it's almost, it almost feels like you're not using another chatbot at all. We're just using the inbuilt HubSpot with AI functionality. But there is so many other ones out there. Chat base and sorry the name, the other names escape me but if you do a quick Google for chatbots, there's, there's plenty of them out there. Some other ones that would would be useful for our community to look into as well is some of the advanced features of ChatGPT. So not going outside of ChatGPT but going back into it. They're called custom GPTs. They're basically prepackaged. You could call them apps that are within the chat GPT app store I guess you can call it where they are. They have been pre configured by companies or individuals to do very specific tasks. So for example, you'll find GPTs that will analyze a building plan. For example I found one a few months ago. You'll find GPTs for writing standard operating procedures. So if you have a transcript of a process that you've recorded, you can use one of those GPTs that have been pre built without knowing how to prompt it all. Put in your transcript, it'll generate the standard operating procedure for you. So there are a lot of great pre built GPTs out there or you can create your own GPT as well. If you've got a paid account you can create your own custom GPT for use cases that are you that you're doing regularly on A day to day or a week? Week to week. I created one last week called a varant emailer where I where I loaded around about 10 or 15 of my professional emails, not my internal emails that I do with the team. I loaded about 10 or 15 professional emails. I gave it a custom prompt and now all I have to do is speak to my chatgpt via voice and it'll generate an email in my voice, in my tone and with my unique words that I use all the time. So that saved me a ton of time and it's only getting better as I give it more custom emails for it to refer to.

How does that custom GPT work? Is that literally just one prompt or can you create them with multiple prompts as well?

It's generally one prompt, one big prompt. So it gives you an interface where you can put in your instructions of what you want that GPT to do. And it. And there's some Funky features in GPTs where you can use one GPT to refer to another GPT so you can get the output of one and then push into another. Usually what you want these GPTs to do is a single task or a single, you know, a single use case. Trying, trying to get it to do multiple use cases will produce, you know, substandard results. So keeping it to a single use case will get much, much better results because it's specialized so you can use multiple GPTs for different use cases. And the last one, which is another area that we're looking into at the moment, we've got a small project is AI callers. So this is voice, AI voice calls that AI is doing on your behalf. And that has blown up quite a bit over the last couple of months in terms of advancements and the types of software that are coming out to support this. One notable one to mention is Bland AI. It's a Y Combinator launched startup that have basically created a platform where you can go in there, create a prompt, choose a voice, whether it be American or Australian or you know, any other, any other accent that might be available on their platform where you can train your voice if you want, if you want the speaker to be in your voice, you can actually load up your voice file samples and it'll speak like you, which will make, which can make phone calls on your behalf. And the use cases that that's being used, used for is appointment setting, appointment confirmation, invitations to events. And you know, some companies are even using it for completely outbound sales calls, which I don't exactly agree with. But you know, you can see that there's quite a few use cases there. And these tools are starting to get better and better in terms of what you can do with them and how you can integrate them. So, for example, with bland AI, you can create conversational pathways. If someone says no, you know, hey, I'm calling you to invite you to an event, for example, and they say no, it'll take you down a different path and give you a different, a different answer. Or you can integrate it with your SMS tools or email tools or your CRM so that the conversations can be captured, logged, recorded. You can get a sentiment analysis on what the customers have been talking to your AI bot about. And if the customer has responded positively and you want an SMS to be sent with a certain link, you can automate all those functions and features as well. It might not be an easy thing to set up for a novice, but there are a plethora of contractors that you can pick up on freelancer sites such as upwork or freelancer.com that can you pay by the hour. They can get these tools set up for you.

Where do you sit on the ethics of AI calling and how do you think builders or business owners in general can handle that side of it? So they're not. Yeah, not ruining their reputation.

Yeah, great question. And legislation hasn't yet caught up with AI completely, so I'm sure there will be legislation that get passed down here in Australia. And I think there's already been some legislation passed in the US on putting some guardrails around these outbound AI calls. And I think on the ethical side, definitely announce that it's a digital agent of some sort. I think it'd be very, you know, it'd be very unethical to pretend that it's a human that's making the outbound call. You know, it's, it's. I think you get a worse reaction from a prospect if they, if you were pretending that the caller was a real human. It's better to be honest and transparent up front. And that way you don't burn bridges, you don't destroy your credibility or your trust. And I have heard of a lot of cases where companies have completely shunned that and sent these AI callers out without announcing that they are AI caller. And it's super obvious. I mean, there's a, there's a slight delay. You know, technology isn't perfected, perfected to 100% yet. If you're, if you've got your ear out open, you can still hear some of the Pauses and gaps between responses or when you interrupt it, it might speak over you and you'll definitely know straight away, hey, that's an AI call. I mean, they've got their time and place of where to use them. Like I said, I don't think they're, they're not as advanced yet to be using them as an all out individual sales call, but they're certainly getting better as time goes on.

Yeah, that makes sense. And to your point, I think legislation is always way behind the eight ball in terms of where AI is going. So I imagine that's going to play a big part in the next couple of years, would you say?

I have no doubt, I have no doubt. Especially when you start seeing, and it's already happening where people are using these technologies for, you know, nefarious use cases, whether it be spam calls or, you know, harassing, harassing individuals, adding individuals, that kind of stuff. Exactly. So there's a lot of that going on at the moment. It's only going to increase over time. I mean, I rarely pick up my phone these days because I'm starting to get a lot of, a lot of spam calls and some of them, you can tell they're AI and they don't announce that they're AI. When they do announce it, I stay on the phone, I play around with it just to see how, how, how good the AI tool is at responding to weird and wonderful requests.

Yeah, that's cool. Being inquisitive again.

That's it.

Okay, Vrant, where do you see AI? What are your predictions for the future in the residential construction industry?

Wow, That's a big question. Look, we're still in the early days, so I think, I think it's safe to say we've seen nothing yet. There's definitely much more to come at the moment. Across the building industry and construction industry, whether it be commercial construction, there already is a lot of AI being integrated into some of the, some of the tools that the industry already uses. So, you know, project planning, for example, risk cost planning, you know, risk assessment. There's, there's, there's already tools that have got AI integrated into it. But in terms of looking forward into the future, I mean, you've got things such as AI for design, being able to, you know, imagine a future where a builder can put a piece of software, AI, smart software next to them while they're speaking to a customer. And as they're, as the customer or the prospect is giving their preferences, the AI tool can design a house on the fly. Using the materials that the builder uses, uses the designs that a builder is accustomed to and change and tweak it on the fly. Because, you know, at the moment, draftsmen and designers and architects are doing these designs. But as, as AI improves and as software gets better, I see a future where you can actually get a customer to speak to a software and it'll design a house for you on the fly. You know, adding to that virtual and augmented reality, I think is going to be a big thing. I've already seen some use cases where people are dibbling and dabbling in this, but it's not widespread yet where a customer will describe the house that they want, or a builder has designed something in concept and a prospect can put on, you know, 3D glasses and walk through the house that they are going to potentially live in in the future and get a feel for the space, get a feel for what the house looks like, what their appliances look like, how certain furnitures can be set around. I think that's, that's, that definitely will become more mainstream as time goes on. We're also in the age of Internet of things, and Internet of things basically means devices that are connected to the Internet, whether it be through WI Fi or through a SIM card. And we've got these devices all around us. We don't even know. Some of us won't even know that they are Internet of things. So, for example, the, the ring doorbells, you know, the WI Fi doorbells. You ring the doorbell and it rings on your phone, you might not be at home, you might be on the other side of the world and it might ring your phone. I think the Internet of things is coming, is coming in a big way for the building and construction industry because I'm seeing it being used in commercial building, where they have various sensors and cameras put throughout the sites to basically keep track of the project, be able to give progress updates, be able to flag risks and hazards that might be on a work site. There's a really cool company called Drone Deploy that does this for the commercial builders for buildings where they've got drones and the company that is Drone Deploy, where they've got drones that can fly around the site, take pictures, but they've also got now robotic dogs that can walk around the site and just take a. Capture, a visual capture of their surroundings. And using AI, they've been able to train what a hazard might be or what a risk might be. So, for example, a ladder might be placed in the wrong place, and that will, that will flag it with the site supervisor to say, hey, there's some, there's some safety issues on the site. Here's where it is, here's where it is. This is what you need to do to action. So I think that's gonna, that's really gonna come in into play over the next couple of years and give builders, you know, a bit of an advanced warning of things to come or a better, a better understanding of what's going on on site and how that project is proceeding while they're not there and they've got their subcontractors there. And in adding to that, you know, predictive maintenance, I think will be a big thing. We're already seeing an insurance company here in Australia called Honey that with every insurance policy that you take out, they send out a number of sensor devices to detect smoke and water. And that way the owner and the insurance company gets notified when, when it detects that there is something about to go wrong. So there's a high level of moisture, for example, and that comes into the predictive maintenance side of things. So I'm super excited as to what's to come down the track. I think every single industry is going to be disrupted. Slash. Improved, Slashed, you know. Yeah, disrupted is the main word that I'm thinking of.

Yeah, I like, I like reframing it to improve. Because for all the builders out there, if you are keeping up to date with the latest trends in AI and technology and making sure that you're applying these things before your competition, you're going to be ahead of the game. You're going to be able to, you know, it's not going to fully replace building companies. They're just going to look different. You know what I mean?

Absolutely, absolutely. I think the earlier people get on to these new trends. You know, there's that famous adage and that saying that keeps going around, especially in the last year and a half, AI will not replace you, but someone who is using AI is going to replace you. So I think the builders and the home builders that pick up some of these innovative tools early on will really be able to catapult the competition just because they're providing a better service more consistently faster. And that's going to lead down to making sure that their projects are delivered faster, which means ultimately, happy customer in their dream home without complaints. And I think that's going to be the big difference between those who adopt it early and those that don't pick it up until. Until the masses have picked it up. And once the masses have picked it up, there's not going to be a big difference between individuals.

Absolutely. Well, Ferrant, I've taken an hour of your time and I've absolutely loved every minute of it. You've provided so much value. You're so passionate about this topic and it's really come through today. So thank you so much for coming on the show. Is there anything else you'd like to share with the builders out there before we wrap?

Up, up. Thank you for having me, Will. I've enjoyed this conversation as well. And yeah, the only thing I would want to pass on to anyone that's listening, especially if they haven't dibbled and dabbled into it, is AI is not scary. It's going to be synonymous with our life from here on end. The same way that social media made its presence in, you know, 2006, 2007, and we thought, wow, look at this Facebook, look at this, look at these new social media platforms. I think AI is here to stay and it's going to only going to increase over, over the next couple of years in terms of how it infiltrates our life. And I think getting onto it now, even at the smaller, at the smaller level, playing around with it, experimenting with it, jumping, putting it on your phone, using some of the free ones that are out there. I think most of the big language models, ChatGPT, Claude as well, they've got a free version, so it doesn't cost you anything to start. All I would suggest is get started, get started now, because if you don't, you will get left behind. And it's a, it's, it's, it's the next big thing since the Internet was released to the world. So I'd be jumping onto it now. It's an exciting time.

Great advice for the builders out there that want to know what to do next. For members of the association of Professional Builders, we have an AI action plan. We'll have another iteration of that coming out later this year, so that'll keep you at the forefront of the industry. We also have a AI toolkit available if you're a non member, so you can get access to that. For Australians, head to the Professional Builders Summit to hear Vrant talk about AI. Obviously, you know, jump in and listen to the podcast that Vrant mentioned, the newsletters he's talking about, and just try to stay on top of everything as much as you can. Would you recommend them, them doing anything else?

No, I think just getting started is a big step and definitely finding some of those newsletters that you can vibe with and a few of the YouTube channels and influencers. There's quite a few great influencers on LinkedIn that constantly produce short snippets of AI goodness that you might be able to get an idea of where to implement in your businesses. So definitely advise jumping onto those platforms and following the experts that are sharing a lot of these contents because yeah, that's where the idea start.

Well, that's the one thing I forgot to mention actually. You're doing the same thing in our Facebook members group as well. Weekly sharing AI tips and use cases in there. So that's another place.

Yeah, definitely bump into a members group or if you're not a member yet, become a member and you'll be.

Exactly. That's what I was getting at. Awesome. Fran well, thanks so much for coming on the show today. I've had a ball.

Thank you. Will appreciate you having me on.

Also thanks to all of our listeners out there, wherever you are in the world. If you like the show today, please subscribe to Professional Builders Secrets on your platform, platform of choice and if you're feeling generous, leave us a review. But until next time, have a great day. This episode of Professional Builder Secrets is brought to you by Apparatus Contractor Services. Apparatus provides full scope, end to end QuickBooks, Precision Construction accounting and CFO services for residential builders. If you feel like you can't completely trust your weekly and monthly financial reports or you feel like you're flying blind, Apparatus Root will rebuild and operate your QuickBooks accounting system and give you precision construction accounting capability you can trust. Learn more@apparatusteam.com or on the APB Rewards Partner Directory.