Painting Businesses + AI: What Actually Moves the Needle

Jon Bryant: Hey everybody, welcome back to Price. Sell. Paint. I am Jon Bryant. This is Michael Murray. How you doing today, Michael?

Michael Murray: Hey, what's up Jon? Doing wonderful, how are you?

Jon Bryant: Wonderful as always - maybe not as always. Today's pretty good though. I know we've been chatting a little bit about AI and the painting industry. I think everybody that I see chatting about things online is also chatting about AI and the painting industry. I think it's a fun topic. It's one that's interesting. I know you and I both have a lot of thoughts about this. Maybe a good place to start is we've been talking about what's going on in the world today. It is October 2025. Fill us in Michael. You were telling me some news on the painting industry and I think it's quite interesting and it leads into a discussion about AI for sure.

Michael Murray: Yeah. There's a lot of news recently about layoffs and different companies - Sherwin Williams, we were talking about 401ks and different things have been going on. It just seems like there's a lot of conversations around in our industry about maybe some softening, let's just say, where maybe the economic trends aren't in our favor. We have to maybe be a little bit more creative as to adding overhead and things like that. That's some of the conversations that we've had at Textbook and trying to figure out creative ways that we can use AI.

I think there's a lot - the rate of change that's happening in our society at large is increasing exponentially. We're having a conversation around AI. To your point, it seems like kind of everybody's talking about it now. But this is something that I've really loved learning about and playing with since it came out - November of '22, I think. It's something that we've been trying to figure out ways to use since that point. I feel behind, and it's because of the rate or the pace of change and how quickly models are coming out and just new capabilities are coming out. It's like, man, this is hard to keep up with. I think the future is kind of crazy when you think about if it's only been a few years from when this whole thing started, what's it going to look like in three to five years from now? That's almost unfathomable to a level, but it's fun to think about.

Jon Bryant: It is fun to think about it and hopefully we can think about it today live with people listening. I think certainly not. I've experienced it from AI painting contracts perspective, as well as from a software perspective. It's interesting. I'm not as maybe scared as most people are of this exponential future. I think there's a lot of hype. There's a lot of "this is going to change everything." I'm just not sure that's true.

I know here in Canada, cannabis became legal. There was - no word of a lie, Michael - there was like five cannabis shops within a block of my house that opened up. I know people are excited about this, but five stores? That's a lot. You go to the stores, you talk to people about the stores or whatever, and it was like everyone is like, "This is it. Everyone is getting in." Give it two years and the major manufacturers - the stock has plummeted. There's one store barely left. It has a similar ring to me of hype where everyone goes to the next shiny thing and this is that shiny thing.

Now don't get me wrong. There's a lot of power here and a lot of things that if it were to go away tomorrow, I'd be sad. But for our industry, I think understanding what the potential is and how to use it in a really simple way to start is important. Get a grasp of it.

In my world and the story we were talking about earlier was this conversation I had with somebody who was at a software company. They're in the talent acquisition space. We were talking about why they might be interested in getting out of that space. What they said was there's just a lot of over-saturation and it gets harder to grow. AI is empowering people to do more with less. That's going to cause a problem down the road for companies like ours.

Looking at the trades, they had just had this experience of trying to do a repair on their home. They had purchased a home and it had some building envelope problems. It was requiring them to take the stucco off and all the insulation and go all the way down to the studs. This person - it was their first home and they thought, "This must be pretty straightforward. I'll call some companies and have them come out."

He said what the experience actually was was the most eye-opening experience of - what would you call it even? They've been operating in a software world where they thought it was all obvious and straightforward and they came to what felt like it was the caveman era. The stone age overnight. It was "let me come out to your house in a week or two. I'm going to show up late. I'm not going to tell you maybe I'm coming even. I'll show up. I'm going to give you an estimate in a couple of days. The range that I'm going to give you is between $10,000 and $120,000 and it might be more, we're just not sure." He was like, "This is crazy. The amount of opportunity here for trades to just advance a little bit - there's just a little bit of low hanging fruit here - is absolutely absurd."

Michael Murray: And you said too though, how long did it take to get the quote?

Jon Bryant: It was like a couple days, maybe a week. Nothing was on the spot. Nothing was concrete. Like "this is it, sign me up for this." It was crazy. So I started to think about this problem that we're talking about - AI and everyone's talking about this magical world of the robots are going to start painting people's homes or commercial properties or whatever. And I'm like, well, maybe we could just get a lead intake form and have someone show up properly. That would be wonderful. Getting wild.

Michael Murray: You're getting crazy. Now you're just getting crazy over here. My gosh.

I had some tree work done at our house recently. I had some trees taken down. It's like the exact same experience. We get a quote, pay per form - it's like $5,000 or whatever to get these trees taken down. "Okay, cool." So we signed up. They don't ask for any money. It's just like, "All right. We'll come out when we can come out." "Okay." No conversation. Nothing happens for three weeks. Then I joke you not at 8:45 at night, I get a text message. "Hey, we can be there tomorrow morning." I'm like, "Okay. Luckily I've got some work flexibility. I guess I'm working from home tomorrow. All right, what time are you guys going to be here tomorrow?" "We'll be there at 8:30." "Perfect. Okay." 7:45 in the morning they show up. And they're done early afternoon. It's just like, this is crazy. My first thought was like, "Man, I need to go into the tree business because there is a lot of opportunities to turn up some customer service and sophistication here."

Jon Bryant: Totally. So I think it's wild. Here we are having these high level discussions of what AI is going to do and let alone you can't get an accurate time when someone's showing up. We can't get estimates. They're not even digital. There are people who can't even answer the phones. It's like, wow. The level at which we're thinking here is pretty high. Not to say that AI can't be helpful. For sure it can be, but I think there's some real tangible things we can do to use it to our benefit to just do some simple things right off the start.

Michael Murray: It's interesting. I thought what we'd talk about today is some of the ways that we're using AI currently, as well as just how we're thinking about it going forward.

Jon Bryant: Sure, yeah, sounds great.

Michael Murray: I had a conversation with our leadership team here at Textbook just this week, and then it became a side conversation with our HR manager about what I think is the future in the painting industry and how we need to use AI. Let me just recreate that here. I think it's really relevant. I don't know that I'm thinking about it the right way, but it's how I'm thinking about it.

It goes something like this. I think that there's many other industries that are already being impacted - many professions where you don't need to hire as many people because AI can augment or superpower what one person can do. They're more capable. If one person can produce the same outputs that maybe two could before, we need less people to produce the same outputs. Then I think what's going to happen is talented, intelligent people are going to start to look for other places to go work.

I think one of those places will be in the trades. Similar to the experience with the guy that you said that you were meeting with, people are going to look at these industries like they've been doing for at least 10 to 15 years with HVAC and others and saying, "Hey, we're going to go over there." I think there's some unsophisticated competition. I can go over there and do well. I think that's going to continue to happen, but I think it more so happens than maybe our industry specifically has experienced.

Then the first thing that they're going to do is fix what we're talking about. They're going to come in and compete on marketing, sales, some communications, sophistication, some of the things that we think are kind of no brainers - like, "Yeah, we're going to tell you when we're going to show up at your house." That sounds pretty good. Then what's going to be left if we're all pretty good at marketing and sales - which we're currently not but if we were - what would be left is who can actually execute? Who can send the crew members that are the best that are going to do the great work that are going to provide an awesome experience? Who's going to look nice and smell nice and talk nice and do all the things that we all know, especially in residential painting, that is kind of important?

There's actually going to be - and again, I think this is already happening. I think it's going to speed up - but there's going to be an arms race towards who can hire the best people to actually do the work. Because again, getting the work is going to be relatively easy, which I would argue it kind of is. Not to say that sales is necessarily always easy, but it's just compared to hiring and delivering the services.

Okay, if we think that's going to be true, what that means then is we're going to have to pay our painters more. Which again, I would make the argument we're already seeing. I know that's been our experience at Textbook in the last five years since COVID. Wage pressure is just very high. If we want to hire somebody to come in here and is an awesome person with or without experience, we need to pay them significantly more than we did in 2017, 18, and 19. You agree with me so far? Make sense?

Jon Bryant: So far I'm on the same page.

Michael Murray: Yep. So it's like, okay, cool. If we say we're going to have to do that. So then what are we going to do? Well, we can raise our prices to afford to pay painters more, which we should always be doing. There should always be the question of how do we raise our prices? How do we deliver more value? Both of those things have to happen at the same time or else we have problems. We talked about that before. We could make less profit. If we pay our painters more, we don't raise our prices, nothing else changes, we make less profit. I don't know about you. I don't like that answer. So I'm going to cross that one off the list.

And then the third one is I have to be able to deliver my services with less overhead. For a company like ours and PaintScout and there's significant overhead involved. It's like, how do we more efficiently do what we're doing? The answer to that, I think is, okay, how do we use AI in our painting business? Not to replace the painters, but how do we use it to augment, not necessarily replace, but maybe replace some of our overhead expenses. How do I bring down my overhead expenses as a percentage of revenue so that I can shift more of that revenue towards paying our crew members? That's the whole paradigm shift that we're focused on right now.

Then I go to marketing, lead intake, admin costs, vehicle expenses. We start to look through what are the big overhead buckets and what are ways that we can increase our revenue without necessarily having to increase those expenses at the same rate, hopefully by using some technology and things like that. Not just asking people to do more, but actually use technology so that we don't necessarily have to hire a second person to answer the phone. The first person can be more efficient within that role and things like that. That's how I'm challenging our team to start to think about AI in all the different facets of the overhead of our business.

Jon Bryant: There are a few things come to mind there. One, I think obviously we see this already. There's a massive amount of people needing the trades these days. There's a lot of jobs that are going to be transitioning because of these types of things that every business is experiencing. Right now though, there's a lack of supply for trades and demand I think is going up. Generations, younger generations aren't doing it themselves. DIY is a little bit of a lost art. So that's going to offer even more opportunity to increase prices, which will allow us to have these types of good systems behind the scenes and that professionalization will keep increasing. I agree with all of that.

I think learning to use these things - and I know a lot of our podcast is about sales and how do you use this to be more efficient from - there's obviously the business operations side, there's lots of opportunities. But from a sales perspective, if you're a sales rep listening, do you have a sense of specific things you're encouraging your team to do, Michael, with AI or tools they're using to amplify that for them?

Michael Murray: For sure. I think that sales is a big one. If I'm a business owner, I'm just thinking about the sales side of the business. Let's just say, what would you say would be a top line sales number for a residential repaint rep in a single year? How much should one rep be able to sell?

Jon Bryant: What year then?

Michael Murray: Let's say they're somewhere in maybe not first year, but year two or three, somewhere in - not super seasoned, but definitely not new either.

Jon Bryant: $1.5 million.

Michael Murray: That's where my number was too. Okay. So historically then as a business owner or whatever, we would just do that math. If I'm going to run a $3 million top line revenue business, I'm going to need two, at least two reps, somewhere in that neighborhood. I have to start to think of that exponentially and say, "Okay, well, what can I do? How can I get a sales rep to sell $2 million?"

Well, what we've often thought of in the past was like, "Oh, I know we'll just get them to work more. They'll just do more estimates. They just did - what if we just, instead of working 40 hours a week, we're going to work 50?" It's like, okay, well that could work. Problem is that there's fewer and fewer people, especially younger people that are like, "Oh, tell me how I work more hours and I'm going to sign up for that job." Myself included. That's not my goal in life either. I'm not judging anybody.

So it's like, okay, so then how can we help our sales reps just to be more efficient with their time? For an example, we're currently building custom software to help us with our lead scheduling and route planning. We have reps that will often drive - our territory is about an hour in any direction. Meaning from one end to the other, it could be a two hour drive. Now, typically we're already better than that, just with some very simple thoughts or whatever. But it's like, how do we eliminate just some of the driving using AI and Google maps and some other tools to route plan better?

In other industries, this stuff is a little bit more common. But we haven't necessarily felt maybe the pain and the problem of that because it's like, "Meh, it's not that big of a deal. Our sales reps are selling really well and they're driving around a little bit. Gas isn't that expensive. It's not really the biggest problem to solve." But if I start to think about it, it's like, what if I could get my rep to just limit drive time, if they could just do one more estimate a week? It's kind of staggering.

If you start to think about, well, if they sold half of those and they did, let's just say 50 more estimates in a year, that's 25 more sales. Everybody's average job size is a little bit different, but you're somewhere probably between $100,000 to $150,000 more, pretty easily, if not more. That's with no more extra effort, no more time, just eliminating some drive time by some better route planning. That can be a way that we can use a tool like AI or whatever to be able to do that. One of many ways.

Jon Bryant: Man, I'm so curious to see how that comes together for you. I've thought of this problem so many times. I think the variables of fit of sales rep, how their day is planned - there's a bunch there. So I'm really curious because I think you're right. That type of stuff should be possible. It's just been - we've chalked it up to the variables cause it to be difficult to get right. Because we don't have the same volume. If you're an HVAC plumbing type outfit, they have six stops per day and they're all over the place. Whereas we might be talking about - like you said earlier - it's a small data point and you're like, well, there's three today. It's not that hard to, if we mess it up, it's no big deal.

But I will say that my brother who has been on the podcast before and people know - route planning takes a lot of time. There's a lot of driving. If he could save that time - I think he was in the car the other day, he was in the car for four hours. It's like, well, that's not a good use of time. The virtual side of things is not effective for what he does. So he has to go see these things. Would it be nice if he could just batch a little better? So I see the value for sure.

Michael Murray: One of the tools that we've been using a decent amount - I would say we're a little late to the game with it. I put it off thinking it wasn't really going to add value. That's one of the things that I'm really conscious of with AI - I don't want AI to just add more noise and stuff. I went to this AI workshop and all they talked about was you can use AI for marketing and lead generation. It's going to be awesome. You're going to pump out five podcasts a day and 10 videos and 50 blog posts. I'm just like, nobody wants that. Why do I need to put out 10 YouTube videos a day of just complete hot garbage that is all AI generated?

Maybe that works for lead generation for six months. Then it's only a matter of time until Google and YouTube and wherever we're trying to generate these leads from just get sophisticated enough to say like, "This is just crap, we're going to stop showing it." It's like SEO was like that back in 2006, seven and eight, where if you just put the word "painting" on the web page enough, you were going to rank well. I think a lot of us can still remember those days and it quickly just went away.

Jon Bryant: Those were the good old days, when you could just stack a ton of words all in the same color as your background. It's like we have a white background and I have the word "painting" 7,000 times in white.

Michael Murray: That was Ninja level. What you would do is you build a course teaching painting companies how to do that with AI and that's how you make money. So easy. Anybody can do it. Anyways, it's just like, no, I just don't want to create more noise.

Anyway, I was getting onto this - the tool is really just through Company Cam. I love the PaintScout Company Cam integration. I know a lot of PaintScout users are also users of Company Cam. But their AI features are pretty good. They have these walkthroughs essentially, I think that's the title. It's awesome. You can just talk into the video and it builds like an output essentially for the production team, where they can better understand what it is that we're going to be doing on the project and see the photos and different things like that.

Where before somebody might have had to - we might have asked a sales rep to, "Hey, on the first day of this project, if you could just show up and walk the crew through, make sure everybody knows what they're doing. This is a big project. We don't want to have any issues." It's like now with tools like this Company Cam AI features, we can get the video, get a bunch of pictures and create this detailed output that the crew leader can then have and really help to streamline communication. It can be a lot more asynchronous. We don't have to get everybody's schedules to line up and free up the sales reps to go focus on selling work. That's a tool I would say that we're new to, but I'm seeing a lot of potential there.

Jon Bryant: Yeah. Communication is a big part. AI for checking emails, generating better communication on emails, especially that note taking element is really strong. I think it's such a great use for what we do, which is sales reps spend so much time on that, just trying to communicate the job. Yeah, it's super helpful.

Michael Murray: One of the tools that we haven't talked about, which I think is the one that probably many people have heard about, is just the voice recorders. There's a handful of these - RILLA, Ciro, Craft, which is a newer one that I wasn't as familiar with, but they all do similar things, where essentially we're going to record the entire sales appointment and we're going to give you some AI-generated coaching and feedback.

We have - I'm open to trying that. We have not done it mostly because I feel like it is way expensive. I also know that it's just all of these technologies are built on these AI large language models. They don't - they're not using their own large language model. They're building it off of these other tools. I just can't justify spending $10,000 plus a year on something that I know is just being built on the software that I have at my fingertips.

But I do think there's a lot of potential if used correctly and embraced in different things where we can get coaching and feedback and eliminating a lot of the sales manager ride-alongs. I think that can be another awesome tool in a way that we could augment the sales aspect of our industry.

Jon Bryant: Totally. To your point though, I'm pretty sure you can just record on your phone, upload to ChatGPT, and it might do the same thing to some extent. I'm totally with you there. I think the value is diminishing quickly because of the availability of other things that do the same thing. ChatGPT 5 will take that stuff in in a second and just give you feedback. If it's like, you might want to load it with some type of a sales system and be like, "How did they do based on this system?" That's not a hard thing.

Sometimes I see some of these services as packaging it nicely so people understand the prompt they should have used in ChatGPT. Maybe I'm wrong. Maybe I'm saying it all out loud and if there's somebody listening from those apps, I'd love to hear more as to why there's a special sauce to it. Because it's possible. I'm just not educated enough on it. But I've used ChatGPT to summarize lots of different things. Even our own podcast - tell me what we're doing here. What actually did we talk about? Because I can't remember. It's been 50 of them now, Michael. So 50.

Michael Murray: 50? Man.

Jon Bryant: 52, I think. We hit our rough. Yeah, super rough.

Michael Murray: Hopefully we're getting better because I'm sure those first 30 weren't any good.

I'll tell you another tool actually that we're using AI and I think it's a cool use case. I think I talked to you about this months ago when we first did it. I created a custom GPT for our sales team, which is our Textbook Sales Coach GPT. I put into it all of the transcripts from our podcast. I put in all the transcripts from the PaintScout X conference videos, some other content that we had over the years on just our sales system and sales training and just written things that I had. Some other sales training materials. I trained it all on that. We use that in our sales meetings for role playing.

I encourage our sales team to use that while they drive around. They can use the voice mode. Role playing within the custom GPT is an awesome way for our sales team to sharpen their skillset a little bit.

Jon Bryant: We've talked about this, but I think for people listening, what you did is fascinating. Maybe you can just walk through a couple more steps of that because it's super valuable. I've built custom GPTs. It's just a funny term to say - a custom GPT. What does GPT stand for? Do you know?

Michael Murray: We should probably answer that question accurately instead of guessing.

Jon Bryant: Because I've thrown this term around a lot and so many times I'm like "ChatGPT." What is that? Can you ask ChatGPT what GPT stands for?

Michael Murray: According to Gemini, because I just Googled it, GPT stands for Generative Pre-trained Transformer. So let's go. I got a second source that is telling me that that seems accurate. So I'm glad I didn't guess because I would have gotten that wrong.

Jon Bryant: There you go. Generative pre-trained transformer? Wow, that's technical. But anyways, we've created these and we're acknowledging we don't know everything, which is always special. So moving on quickly.

Michael Murray: I think most people listening should already know that, but anyways, we'll move on. We're just a couple of knuckleheads that are talking about stuff. We don't pretend to be experts.

Jon Bryant: But having developed these now a couple of times, they're super helpful. So maybe just walk through your process. Because maybe it might be a little different.

Michael Murray: I think again, this is, I believe, only available if you're paying for ChatGPT. So the $20 a month plan. I would suggest that's - if you're not paying for it, just do that. Just start paying for it. That's one of the things I'm encouraging anybody who's getting into AI - just pay for it. It'll be a forcing function. It's going to force you to learn how to use it better. It's also going to make it more comfortable when the cost starts to go up. And I think it will.

Anyway, so we have the paid plan. There's a custom GPT section within there. There's a database of publicly available ones. You can find them on the Sandler sales system. You can find them on EOS for business owners that are fans of Traction. There's a lot of custom GPTs that have been created by others that are made public. Or you can create your own.

You can create a custom GPT that's private only to you. You can create one that is private - if you have the link, you can get to it, which is ours. Or you can create a public where it's available, you can search within the database or whatever. Then really what it's all about is you're just pre-training it on your specific information. So instead of just all of the large language model training data, which it still has access to, it's like, "I want you to specifically reference our steps and our sales system. Reference how we teach how to handle a price objection or different things like that."

We have a custom GPT for HR where we've put in our employee handbook, we put in our job descriptions, interview questions, our core values, all that kind of stuff. Then we use that - specifically our HR manager mostly uses that to create job descriptions, interview template questions, and things that are relevant to our core values. They're not just - you could go into ChatGPT and say, "Hey, give me 10 questions I could ask somebody in a painter interview," and they'll give you some. But if you pre-train it on specifically the things that your company does and is looking for, you're going to get a better output. That's basically the idea of a custom GPT.

Jon Bryant: So when you say train, you're just giving information and saying act like you know this.

Michael Murray: Just you're giving it - you're literally putting in tons of information, depending on whatever you want that to be. For the sales one, it took me probably an hour or two. I'm getting the transcript. I'm downloading it, copying it into a text file, I believe it is, that I'm uploading that or copying and pasting it, whatever that looks like, into the custom GPT to have it reference that later on in conversations.

Jon Bryant: Has it been helpful for training? You notice a difference?

Michael Murray: Super. It's the amount of, again, just amplifying your ability is huge. Now you don't have to sit with those ride-alongs as much. Sales reps are getting answers and that's a perfect example. I love that example for what we do.

I think an awesome use case is you're on an appointment and a customer stumps you with something, with an objection. Whatever it is, it's just something that you just don't feel like you handled it all that well. They said, "Hey, we're going on vacation next week and so I'm not going to be able to make a decision for a few weeks," or I don't know, whatever it was. You're just like, "Man, I could have handled that better. What could I have done?"

You can just pull up your custom GPT within the ChatGPT app and go into voice mode as you're driving to your next appointment and just say, "Hey, I need you to role play with me. I need you to give me some coaching. This came up on my last appointment. The customer said this. I don't think I handled it really well. What are some other - give me three other ways that I could have handled this better to get to a better outcome, to get to closure, to have been able to sell the job? What might my next - what should my next question have been?" Different things like that. Using the voice mode while you're driving, I do that all the time. I find that to be an awesome way to use this technology.

Jon Bryant: That's super cool. Man, what's left? There's nothing else. That's the end of AI. Okay, yeah. No, there's so many things.

Michael Murray: What else we got going on? Well, actually, I thought of one more thing. That's the end of it. So no, there's like a million things. This is as we sit here longer. I'll give you another way that we're using it.

We're using it for before and after renderings. We were using ChatGPT. Now we just, in the last little while here, we switched over the last couple weeks to Google Gemini with their Imagen 3, which is just fun to say, which is their image generation tool that came out, I think, in September. It is incredible. If you're not using it, if you've never done that before, it'll blow your mind. Just take a picture, a before picture from a project or whatever, and just ask it to paint whatever.

So kitchen cabinets - we do a lot of kitchen cabinets. Here's a picture of somebody's wood cabinets. I want you to paint them with the Sherwin Williams color naval. And it's like 10 seconds go by and it's like, "Boom, there you go." And it is incredibly perfect. It's to the point where it's like, "Are we sure that's not actually just an after?" It happens instantly. Our sales team is supposed to be doing that ideally on every single appointment. It's very impactful. What we do is very visual. To be able to show a homeowner, "Hey, this is what it could look like if we did this for you," is pretty cool.

Jon Bryant: Have you been trying those on estimates for customers? Is that what you're doing with them?

Michael Murray: That's what we're doing. We're doing it live at the estimate.

Jon Bryant: At the estimate and they're still on their phone. They take a picture, load it in Imagen 3. Yeah, cool. That's really great. The before and after - we've struggled a little bit, at least in my past life with trying to get before and after images. Then you've got to send somebody there before and someone there after. It probably is easiest to get them there after, but you don't get the before picture. Maybe if in a perfect world, we've seen them before and then we just say, "Imagen 3, change this to what we did." That could be really good because it's hard. I mean, to go in the opposite direction would be difficult. Like make this look like it was painted in 2000.

Michael Murray: One of my favorite - we have this picture. You know, just getting actual before and afters is hard. We try to get our crew members to do this in Company Cam or whatever. We have this one kitchen we did probably four years ago. It was beautiful. It's this amazing transformation. It's one of my favorite - the countertops and it's just so - the work we did was just awesome. In the after there's a picture, so prominent in the foreground, of a Chipotle bag just sitting right there on the island. I'm just like, it's so frustrating because I can't get - it's been this never ending battle for our crews to not have the five-in-one sitting in the after picture or whatever.

I was like, "You know what? I wonder if I could go in." I think I was using this a month ago. I think I was using ChatGPT at that point. I was like, "I wonder if I can ask you to just remove the Chipotle bag from this picture from four years ago that's always driven me nuts." And it was just like, "Boom, there you go." I'm like, "Wow, that's incredible." I can just give me a decent after picture and I'll fix it all in post-production using ChatGPT. Because I don't know how to do that in Photoshop or whatever. I'm not smart enough. But I don't need to be anymore. That kind of stuff is mind blowing to me. It makes me so happy. It's the simple things.

Jon Bryant: It's so powerful.

Michael Murray: I don't know if you saw it just recently - I really geek out on this stuff. I love learning about it. But OpenAI just put out a study that they did on basically showing essentially how viable are AI models in replacing the actual output of real world business scenarios. I think it was called - I think it was GDP Bench or something like that. It just came out this week. If you search for it, you can find this and reference what I'm talking about here. I'm not going to do the best job of explaining it. It's pretty technical. I think it was like a 30-page paper. I took that 30-page paper, put it in Notebook LM and learned all about it, or as much as I could I guess.

What was really interesting - they took some of the biggest industries, the biggest segments of the United States economy - healthcare, government, whatever. They said, "Okay, we're going to do a blind test. We're going to take real humans, experts - on average, these people had 14 years of experience in their fields. Marketing experts and all these different things. Then we're going to give the same instructions to all the different AI models and we're going to compare them in a blind test." So it might be create a marketing plan or create a customer service engagement or whatever, handle this customer inquiry or things like that.

Then they took that and they put the outputs in front of another set of industry experts. And it was blind. So you don't know which one's which. What was interesting is that about 47.5% of the time, the experts chose the AI generated output. More than half the time, we're still choosing the output that an expert in the field with 14 years on average experience is putting out there. But half the time we're not. Those are based on the current AI models. It feels like every two or three months we get a new one.

It's kind of crazy how quickly that percentage is going to go up where people are actually saying, "No, the AI output is actually better than the human expert" in these real life examples, things that people do every single day in these industries. It's kind of crazy how quickly that kind of stuff is happening.

Jon Bryant: Yeah, it's so - to the people listening, we've got this level of pen and paper, get back to you in two weeks. Then you've got what you're talking about. One of the biggest barriers here about AI in general is having people learn how to use it. You mentioned a little bit earlier - paying for ChatGPT, force yourself into a world where you have to learn some of this stuff, just to get a feel for what's going on.

Because yeah, you've got these crazy models, these things happening. Then what actually happens in real life takes a lot longer to incorporate. I sometimes think about our business and getting people to change is really hard. Our encouragement has been get to know AI. If you don't, you're probably going to lose your job at some point to someone who does.

Michael Murray: You'll get left behind for sure.

Jon Bryant: And it's cool to think about all the things you can do, but just start doing something with it today. That's been my major suggestion, because we see these crazy things, especially on the software side. We're just like, "Hmm, that's interesting. I wonder how that's going to go. That's wild."

One of the interesting things too - this is a total offshoot of this conversation - but it's where the economics of AI goes. I don't know if you've done much reading on that or how it works, but you made a point about ChatGPT being 20 bucks a month. It's just like how much money they're losing and how much power goes into these, essentially writing a nicer document. You start to look at it being like, there is not enough power. There's not enough capital, not even close. We're running on this world of "all this is possible." What's actually economically possible? Are you willing to spend a thousand dollars a month on ChatGPT?

Michael Murray: Am I? Yeah, I am. Yes. But the point is going to be though, that's why I'm saying to somebody you've got to start spending the 20 because if you don't, you're never going to see the ROI on that. Then when somebody else is willing to spend a thousand, you're going to look at it like, "Well, that's crazy talk." It's just like, but if I'm already getting more than a thousand dollars a month of value out of it, when they raise the price to a thousand dollars, I'm going to be like, "Yeah, okay. I'm already getting that."

I want to assume if I'm getting more than $1,000 of value out of my $20 subscription, holy crap, what kind of value am I going to get when I start paying for the upgraded super duper one?

Jon Bryant: It almost becomes like a differentiation factor. I mean, there's all these big macro questions of can you actually find - can the customers support a thousand dollars a month? But to your point, it's like, get involved now, start seeing it from the sales rep's perspective. If you're a sales rep, there's huge value. I think even, we can talk about PaintScout too for a minute, which is figuring out how to write better proposals, getting estimates up and running faster, using that scheduling type functionality that are all - those things play with this type of functionality. You've got to start just getting into it a little bit.

Even little things like identifying prospects, which is pretty easy to do with the right prompts. Here's 10, 20 companies we've worked with - give us 10 or 20 more. That type of thing is great. For business owners, the systemization, the training, the backend knowledge bases that you've created, which I think is really a great starting point for most people. It's just cool. It's cool to do. And it's so helpful and such an easy way to get value with this stuff, I think.

Michael Murray: Again, I don't believe I'm still in the early learning of what's capable here. I'm trying to force myself to just use it more because I feel like I have way more to learn than I already understand. That's based on the current capabilities, which next week or whatever, it's kind of like double in capabilities. And it's like, "Oh my goodness, this is crazy."

Jon Bryant: All of a sudden AI is painting houses. And that's a scary world.

Michael Murray: Yeah. I mean, I would say when AI is painting houses, all of the other things have already been replaced. Everything that happens on a computer has already been replaced. Nobody's interacting with a computer the way that we currently do. That's not a job. Will AI come for the painter's job? Maybe, but it will be the last one taken. At that point, it'll just be some - either societies and governments have figured that out or we haven't. Then you can go and read about that side of this whole story and it's a little scary. But yeah, I mean, the power consumption is crazy enough to your point.

Jon Bryant: Can you imagine how many people will be available to paint? How many - your production issues will be solved. There'll be no problems. But then the problem is everybody can paint their own house. So it's kind of a chicken and egg problem.

Michael Murray: Yeah, except the problem is that nobody wants to do that work, which is what you talked about at the beginning. We decided as a society, decades ago, that manual labor isn't a worthwhile task or whatever. It's not a great career path, and it's just not desirable. So now there are active efforts towards changing that sentiment, but it's going to take some time. I don't know that it's going to happen fast enough.

I was reading something. It was like, there's literally trillions of dollars is what it's going to cost to build the power plants to make the machine - to make AI work in five years. So if you're smart enough to figure out how do we create power plants that are a hundred times more efficient than current nuclear power plants, because that's going to be something - I have no idea how to do that. I can't even guess. But that's going to change. There's no way that we're creating power in the same way that we currently are because it's not sustainable to your point. That technology has to increase or that will be the limiting factor of what we're able to do with this stuff.

Jon Bryant: It's interesting that I always thought The Matrix was a documentary. Here we are realizing it was. So the machines need the power and we become the power source. Can't wait for that day.

Michael Murray: Oh man. It's like, is that I, Robot or some movie? All these movies from 20 years ago or something are starting to become reality. If anybody was wondering, this was not an AI generated podcast, which is a thing now, with Notebook LM. I just had it create a podcast for me out of that report. And it's just like two people that educated me on the topic. So maybe one day the PriceSellPaint podcast will just be AI Michael and AI Jon. But until that time comes, we'll continue to get together here for a couple hours and chat.

Jon Bryant: I thought the future was friendly. Turns out it's not. Michael, thanks man. If you're listening today, if you enjoy this content, feel free to give it a like and subscribe. As always guys, thanks for hanging out with us and we'll see you again soon on the PriceSellPaint podcast. Thanks Michael. Peace.

‍