Episode Transcript
Kyle James (00:01.006)
Hey, welcome to the Adix Chronicles podcast. I'm your host, Kyle James. And today we're going to be diving in headfirst into how a marketing agency called GrowHouse is using AI inside of their own business. And we'll share the exact steps that you can take in order to implement AI for yourself. Now, before I dive into that, listen closely. Are you looking to implement AI inside of your own company or just struggling to get your AI to stop hallucinating? Speak to GPT Trainer.
GPT Trainer literally builds out and manages your AI agent for you, eliminating hallucinations for good. Go to gpt-trainer.com. I promise you, it'll be the biggest time saving decision you've made all year. Trying to set up AI on your own is like trying to build a house from scratch. Sure, you could do it, but the time and frustration it's gonna take you to get it finished just isn't worth it. It's a thousand times faster and safer to hire professionals.
Once again, that's gpt-trainer.com to schedule your consultation today. Say hi with me, Drew Donaldson, who is the founder and CEO of GrowHouse, a data-powered marketing agency that focuses on hyper-targeted lead generation, AI-powered automation, and return on ad spin optimization. So excited to have you on the show today, Drew. How are you doing, my friend?
Drew Donaldson (01:17.181)
I'm doing great, Kyle. Thanks for having me.
Kyle James (01:19.276)
Yeah, super excited for this conversation. give us some background. You know, what exactly is Grow House and like, how did it, how did it come to be?
Drew Donaldson (01:27.379)
Well, like a lot of marketing agencies, we've really evolved over the years. mean, when I first started this, this was a straight marketing coaching business and that was the whole model, but we ended up in the agency space because a lot of our advice required some level of execution and we had built these relationships with clients and they didn't want to go anywhere else. And so I kind of reluctantly started this agency as like just a way to, to, you know, calm the rioters. And so after we
launched it, it became a huge profit center for us. And so over time, I've just kind of eliminated the coaching business from my main kind of profit center. I still do it. It's more of a passion project. but really we, got deep into lead generation because when we were first starting out, lead generation was key. And so we, that was really the, the, you know, in the, B2B services space, we just got really, really good at that. And then over time with the advancements in AI and the ability,
Kyle James (02:06.254)
Mm-hmm.
Drew Donaldson (02:23.377)
ability to do hyper targeting, we just got more and more specific about like how we can generate leads, not just, you know, some tripwire funnel that collects a bunch of people for an email list. It's like, no, like how do I get a decision maker in front of you so that you can make your pitch and sell your product?
Kyle James (02:25.592)
Mm-hmm.
Kyle James (02:37.836)
Yeah.
When you say hyper targeting, I'm thinking definitely niching down. What exactly do you mean by that?
Drew Donaldson (02:46.493)
Yeah, so it's using buyer intent data in our campaigns. So a lot of people don't realize, but every time you visit a website and you accept a cookie from that website, those cookies are pooled. Every action you take on that website, every page you visit, every form you fill out, it's all collected and shared on these things called co-reg networks. And there's hundreds of them. And these things are essentially
giant data collection points. And if you buy into one of these networks or one of the outflows of these networks, you can get lists of buyers to feed into social media campaigns, cold email campaigns, cold calling, know, what have you, and use them, pay-per-click as well, I should mention. You can use this insight into who is actually expressing buyer intent in a specific category. So instead of just going to Apollo and buying
thousand leads in a particular category, I can say, well, you don't really need a thousand leads because 800 of those aren't going to do anything. They're not buyers. They're not in the market. They're not interested. can now set, eliminate those 800, give you 200 that are in market and run much more efficient campaigns. can also use that data to do create better lookalike audiences because a lot of times with lookalikes, know, people it's a good concept, but if your initial data is poor or not, you know, enough
Kyle James (03:44.174)
Sure, right.
Kyle James (03:50.988)
Hmm
Kyle James (04:09.624)
Yeah.
Drew Donaldson (04:12.735)
quantity, you can't really use it. using buyer intent data in lead generation campaigns is really the next step, especially for high ticket business B2B service based businesses.
Kyle James (04:23.094)
Yeah. I mean, that's like a, such a big thing, right? With like, with leads coming through is, is like, are they qualified leads and like, are they relevant to, they, are they, would they actually appreciate your services based on like, what you're saying is like the data on the backend, like the buyer intent that's coming through. and within like grow house, like I know you are using AI to some degree, tell me what exactly made you want to decide to start using AI in the first place at grow house and like what types of like.
challenges were you trying to solve?
Drew Donaldson (04:55.891)
Well, initially I've been in the machine learning and AI space longer than chat. GPT was publicly available. So I was using AI based tools and machine learning and writing my own algorithms from way back before any of this stuff was public. And because, I came from that background, it was always something that was like, if you had a very specific thing to solve. like I had this one project that we wrote in Python and it had a very specific beginning and end. wasn't really this, this
Wild West that that chat GPT has kind of opened up. And so in those cases, it worked really well, but getting to that point was hundreds and hundreds of hours of, you know, kind of fine tuning this thing to solve this particular problem. So I always kept it in the back of my mind and I always, would use these tools and I would test them, but they were never ready for, for stage time. They were, they were always kind of like, you know, maybe we can solve this using this, maybe not. So
Really around the time chat GPT became public, I started to take a lot more notice because I was like, okay, the widespread adoption of this is going to mean that I can explain this a lot easier because people are going to understand what I mean when I'm talking about AI and not just think about that movie with Will Smith, right? Like they're actually going to have some context around it. And it wasn't really until I would say like, three GPT three came out or three five, last summer. I think that was around the time I forget.
Kyle James (06:12.877)
Right, right.
Drew Donaldson (06:24.115)
the virgin history of all of them. But it was only around that time where I had an inflow of clients and I was doing a lot of the copywriting manually myself, because that's what my background is in. I've written thousands of pages of copy at this point, from landing pages to ads and everything in between. I know the systems and structures that work, but writing them all out manually is just very time consuming.
Kyle James (06:38.402)
Right, right.
Kyle James (06:48.589)
Right.
Drew Donaldson (06:48.989)
especially when you have, you know, a backlog of work to try to get through and you're trying to run a business and you're trying to close more clients. So my first kind of for, my first kind of venture into this was how can I replicate my copywriting style? And that's where I really got deep into prompt engineering, because if anyone has ever just opened chat, GBT and said, write me a cold email sequence, you don't get good results. hate to break it to you. It's, it's, it's pretty terrible. And so.
Kyle James (06:54.094)
Sure.
Kyle James (07:12.606)
No, no you don't it's it's pretty generic
Drew Donaldson (07:18.725)
It, I really got into, know, like hearing you talk about the prompt trainer software. was like, I wish I had that when I was starting this, because it was a lot of time at my desk trying different things. And, you know, back then your prompts had to be structured so carefully. Now there's a little, it's a little more, you know, you can be a little bit looser with it, but you still need to know good structure. You still need to know good instruction. You still need to make sure that you're not.
overloading it with information, which is the biggest thing I see when clients come in and they're like, can you fix my AI bot or my automation? It's like, it's cause they've overloaded it. They've tried to make it do too many things at once. And the context window just isn't big enough for whatever model they're using to take in all of that and actually appropriately handle the task at hand. And so when I was going through that, I trained my first couple of watts doing copywriting. And then we piloted those with a couple of clients.
Kyle James (07:49.432)
Mm-hmm.
Kyle James (07:57.121)
Right.
Drew Donaldson (08:13.981)
those turned out really well, like they produced results and we were like, okay, this is actually, we can use this.
Kyle James (08:19.238)
And so did you like, when you said you pilot it and then send it to your client, were you selling it as a service to them or just saying, hey, if you're a client of ours, then you have access to this free, I guess, AI agent.
Drew Donaldson (08:34.031)
It wasn't even that it was us doing all the work in the back end. so essentially like, you know, say we were writing a campaign for a client, you know, say we were writing three ads, one of those would be AI based and we just kind of throw it in. We'd send them all for approval. They would approve it. They'd like, yeah, I like all three. We'd run all three and then we'd see what the results looked like from that ad. And if it performed, then we knew, okay, it's, good enough to, to meet the human criteria.
of like what we would be able to do if I just sat at my desk and type this out manually, it can do the same thing. And so we didn't sell it until I would say like late last year, early this year, because we were still building internally. So I would say for the first six months of our deployment was all internal tools. was like we didn't like hide it from our audience or from our clients that we were using it.
Kyle James (09:22.062)
Mm-hmm.
Drew Donaldson (09:30.899)
But we didn't come out and say what you're buying is AI generated because of the stigma around it. And because people, because people go into chat GPT and say, a cold email and that's their frame of reference. It's like, they don't realize that when we spent a hundred hours getting these things to perform the way we want them to. Yeah.
Kyle James (09:35.406)
Mmm.
Kyle James (09:41.102)
Yeah.
Kyle James (09:45.646)
It's like spending that time on the back end of it. That's why you're able to get so much of the better results there is because when you just go to the large language model and just type in a two sentence prompt, build me a sequence of emails or build me some content for social media posts, it just gives you basic. it seems like you're getting some pretty good
Drew Donaldson (09:52.959)
110.
Kyle James (10:09.258)
Results with it like what type like specifically though like when you were doing that test phase sending them those like three templates of content Like what specifically like in regards to results were you seeing? Even like maybe your clients like once they started using the AI templated version, I should say
Drew Donaldson (10:28.979)
Yeah, so for, you know, like paid media, we would use AI to write pay-per-click campaigns. We would use it to write social campaigns. And because of the backend time we spent with it to get it to where it is today, we were seeing benchmark results on all of those campaigns, some that even beat benchmark. So that's really how we deployed it was just like we would have, you know, I have my cold email bot.
Kyle James (10:33.038)
Mm-hmm.
Drew Donaldson (10:54.947)
and I have certain forms, form fields that I have to give it. And then once I give it those specific criteria, it's able to spit out ready to mail emails. Same with the...
Kyle James (11:06.51)
Yeah. you started to test this out and see the results, did it change your business model and maybe your workflow over the next couple of months once you did see the traction with it?
Drew Donaldson (11:25.049)
massively sped up client engagements because now it wouldn't take me two weeks to write an ad. I could get it done that same day. I could have the ad written before I even had the call with the client to discuss the ad because I've been in this market for like 20 years, man. Like I, I've been around the block. if a landscaper comes to me, I kind of know who their audience is. know what they're going.
Kyle James (11:44.334)
Yeah, familiar.
Drew Donaldson (11:46.235)
Every once in a while, some crazy business will show up and I'll be like, I know nothing about what you do. Like, like sit with me for two hours and tell me everything about this. But nine times out of 10, that's not the case. And so I can take what I know and what I know works and then have that ready to go on that initial call, which not only kind of it's that surprise and delight element. It's like, they think they're coming in to write this ad or write work on this landing page or whatever. And I'm like, it's already done.
Kyle James (11:54.242)
Yeah, yeah, because I'm close. Yeah, that's funny.
Kyle James (12:12.654)
Sure. Yeah.
Drew Donaldson (12:14.259)
Like all you have to do is read it and approve it. And so it sped up our time delivery significantly. It reduced the amount of manual work I was doing. I still go through and like copy edit, like AI is not perfect. It will get, you know, I always tell people it'll get you 80 % of the way there. The other 20 % is that human element that it can't pick up on. And I think the big thing just going back to like the mistakes people make when using AI is that what they don't understand is that AI prioritizes two things, pattern recognition and efficiency.
Kyle James (12:24.398)
sure, of course.
Drew Donaldson (12:43.783)
So sitting at your desk, like if we imagine like the old ad writers and art directors in the fifties, I like old will be right. They'd sit in a room for eight hours kicking around ideas for, you know, one ad for one client or campaign or whatnot. Right. It's not the most efficient way to do it, but they came up with some really amazing campaigns because through that brainstorming process and trying different things, they were able to hone that message and hone that tagline. AI doesn't do that.
Kyle James (12:52.91)
Mm-hmm.
Drew Donaldson (13:13.183)
AI looks for what is the pattern in the data and what's the most efficient answer I can give them to deliver that result. And so that's not to say the results aren't wordy. It's just to say, it doesn't want to work as hard as you want it to work. So you're still going to need to go in and fix that 20 % where it's just like, yeah, I wouldn't say that in an email. don't, or like they, sometimes AI will like throw in weird things like, I hope your day is going well. It's like.
Kyle James (13:35.31)
Mmm.
Kyle James (13:41.102)
Hmm.
Drew Donaldson (13:42.723)
I'm trying to get it like 50 words. Like, why are we wasting 10 words on like, I hope you're doing well, or three words, right?
Kyle James (13:45.026)
Yeah, yeah.
Yeah. Yeah. That's, that's so funny. That's it's crazy how like, um, I mean, just, you know, that's the thing, right. Is because more and more companies are using AI, like that last 20 % that you're talking about where like you'll go in and fine tune it. Your team will fine tune it. Like that's what makes a difference. But the initial going from like one, you know, the one Valley to the hilltop is like the AI can get you that first hilltop, but that last bit of climb that's, that's where the real skill kind of comes into play, which is the human aspect of things, which.
I mean, humans will always be here, right? Can't get rid of the humans, thankfully, hopefully not like Terminator style, right? But you know what mean? So it's like, it makes that big of a difference when you have like, we had a conversation on a last podcast I did and they made a bookkeeper AI agent. And they said the funny thing was the people that were using the bookkeeper AI agent was bookkeepers.
Drew Donaldson (14:24.605)
Yeah.
Kyle James (14:44.738)
Like they were using it to like speed up their process. And so like, I see that like that trend there where it's like the people who are using the marketing AI agents are the marketing professionals. So they can just amplify. It's not like they're to be the regular Joe out in the street using marketing, like, that's not their expertise. It's the experts using the expert AI agents that will make their next project that much more efficient and more effective. So.
Drew Donaldson (15:09.823)
Yeah, I mean, it's old school illustrators complaining about Photoshop. That's all, that's all it comes down to, right? It's the, people that the teachers that you had in high school that told you, can use the internet for research. Like it's the same kind of push against progress that it there's really no, nothing to gain by doing that. And at the end of the day, like I met with marketers that are a little long in the tooth and they refuse to use AI and they don't believe in AI and they believe AI is the devil. And the reality is.
Kyle James (15:13.357)
Mmm.
Kyle James (15:20.974)
Mmm.
Kyle James (15:28.951)
Mm-hmm.
Kyle James (15:38.114)
Hmm.
Drew Donaldson (15:39.529)
they're going to be the ones that are out of the jobs. It's not going to be people like us that adopt it and use it and leverage it. Cause at the end of the day, it's still a tool. That's all it is. And like the most clear way, you can, you can see this is bought because AI doesn't understand the core human elements that make up that last 20%. It doesn't understand subtext context clues. doesn't understand empathy or emotion because you can't program those things yet. Now, maybe there comes a time.
Kyle James (15:47.395)
Mm-hmm.
Kyle James (16:01.326)
Hmm
Kyle James (16:07.406)
Mm-hmm.
Drew Donaldson (16:09.343)
where we get to a hundred percent, but the climb from 80 % where we are now to a hundred percent is going to take a decade to get there. It's not going to be the rapid pace of advancement we've seen because it's the same thing with internet, right? Think about how long we a diet, we had dial-up and then a couple of years later, DSL came out and then after DSL came cable. What's, do you have faster internet than you had 10 years ago?
Kyle James (16:17.687)
Right.
Kyle James (16:30.076)
huh.
Kyle James (16:37.026)
Yeah.
Drew Donaldson (16:37.095)
Maybe if you're paying for a higher plan, but there's not some new fancy thing, it's incremental improvement, right? Now we have fiber optic and that's a little bit better, but it not, everybody has it. And the next thing, you know, who knows what it is, but we've hit a point where the advancement in that space is going to be much slower from here on out. There's still going to be amazing things. There's going be amazing advancements, but they've already run. I mean, you got to 80%.
Kyle James (16:39.982)
Hmm
Kyle James (16:46.03)
Sure.
Drew Donaldson (17:05.576)
Like you only have 20 % to go, but that 20 % is the part that's the difference between making something sound robotic and like ChatGPT wrote it and making it seem real and human and authentic.
Kyle James (17:05.58)
Yeah.
Kyle James (17:16.002)
Yeah, that's, it's, it's funny to think like that, that perspective of like how, how much of a climb we still have when you get that first, like, like I said, mentioned going up to the Hill, that first Hill type, man, we can go to the second Hill top just as fast. not so much. Like it's going to take time just looking at history, which is, think that's a great perspective that you've shared there, Drew. So walking a little bit more through about grow house, like I know you've got some, you're using AI and great capacity on the marketing side.
What are some of the maybe like upcoming AI initiatives that you have planned out for this year, next year? And then where do you see AI playing probably like the biggest role in your operations next?
Drew Donaldson (17:57.327)
AI powered automation is really where we're focused on now is because it automation, you know, I've been in automation for a long time and a lot of the manual work that you would have to do to get an automation to work is quickly being replaced. Now I think automation and agent base still has a long way to go because there's still, you know, we go try to build an agent that checks your calendar for free and busy using Google calendars API.
And then tell me that agents are ready for prime time. They're just not. Dude, I spent hours on simple agents that just won't work because of whatever, whether it's the prompting or the API call data coming in or the context window or whatever, they're just not ready for prime time. That said, there's a lot of things you can do with agents now that is fascinating and streamlines workflow.
Kyle James (18:30.08)
Yeah, far off.
Kyle James (18:36.396)
Yeah, I've done the same.
Drew Donaldson (18:52.585)
But I think we're coming into a place where people are seeing these amazing things. They want them for their business. They have no idea how to do them. Or if they've tried them, they've got so lost in the sauce that there's, they have no path forward. So it's kind of like standing outside a car dealership and looking at a Ferrari. It's like, it's so close. It's right there. I could just get through that window. I could touch it, but the car dealership's closed and you're not getting them. Right. So our,
Kyle James (19:10.024)
Yeah.
Kyle James (19:16.045)
Yeah.
Drew Donaldson (19:17.683)
big initiative moving into Q3 and Q4 is really going to be pushing some of those services in terms of like building out and maintaining them. Because I think that's the other piece that people don't realize is in the old days of automation, once you build something, you're kind of done. Like as long as the APIs you were using didn't change, which most of the time, if you're using legacy services, there's not going to be a huge API change. The agent function of this changes constantly.
Kyle James (19:39.639)
Right, right.
Drew Donaldson (19:45.427)
And you constantly need to go in and check to make sure they work best in that, you know, the newest version of, of chat GPT or Claude or whatever you're using is performing in the same way. And if they depreciate, like, you know, say you built something on like chat GPT two and it's been depreciated. Well, you need to now update that prompt and update it and test it and go back through that, those rounds of revision. So we see it as not only a service to build, but also maintain.
Kyle James (20:09.74)
Yeah, absolutely. Yeah, that's a good space. I mean, automation, that's essentially what AI is, right? It's just automating things as much as possible, whether it's like even through AI, make.com, N8N, Zapier or some other automation. that's, it's saving, it's going back to saving time for people and for businesses. That's take your time, save it, put it towards something else that's going to bring in more revenue. So, and as we wrap up here, Drew, I know we've got a few minutes left here. Like for those listening in, like,
Drew Donaldson (20:32.318)
mean, time is money.
Kyle James (20:38.722)
Very curious to learn more, but where can they learn more about you and then maybe a little bit more about GrowHouse?
Drew Donaldson (20:45.279)
Sure, so I will say I'm kind of a social hermit. I don't really post all that much on social media. I should my team post stuff for me, but I'm kind of more of a lurker. And the only place I'm really active on is X. That's kind of where all marketers and AI people have kind of congregated. So that's where you'll find me. And it's GR0, H-A-U-S, which by the way, if anyone knows Greg Rowehouse.
Kyle James (20:53.827)
Yeah.
Drew Donaldson (21:14.111)
Uh, if you could get me in touch with him, I'm trying to buy his handle for like the past three years and I cannot, I cannot get a hold of this guy. Um, so it's GR zero H A U S or you can visit us on the website, g ro h a us.org. Uh, and you can book a call with me. I'd love to talk to you and see what you're thinking. See what's possible.
Kyle James (21:33.422)
Awesome. I love it. Thanks, Drew, man. I appreciate you being on the podcast today, AI Chronicles. Looking forward to maybe potentially having you on in the future. And for those listening in, remember, if you're looking to implement AI into your business today, don't try and do it yourself. The time and stress that the AI could cause just isn't worth it. Schedule a call with GPT Trainer and let them build out and manage your AI for you. Once again, that's gpt-trainer.com. Signing off for now.
Have a wonderful rest of your day and looking forward to seeing everyone on the next episode of AI Chronicles.
