How to Build an AI Agent Team for Your Marketing Agency | Tristram Dyer
How do you build an AI agent team for a marketing agency? Tristram Dyer, CEO of Webtopia & OAKS, walks through the exact system his agency runs: fast AI that speeds humans up with MCP data connections and AI creative tools, and slow AI, a team of scheduled background agents that detect performance anomalies, pace budgets, write morning briefings, and operate Meta ads with a human approving every change.
We discuss:
Fast AI vs. slow AI: the Thinking, Fast and Slow framework for agency work
The creative workflow connecting Claude, Arcads, Claude Design, and HeyGen
Why every Webtopia employee gets a paid Claude account
The full agent roster: anomaly detector, agent 00 morning briefings, the Meta ads operator, and Mia, the Google Ads specialist
Will AI replace ad agencies, and how to incentivize a team to adopt AI
Why agentic ad buying may be only three months away
Watch the previous episode with Cody Plofker: https://www.ai-dtc-wtf.com/podcast/why-ai-ugc-is-a-trap-and-what-actually-works-cody-plofker
This Episode is sponsored by Motion's very own Runneth. Runneth is the AI Tool that is quickly becoming the go to decision maker brands turn to when it comes to creative decisions within their ad accounts. From Creative stratigest to hook insporations and creative briefs, runneth can do it all. See the case studies and try runneth yourself https://runneth.motionapp.com/
Learn more about Tris and Webtopia: https://www.webtopia.co
To connect with Andrew Foxwell reach him here Andrew@FoxwellDigital.com
To connect with Will Sartorious DM him here https://x.com/will_sartorius
To Connect with Thomas Moen DM him here https://x.com/thomasmoen
To learn more about Foxwell Founders and conversations like this one, go here: www.foxwellfounders.com
Learn more about Foxwell Digital: https://www.foxwelldigital.com All episodes: https://www.ai-dtc-wtf.com
AI DTC WTF is the super practical podcast about leveraging AI to get an edge in DTC, hosted by Andrew Foxwell, Will Sartorius, and Thomas Moen.
Chapters
00:00 How to build an AI agent team for your marketing agency
01:22 Fast AI vs. slow AI: the Kahneman framework
02:23 The fast AI stack: Meta, Kaizen, and Motion MCPs
06:26 Kaizen: adding qualitative client data to the numbers
08:27 Why every employee gets a paid Claude account
09:28 Creative workflow: Claude → Arcads → Claude Design → HeyGen
11:29 Tangible results: faster decisions, happier clients
13:31 The roadmap: putting agents in founders' hands
16:33 The agent team: BigQuery warehouse and anomaly detection
18:34 Morning briefings and a self-improving agent
19:35 The Meta ads operator with human-in-the-loop
22:40 Will AI replace ad agencies?
25:46 The Cody Plofker landing page example
27:48 How to incentivize employees to use AI
30:51 Meet Mia: the Google Ads specialist agent
Full Transcript:
(00:22) Welcome to another episode of AI D2C WTF. We are excited today to have Tris Dyer on the show. So AI DTC WTF is a super practical podcast, so we don't have time for introductions. So we'll just go jump into the good stuff. So thank you for joining us today, Tris, and excited to talk a little bit of AI and what you guys been doing lately.
(00:47) Yeah, lots. What haven't we been doing? I guess, yeah, look, I think overall, one of the things we've seen through AI is it started off about a year to year and a half ago where we saw a few things on ChatGPT and we're plugging it in, seeing what happens. And we've gone all the way now to pretty much semi-autonomous humans doing what they're doing with AI, which is amazing to really see the development.
(01:15) But so many people find it scary, but I'm only enthused about where we can go with it. So yeah, I can go top to bottom if you like where we're going. Yeah, for sure. And I mean, I agree. There's just lots of opportunities, but also, of course, kind of lots of fair and stuff like that. So yes, just tell us a little bit about kind of how things have been evolving lately and what you guys have been doing.
(01:37) It looks like you guys are using agents and are like really diving into the whole thing. Absolutely. Yeah. We base our theory on, I think it's Daniel Kahneman's book, Thinking Fast and Slow. We talk a bit more about kind of, you know, there's two different ways of approaching AI. So first of all, especially when it comes to work.
(01:56) So from a fast point of view, you're kind of thinking of, you know, humans, speeding humans up. So, you know, it's a person's work. They take their grunt work, take all the stuff that, you know, was done by execs and et cetera, and that's moved into AI. And that's not just, you know, automating work. It takes things from minutes, from hours into minutes.
(02:13) It speeds things up significantly. But then from a slow standpoint, we've got kind of that background work to make sure there is, you know, scheduled ongoing work that's continuously needed. The kind of slips through any cracks, anything that, you know, you might as part of a team might miss or whichever that kind of holds underneath.
(02:30) And then it kind of builds out a strategy that it helps as well to inform. So I'll go through in detail. I suppose that's kind of the overarching look at it. But from a fast... Yes, absolutely. Fast AI and slow AI, kind of. That's the overarching... Yeah, yeah. Well, it makes sense from the slow AI standpoint, because so many people think, oh, AI is just going to speed everything up faster and faster and faster.
(02:51) But actually, there's stuff that you can do to help you slow your thinking down as well, because you still need to think as a human as what you're going to be doing. Got it. Okay, so take us maybe through then. Let's start with the kind of the fast AI thing. What kind of tools are you using? What are you using it for? Kind of give us the rundown on that.
(03:10) So, yeah, I suppose on a fast AI standpoint, it's very MCP-based. It's very focused on trying to get fast, bite-sized data that you can use to analyze. But it's getting larger data sets to analyze quicker. So, we use things that we obviously use the meta MCP, the Kaizen MCP and Motion MCP. I'll talk about Kaizen now in a second.
(03:29) But bringing in the data really, really quickly into different dashboards that allows us to understand what's actually happening in the world. So, if that's performance creative, if it's performance of an actual company, how it's actually working, what's making profit, what's not, where are we spending, all that stuff comes in really, really quickly and allows a user to use the AI like a tool.
(03:52) So, if you think about going from using a manual screwdriver into an electric screwdriver, that's what we've done there. So, we've split it up quite quickly. Obviously, the sexy kind of making creative has been a huge jump for us as well. So, we've gone from a couple of designers making some kind of craters here and there to jumping up and using something like Arcads or Claude Design.
(04:11) So, the two of those connect. They work together. So, Arcads allows us to create the UGC that is AI generated. But we then take that and move that into Claude Design and actually to have the overlays of the different brands that we have over the top. So, there's a couple of different connectors that have worked quite well.
(04:28) Arcads has been really revolutionary for us because it's allowed us to be able to build out the type of creative that we want without having to go and ask the client to come back and actually make us more and more content. So, it's taking out that roadblock that's slowing things down. We all hear a ton about AI.
(04:49) We all hear about what it can do and how amazing it all is. And I don't think that there's anything incorrect about that. But this episode is brought to you by an AI brain that is absolutely incredible for your marketing team called Runneth. It's from Motion and therefore, you know it's awesome because it's from Motion.
(05:06) But really what Runneth does is connects every single function of your marketing department into one place. It can do so many different things. It can build you ads. It can verify that your brand looks proper. It can verify that what your top creatives were and you can ask it. You can have it live in Slack or you can have it live as a CLI so you can interact with it any way you want to.
(05:28) The thing that's really insane to me about Motion is how absolutely intelligent it truly is and how it works cross-functionally across the team. There was recently a marketer that Reza talks about in the landing page on Runneth, which you should definitely check out, where basically she went on vacation, the CMO did, and said, if Runneth says the ad is okay to run, then it is okay to run.
(05:50) So she uploaded all the context for the team. They ran it into Runneth and Runneth said, yeah, that ad is totally on brand. It's absolutely crazy the number of things that this can do. And I think the fact that we all live in silos and the fact we all live separately is really, really challenging and one of the hard parts about AI.
(06:06) So that's not what this does. This brings it all together with shared context, shared resources, and it also looks at thousands of ads every day. So it's always up to date on what actually is the best performing ads that it's seeing across Motion. And so that's another whole thing that it does. It's really crazy because instead of one more dashboard, you get a teammate in your Slack that knows your performance data cold.
(06:28) As I said, you can ask it really anything and answers instantly. Weekly recaps, top creatives, which creators your audience trusts. It learns your brand, your playbooks, and gets smarter every day. E-commerce and D2C teams are using it to turn weekly reporting from a two-day job into a two-minute conversation.
(06:45) Head to runneth.com and see it for yourself. And it also has a 30-day money-back guarantee. So if you need a cheat code in e-commerce and D2C with your ads and your ad creative and your marketing team, check out runneth.com. 30-day money-back guarantee. It's absolutely incredible what this thing can do, and I encourage you to give it a shot.
(07:10) Okay, so just to kind of, so I understand because there's a lot of things. So you have then MCP connections into typically cloud, cloud code, cloud code with meta, motion, and Kaizen. And that's kind of where you do the brainstorming, creating scripts or working on strategy. Yeah. And then, yes. Let me put you on Kaizen because that's actually an interesting one that I glazed over there.
(07:37) So, I mean, it's really good to bring in motion and meta information. That's all the numbers, right? And that's all the quantitative data. But what we miss, and a lot of times marketers miss this, is the qualitative data that comes with it. So talking to the marketing manager, talking to the founder, like what they're actually trying to do.
(07:52) And everyday conversations, it's not just conversations of like, we're sitting down to talk about this. It brings in the conversations from the account manager with them, or it might be talking to the operations manager in the company. All of that's recorded and brought into Kaizen, which is where we record our calls.
(08:07) But it actually allows us to understand sentiment analysis, understand what's really good and what's really bad, what they really like. And brings that information, too, into the strategy. So it's like the founder might have talked about it this way, or whatever the conversations might happen. That connects in them with motion and meta data to be able to give you information in a much richer format rather than just looking at numbers.
(08:29) Because we're not. Okay. So Kaizen is like connecting all the transcripts and all that data. So that is kind of the intelligence, the client intelligence layer of it. Yes. Got it. Yeah. Okay. And again, your team is using cloud, co-worker, cloud code with these MCPs to kind of connect everything. Every single person in our company has a Cloud Pro account.
(08:56) Straight up. Like that's just the standard. Because so many people in the industry, in the agency world have gone, I'm not too sure, should I be using AI or not? We just went headfirst into it. I said, right, everyone has a Cloud account, a paid Cloud account that they can use that comes with Cloud Code, Cloud Design.
(09:12) And we keep those in Cloud projects as well per client. That's cool. Just to jump in, because again, I'm super nerdy about this and this is exciting for me. So I've looked a little bit at Arc ads and those kind of things. I think that's really, really good. And I see a lot of other agencies are using it. Cloud Design is a bit new, I think.
(09:34) And so how do you work together with those things? Like super practically, how would your team combine those two to create an output? Yeah. So the way we would do it is we create the initial designs and the initial briefs with Claude. And Claude would then be connected. And we'll go into the slow thinking now shortly.
(09:53) But they would design briefs and everything else with that. That would input into Arc ads. It would actually bring in different aspects of winning creative and winning performing creative out of that into a brief for Arc ads. Arc ads then goes and creates the basic structure of the ad with all of the creative in it.
(10:10) It might be a UGC person talking about a certain product with the product in their hand. We then bring that into Claude design to actually be able to build out statics, build out kind of iterations on that. So that we're not just launching with one version of it, but several versions of the one. So it might, and we use something like HeyGen that might take that and actually translate that into six different languages.
(10:31) So we work with people all across Europe that want to translate into those different languages. So it'll take the initial ad and translate those really, really quickly. So it's all about one of the things when you're using across like Claude design and Arc and everything else, it's kind of token maintenance, I call it.
(10:45) So if you just start throwing it all into one AI, it uses all your tokens really, really quickly. But if you're clever about where you brief it, so for Claude design, if you get your briefing and everything else done before you get to the design stage and use it in actual Claude code, it doesn't burn as many tokens in Claude design.
(11:02) And that will actually save you a lot in the long run. It'll be a lot cheaper. Tris, I'm wondering of the stuff that you've done under fast AI, do you see like have client results increased as a function of this? Because you're able to spool stuff up faster and have client relationships have gone better as well.
(11:22) Like, can you give us some tangible examples of like where it's impacted both of those? Absolutely. Yeah. I mean, that's, that's exactly it, right? We always talk about two parts of the job, doing the job and telling someone you've done the job. The first part is actually really increased because what we're able to do is make decisions faster.
(11:37) Is that going to lead to a better result? Is this, is one connected to the other? Is a sale on a meta connecting to a sale on the actual, on the Shopify? So there is a very tangible seat, like speed in terms of the speed increase there in terms of making decisions. But then the second part as well is actually allowing the client to understand what we've actually done and how we've done it.
(11:59) So when we're making changes, when we're doing certain things, we'll allow our agents that we'll talk about shortly to be able to actually write summaries for that in a very simple format, not just spit out slaw, but actually explain, explain to clients, this is what we've done and why we've done it. We'll read it and go, yeah, happy with that.
(12:15) Send it on. So with that constant communication, that constant contact allows that step up in trust, but also allows us to be able to analyze and go, that's what I've done. Yes, I'm happy with that. Or no, undo that and change it and move it back. So in terms of tangible results, that's actually being, if that being creating ads for a specific product in a campaign, it will actually create ads a lot faster and actually start running ads a lot faster as a result of this.
(12:41) I think that another tangible result is actually part of our kind of, one of our daily agents is sending a daily update and actually getting that daily update written and transcribed from six or seven different data sources into one and actually written and then analyzed by a human and sent across. That would typically have been a lot more of the person's time.
(13:01) Now they can actually spend the time going and analyzing that that worked and actually going and making decisions on where they need to go next. So that's some two tangible results on those. The processes that you've already done here with Fast AI. What are the other things that you're like, look, like what's the six month vision of doing some of this stuff, right? It's, it's, I mean, it's an analysis.
(13:23) It's, it's building more, you know, building better ads. It's discovering new personas. Like what are, what's the roadmap for you? Cause I assume what happened is you like had them sit down and say, look, this is where you're spending time. This is what we're, you know, and like, we're going to optimize that.
(13:37) But like, what's the looking forward for you in terms of continuing to make this better? Yeah, I think the biggest thing for us, the biggest kind of growth lever for us is actually not just building it for ourselves, but allowing others to be able to use these products themselves. Cause when I'm starting to realize a lot more quickly is that, you know, we'll build it for ourselves and it'll be great.
(13:55) But the democratization of all of this is going to be huge where people will go, well, I could just make that myself. So we want to put those tools into the hands of owners and founders so they can go and use those information, have an app on their phone, or even be able to connect to when Siri eventually starts to connect.
(14:11) To Claude, we can actually start to make artifacts that people can see on their phone. So it's actually starting to roadmap over the next six months with WWDC coming up soon. They'll release obviously anything that, like I said, connecting Claude to their iPhone, where they'll be actually able to make ad changes on their iPhone or actually be able to get that information directly to them.
(14:30) Because if you think what a founder wants is the clarity and the knowledge that their agency is making the right decisions, but also doing that and leading them in the right direction. So our roadmap is to be even clearer and give the founder even more knowledge and power to be able to go and make things like ordering decisions or marketing decisions they want to talk about.
(14:48) So it's empowering founders a lot more quickly. So creating agents that we like and then making them publicly available. Interesting. That's super interesting. And then as a business model, you're thinking about like a monthly retainer per agent or something like that? Or what's your thinking around that? There's two parts of it.
(15:07) There's either a retainer model on our side where we would say, right, we'll continue to make you those agents. I will make you newer and more improved. Or it's going to be a simple case of this. You know, just because you have a screwdriver doesn't mean you can build a house. You still need somebody to run it, operate it, make those decisions.
(15:22) We always make these, we design these agents with you and in the loop. Because we don't like, there's no point in having a fully autonomous platform. Because what happens is it'll just go and start, you know, making all sorts of decisions that may not actually line up with your overall goal. I don't know if you remember Silicon Valley with Son of Anton, where he started making AI decisions like that.
(15:41) That's the thing we're trying to avoid here, right? So I think, you know, yes, that's a silly example. But I think that's something that, you know, we want to make sure that we still have control and direction and be able to make those calls. But it's something we're seeing for like time and time again, that every time you kind of think you've made something innovative, Claude or something else will release something that will just completely squash it.
(16:00) So if you're making a SaaS product, you know, that's great. But then you can go and spin that up in five minutes, which one of the things I would say is that power is the data. So that you can make a SaaS model, but if it's not based on anything concrete, there's no point. So our business model then around that, Thomas, to a question is like, what do we actually monetize? How do we work it out? It's looking at the data and how we use it.
(16:21) That's the direction. We've always looked at ourselves as a bit of a fiduciary and actually managing people's money correctly. That's one of the things that we will continue to do, but at a higher level now that we've got AI plugged in. Can you walk us through the agent team that you have and like the different things that they're doing? Just run us through the whole list of every day, every week and every month.
(16:42) It's a perfect transition, I think, then, Andrew, from what Chris was saying, is like the data part of this, right? Because your agents are able to do what you're going to walk through now because you have this data warehouse where you're connecting everything, right? Absolutely. Absolutely. Yeah. So this is exactly it.
(16:59) So we keep our data in this very secure data warehouse on Google BigQuery. And what we do is we import the data on a daily and sometimes hourly basis, depending on the connector. But what we then bring that information in is that we have a central store of everything that you need to know about a certain brand.
(17:14) So that'll be whether it be quantitative data, where it's literally all the numbers and everything else, or the qualitative data on like, is this happening or when's that happening? What's the plan and everything else? All of that comes into one place. But then we start to identify, for example, what the goals are, what the ROAS is, et cetera.
(17:33) And every morning we get a performance anomaly detector. That's agent one. And what it does is it catches any ROAS or CPA problems that are happening across all the agents. And it'll DM either me or it'll DM the account manager that's working on it saying, there's a problem here. Take a look. So when they open their phone or open their laptop in the morning, that's the first thing they see.
(17:52) Have a look at this. And we've refined that over time because sometimes the CPA goal might not be clear over what period. So we started to refine it where it gives it a 3, 7, and 14-day view to see if those anomalies are working. And these agents, are these cloud managed agents or are these open claws or what kind of tech is in the background here? So these are cloud routines that are running in the background.
(18:16) Got it. So you have a separate machine that's like, these are all the routines and it's just running through. Yeah, a good friend of mine, Thomas Moen, actually showed me that. Yeah, it was one of the things that, I mean, it's great to share this, but it's like where we actually were able to say, right, we have a machine there that runs all this.
(18:35) And we've got a bunch of other things around, uploads and stuff like that, that runs through the machine. And the idea then is that this is a agent to run off the back of this, but we have a self-improving agent that actually goes back through any feedback that our team gives. So if we're missing, for example, missing ROAS or the daily update we'll talk about in a second is slightly incorrect.
(18:54) That will be corrected in Slack. And then the agent will go back through Slack and read all of that and self-correct. Cool. That's very cool. Okay. So keep going through your different agents. It's just cool to hear how it's in the background as well. So sorry. Of course, yeah. So this is part of the slow AI, as I was talking about.
(19:09) So you've got your budget pacing to make sure you're on budget. That will, again, be part of your overall. Then you've got your morning briefing agent, which is what we call agent 00, because the idea here is that this starts to gather all of the information in one. So how the performance is going, where you are in terms of your budget spend and how your CPA is going, it starts to come together.
(19:28) Think Jarvis out of Iron Man. That's what you wake up in the morning. It's like, this is what you've got. So then we obviously have a daily report. We've got a detailed performance report. Then in terms of underneath it, it goes, okay, so this is how your CTR is going. This is how many add to carts you've got.
(19:42) This is what your conversion rate is and so on per product. If you want to call out something in the previous day that you want to keep an eye on, it will then call those out in those daily reports. So this is almost like having an exec directly with you on each of those. We then obviously have our meta ads operator, which is connected directly to Facebook's MTP.
(19:59) So it verifies everything that comes into in through our connector into the back end of BigQuery, but verifies it through meta. Because sometimes you want to make sure that the data is correct. But then what it does is suggest any ad approval changes we've got in there. So we've trained this agent on our SOPs in terms of ad buying.
(20:14) And then it goes and makes suggestions and gives you the option to thumbs up or thumbs down in terms of making those changes. So those changes, again, it's required as a human in the loop. Because we, again, we're responsible. Regardless if we put an agent running it or not, we're still responsible. So that's really, really important.
(20:30) So, again, I'm just trying to dig down to the specifics of things so I can understand it. So it's cloud and schedules and kind of the outputs of thumbs up, thumbs down. Is it generating artifacts? Is this like built-in notion? What's the interface here? All of this is created into Slack. So we created a Slack app for this.
(20:51) And so what happens then? It drops it into Slack. And the communication is back and forth with a human in Slack or a user in Slack. So when you wake up in the morning, you have an exec. If you're an account manager, you have an exec that's connected to Cloud, powered by Cloud in the background on a separate machine, but it's connected in Slack.
(21:09) And you communicate it like that. So it can be done through your phone. That's super cool. That's super cool. So, yeah, at the end of the day, then you get sent in kind of a flag or any actions that are needed because we usually have external and internal Slack channels with clients. So we have our own Slack channel where we talk about stuff with the design team or with the data team or anything else like that.
(21:26) And then the communication with the clients. And it scans all of that throughout the day and gives a daily brief of like, this is what's happened. This is what you've done. This is what you may have missed that you need to update before you leave. And so that's a very clear kind of close of day, making sure that everything is done.
(21:40) And we've seen client satisfaction results really increase. People are super happy. They feel like we're on the ball. And we are. But this is a real helping hand in that. So that's kind of what about your team? How did that they kind of adapt this new world of agents, giving them a hand, basically? How has that been? It's good.
(21:59) You know, the way I look at it is we've got the right people on the bus. The people who are working at this company are very forward thinking, very innovative and very focused on how they can improve themselves and the work that they're doing, make a more meaningful change. So it works if you have the right mindset.
(22:15) If they come in and have fear that they're going to be replaced by this, they're going to be rejecting of it straight away. The ones that stood up and went, I can make my work 10 times more effective by using these agents and this product that we're building as well. I think it's generally the type of people we want to work with.
(22:30) So the vast, vast majority of people in our company took this and ran with it. A lot of this stuff I haven't built myself. The team, we're building it as we go. So it's really, really exciting. That's really cool. One thing I'm curious about is like, we know that Meta is coming out with more tools as time goes on.
(22:48) We know that AI tools, right? We know that the way that Meta serves ads with, you know, Gem and Lattice and all that shit continues to evolve. You know, like what, and you've been in the agency space a really long time. Like, what do you think is the future of what your ad agency agency is? This is it. I mean, what's the race to the bottom there is the way I look at it, right? What's the lowest common denominator you can get to? And that's the individual business, a single person running an entire business.
(23:19) That is, I'm on podcast here talking about it, but that's not going to happen, right? So where's the low bar in that? What's the race? What bottom are you getting to there? And to be honest, I think it's not as low as you might think, because if you think back to, you know, just five, 10 years ago, Meta have always been kind of trying to make it easier for advertisers to run ads.
(23:40) It's easy to run ads on Meta. You can just turn them on. You get in, you throw your card in, the way you go. That's simple. Running ads is fine. Running good ads is impossible. It's really, really difficult to always just look at DTC Twitter. Look how difficult it is to run ads and actually make a huge amount of money.
(23:56) We spend literally millions a month and actually do it really, really profitably for every client that we've been working for the last six months and being really, really profitable. What I would say to you is that didn't come by accident. The decisions that were made may not have had all of the data that the AI may have.
(24:12) It may be some software metrics and everything else. There's a human element to making decisions that what AI cannot capture. It's just it will never be able to because you have to then program it to feed in this information and that information. Look, context windows can get better. You can feed more information in, but there will always be something missing, something there as intangible, that limbic system that we're missing in an AI.
(24:35) But what I will say is when you start to think about building an AI for somebody, again, the way that we think about it isn't necessarily as building a replacement for a human, as enhancing a human. So the people who think of this as reductive are the ones that are going to lose out. The people that think of it as additive and kind of increasing where we're going, but the people that are going to go.
(24:55) Look at the, I think it's NVIDIA, the likes of those kind of guys. When AI started to come in, they go, right, well now we can go and do this and this and this, rather than looking at the other kind of large Fortune 500 companies that just cut half their people because they're like, we can just AI this. Those are the people that are going to lose out versus the people who are trying to scale with AI.
(25:13) So to answer your question in more of a short form, I think it's going to be in two years' time, I say that we're going to have the same amount of people. We're going to be helping a lot more clients and it's going to cost them a lot less to work with us. I think it's interesting. Like, I think it's already, it helps you unlock so much more in terms of finding new places to advertise, spooling up landing pages that are better because they take every learning you've ever had.
(25:41) Right, like Cody Plofker talked about in a previous AI D2C WTF episode about how he basically, when he builds landing pages now, he takes every call he's ever had with any CRO firm. He took like all these PDFs he's downloaded from CRO experts. He took the literal entire text of a book about a CRO, like from a CRO expert.
(26:05) He did, and like, it's like seven things and he's 10 for 10 on the landing page that he's rolling out, the PDP, improving CVR by 5 to 10%. So he's increased it by almost 100% in two months. And like, that's what I'm saying. It's like, it's crazy how big that can be. And I think it's going to, you know, I think that we also, a lot of times people get worried like, oh, wouldn't the client do this themselves as an agency or whatever.
(26:36) But I think you have to remember, like, I could paint my house. Like, I, you know, I could plumb, I could go on YouTube and plumb my shit, but like, I'm not going to. Like, you know, because it's not. And so people are always willing to pay for an expert. And so I think that staying on top of this stuff, and I'm not saying you have to adopt it all or any of it.
(26:56) I mean, some of it probably, but like being on top of it's massive. And being able to say like, yes, let's try that. Because that's a huge value add. If you're able to say, look, we have three people working on your account, but it's the equivalent of like 15 people. Yeah, exactly. That's a crazy like value prop, you know? The thing I would say to your point there is like, yeah, Cody was able to do that 10 for 10 the whole way up.
(27:19) Cody was. But John down the street, he's not going to be able to do that. So it took Cody 10 years of actual working on stuff, probably longer. Sorry, Cody. But longer for actually go through all of this and say, right, you know, you've learned the basics of it. It'll spit it out and it'll be like Will Smith of, you know, 2023 AI.
(27:38) It won't be the Will Smith of today AI, you know? So it's like it will spit it out, but you still need someone to look at it and go, ah, that's crap because I need to move that there. If Cody spat that in and put all that information in and iterated again, 10 for 10 and didn't touch it, that's when it becomes, okay, this is autonomous, but it's not.
(27:56) Because Cody went in and made those edits and did all the changes, which is perfectly fine. But to your point, Andrew, you could paint a house, but you're not going to, you're going to pay the guy or the girl to go and paint the house and go and get it done right. Yeah. I think it's, I think it's really interesting.
(28:09) I think, you know, the way that you've gone, gone and sort of looked at like enhancing the employees, do, do employees get bonuses for innovating on AI? Like, or how are you incentivizing them? Because a lot of times I feel like the top down edict is really hard because it's like, do more with AI. And then everyone's like, when? Yeah. Yeah.
(28:31) And so you're like, so how do you incentivize employees to be like, Hey, like if you save three hours of your day, like go be with your family. I don't care. I'm still going to pay you the same. Or here's a financial incentive. If it's something the entire team can use. The reason I ask is because talking about Kaizen, like when Kaizen came out, the process of Kaizen in Toyota in like the sixties, they paid proportional to the amount of time that it saved every employee a bonus.
(28:57) They saved like three minutes of time. They paid like whatever it was. And if they saved 30 minutes, it was like much bigger. And every employee was incentivized, even on a micro adjustment to, to do that. So I'm curious if you're doing that. Yeah. I guess the, the, I mean, that worked super well because the, the, the, from a Toyota standpoint, because the goals of the company was to get cars out fast.
(29:19) Right. And so that was what they, they wanted to do. Good quality and get them out fast. Our goals of our company is one to improve the performance of companies, not just get shit done. Right. Get shit done is a function of what we're trying to get to. So while we try and incentivize, we try, we focus on performance of clients.
(29:34) We focus on, on happiness of clients and we bonus them based on that. If they can achieve that in a shorter amount of time, great. But our focus is on keeping that high level, high standard. And we pay on a quarterly basis where we say, if a client is higher on a certain CSAT score, which we would then record through Kaizen, which also, by the way, takes in Slack communication as well.
(29:54) But then we have that and then performance of clients as well. So as we start to scale the client's results, that will then be fed back into our, our people that are doing it. So they're incentivized to improve performance. And if they can use AI to do it quicker, go be with your family. So it's kind of a mixture of both.
(30:10) But we want to make sure that our incentivization is lined up with what we are trying to do as a company. And that's, you know, create commercial masterpieces. That's what I talk about all the time. Yeah. Yeah. Great. Thomas, any other questions? Oh, I'm just excited. And I think this just reminds me, and I think in every interview you've done so far, AI is great, but you need experience and you need some, you need to know what you're doing to actually get this to work in a high quality way.
(30:41) And I, I'm not afraid at all that AI is going to replace my team or myself. It's just going to enable us to help more people. So I think it's just really cool. One more thing on that. And I think this is super cool. We built a very specialist agent that is, is a kind of a different tact to what we're talking about here entirely.
(31:00) But we, we call her Mia. She's a very special agent. And she's separate to everything we talk about here in some parts, but kind of still connected. So Mia is a Google ad specialist. What she allows us to do is a lot of the kind of the grunt work, the basis work. So she runs a service, so she's obviously there access to everyone, but she does research across all Google accounts to understand what a search intent is happening across their accounts with a very similar level of accounts.
(31:24) She'll then bring up kind of search terms that are relevant for like when she looks at the search volume. She'll bring up search terms that we can bring into keywords. She'll bring those into shopping titles and actually be able to work out what we need to turn negative in terms of non-branded keywords. So a lot of the basic works that are in there, she's just started two weeks ago.
(31:43) We've started running that, but actually started to see a measurable increase in terms of performance. Where that's starting to go is actually that's starting to build out that autonomous, autonomous buying. What's really interesting though, is where we're starting to link that into our communication. So when we're talking about what's performing and what's not, that then feeds that into the direct client communication.
(32:04) She's building out Google performance through AI specifically. So we're just, we're hedging a little bit. So we're still focusing on ourselves, but we're building out platforms that are allowing us to actually buy autonomously as well. Because agentic buying is going to be, it's not even six months away, I'd say it's three months.
(32:20) It'll be big. I mean, I just was talking to a colleague yesterday, his team, they spend about, you know, six million a month, probably across the brands they work on. And they've been doing agentic buying and it's like beating any baseline of a human. And so he's like, I don't know how quickly we're going to roll this out, but it'll be faster than I thought.
(32:40) So it's interesting, like where we're headed with that for sure. That's a whole other episode. Tris, thanks for joining us, man. Pleasure. I really enjoyed it, guys. Bye. High-end. Bye. We'll see you next time. You
Made With AI