In this illuminating episode of the eCommerce Evolution Podcast, I sit down with AI marketing expert Russ Henneberry to explore the transformative potential of artificial intelligence in eCommerce. Russ shares invaluable insights on how we can leverage AI as a powerhouse marketing tool by creating customized AI marketing hubs. This episode is a must-listen for any eCommerce professional looking to stay ahead of the curve and harness AI's capabilities to drive business growth.

Key topics covered:

• The concept of "AI marketing hubs" and how to build them using tools like Claude and ChatGPT.
• Strategies for fine-tuning AI models with your business data to dramatically improve outputs.
• The three key AI roles in marketing: Scribe (content creation), Optimizer (asset evaluation), and Strategist (big picture planning).
• Tips for crafting more effective AI prompts using the RTIF (Role, Task, Input, Format) framework.
• The future of AI in marketing, including predictions on how major tech companies will shape the landscape.

Tune in to discover how you can take your AI marketing game to the next level and turn generic AI outputs into tailored, high-impact marketing assets for your eCommerce business.

---

Chapters:

(00:00) Introduction 

(05:14) Mistakes and Misconceptions About AI

(23:02) The Role of Prompts In AI

(27:57) Introduction To AI Marketing Hubs

(45:16) Examples of Well-Built Marketing Hubs

(51:20) What About Gemini? 

(54:14) Conclusion

---

‍

Show Notes:

• Russ Henneberry 
• theClick
• Molly Pittman
• Sway Thai
• Torchy's Tacos
• Nvidia
• ChatGPT
• Anthropic
• Claude
• Perplexity 
• Gemini
• Llama

---

Connect With Brett: 

• LinkedIn: https://www.linkedin.com/in/thebrettcurry/ 
• YouTube: https://www.youtube.com/@omgcommerce 
• Website: https://www.omgcommerce.com/ 

---

‍

Past guests on eCommerce Evolution include Ezra Firestone, Steve Chou, Drew Sanocki, Jacques Spitzer, Jeremy Horowitz, Ryan Moran, Sean Frank, Andrew Youderian, Ryan McKenzie, Joseph Wilkins, Cody Wittick, Miki Agrawal, Justin Brooke, Nish Samantray, Kurt Elster, John Parkes, Chris Mercer, Rabah Rahil, Bear Handlon, Trevor Crump, Frederick Vallaeys, Preston Rutherford, Anthony Mink, Bill D’Allessandro, Bryan Porter and more

‍

---

Transcript:

Russ:

We don't need AI to get better. You need to be better at using ai.

Brett:

Well, hello and welcome to another edition of the e-Commerce Evolution podcast. I'm your host, Brett Curry, CEO of OMG Commerce. And today we're talking about AI and marketing. What are mistakes and misconceptions that people make about ai? How can you turn AI into a powerhouse marketing tool or marketing assistant or marketing team member for you? And my guest is a long time friend, multiple time returning guests to the pod. Although it's been a while. It's been too long. I'm sure it's been a long time. It's been a long time, dude. Yeah. But Mr. Russ Henneberry of the click join the click.ai. We're both Missouri boys. He's in St. Louis. I'm in Springfield, so representing Missouri. But Russ, what's up man? How you doing? And welcome to the show.

Russ:

Welcome back to, I'm so excited to be here and I haven't seen you in person in even longer.

Brett:

Thanks.

Russ:

It's been forever actually. I saw you

Brett:

In Austin, wasn't it? We were both doing, why were we there? It was something with Molly and Sam and I don't know why were, were we in Austin?

Russ:

I'm not sure. You were in an e-comm event. I was at Kajabi's event.

Brett:

Got it. But Molly Pittman was there, wasn't she?

Russ:

Molly Pittman was there.

Brett:

Yeah. So we hung out, waited at a great restaurant, Austin man, Austin does food, right?

Russ:

Sway.

Brett:

We went to Sway. Sway. Look at your memory. Yeah,

Russ:

Sway. Because I go there every time I go to Austin. It's the best Thai food in the world.

Brett:

I love Thai food, man. I am a huge, huge Thai fan. And you had some kind of funky congealed beer or something, if I remember

Russ:

Right. Molly bought me a beer that slowly freezes after you open it and then,

Brett:

Which shouldn't happen. Alcohol doesn't freeze, so it just kind of turns into a gel or something.

Russ:

I can't remember what the heck happened, but a mess, hijinks ensued and Molly just sat there and laughed the whole time.

Brett:

She was cackling. It was entertaining to watch the beer congeal or whatever, and you try to deal with it, it was equally as funny or more so watching Molly's amusement at the whole thing,

Russ:

You basically just, you walk up and you hand somebody a problem and they have to deal with it. You basically have to drink.

Brett:

I'm going to watch this unfold. I'm going to see how you do there, buddy. How are you going to do with this congealed beer? But Austin such fun town to hang out in. They do food, right? Tacos. Oh my goodness. I was just in Dallas for an event with my family and Torchy's Tacos, Austin original. It originated as an Austin food truck. They're in Dallas now. So I ate some, oh my gosh, that was one of the best tacos I've had and in a long time. So Torchy's tacos, highly recommend sway if you're winning Thai food. So there you go. Some Austin recommendations for you. Free of charge on the eCommerce Revolution podcast, free charge, free podcast. That's awesome, dude. So let's dive in and let's talk AI a little bit. Obviously it's the buzzword. Hopefully everyone here has invested in Nvidia stock from way back in the day, and hopefully you're wildly rich now. And AI is all the buzz. We're not really sure who's benefiting from it other than Nvidia who's really benefiting from ai. But you and I are both very bullish on the technology being at least somewhat disruptive now. It's going to continue. But why are you so excited about ai? So much so that your membership site is about ai. I know a lot of the consulting you do is about ai. Why are you so pumped up?

Russ:

Well, because my background, I know yours we're similar age. We came up when the internet was becoming a thing. My first job in marketing, I was in direct mail and then I saw this internet thing,

Brett:

Yeah, was TV radio for me.

Russ:

So we were both doing more old school types of media and we probably, I saw the internet coming along and I was like, I'm hopping on this train. And I never looked back. And I rode that wave. And when I saw this coming out, I pumped the brakes for a while and was like, okay, I'm not going to say anything about this. I'm just going to watch it. I'm going to study it. And eventually I've come to the conclusion there's no future that I see where we aren't using AI isn't just infuse everything we're doing as marketers. So yeah, I'm all in on it.

Brett:

Yeah, I 100% agree with you. I've kind of gone personally from being fascinated by it, maybe partially afraid of it, skeptical a little bit to like, okay, no, this is going to happen. It's going to be good, but not really using it to, now I utilize AI quite a bit. I'll share some stories as we go here, but let's maybe open up with what are some of the mistakes and or misconceptions people have about AI related to their marketing?

Russ:

I think people are frustrated with ai. You mentioned earlier that where are we seeing the ROI in marketing? There's a lot of investors asking if there's a bubble here because there's so much money flying into ai, billion,

Brett:

Billions and billions and billions of dollars being invested in AI right now.

Russ:

And where's where are the sales? So what I think we're seeing right now is that intelligence has become a commodity like electricity or internet access. At 20 bucks a month, I can have access to the leading Frontier model, the most intelligent model. And that's not going to change. We're going to see chat BT five come out, chatt six, claw five, whatever that comes out. And we're going to be able to have commoditized access to intelligence. So what I'm seeing is that it's going to be how we use this resource, just like the way we use the internet. So just because everyone has asked us to the internet doesn't mean we all used it to our best capabilities. Some people won, some people lost, some people won big, some people won small. So it's going to be the same thing here is how is it that we go about using this and the business?

And probably not just the average, but the majority of businesses that I see when I talk to them are using off the shelf ai. So they're using maybe a paid license to cloud 3.5 sonnet, which is right now, a lot of people say the best model, I would say is the best model out there or chat g BT four oh or whatever it is they're using. And they may be paying 20 bucks a month for it, and they're just using the generic off the shelf ai. And what they're getting back from it is generic boiled chicken output. So

Brett:

Boiled chicken, boiled chicken is not something you're going to find at many Austin restaurant menus. So if we're going to just harken back to the good food, it is not boiled chicken. So just wanted clarify.

Russ:

Torch's Tacos has a boiled chicken taco though on Thursday.

Brett:

Do they? Really? No kidding. They do have the trailer park, which is the fried chicken and dude that's so Trailer Park, I love

Russ:

It. But I mean the key to this whole thing that we're seeing reports coming out of places like Accenture, those big consulting companies, Accenture, Deloitte, Dell, IBM, saying that what they're seeing is that Billings for AI consulting are going to go up 400% over the next five years and they're growing 32% year over year right now. And so I started to dig into what it was that they were selling to big companies like Best Buy or Target or whatever, like their clients. And what they're selling is something called fine tuned ai. And this is really the key to unlocking the potential of what AI can be for even a small business. Forget Best Buy and Target. They're paying these big consulting companies to come in and do complex fine tuning projects, but we can just fine tune $20 a month chat, GPT or Claude and get incredible results out of it.

So this is kind of the key is to fine tune it. What fine tuning basically means is that we are going to take an existing LLM model like Claude or chat GBT and adapt it to the use case that we want to use it for. So to keep on our food analogy, if we plugged all of my recipes from my family or something into Claude 3.5 sonnets brain, I can then start to have conversations with that version of Claude about my marinara sauce or about my Texas beans recipe, which if you're real nice to me, I will send you

Brett:

No way. The family, the Henneberry family Texas Beans

Russ:

Recipe, this is the Russell Texas bean recipe. I found it from being down in Austin and going all those barbecue joints and I was like, I got to figure out how to make these beans, but I can

Brett:

The marinara, not to get too focused on food here, but the marinara, you're in St. Louis. So a lot of people don't know St. Louis home of amazing Italian food.

Russ:

It's on the hill, man.

Brett:

It's on the hill.

Russ:

My marinara, I'll put it up against anybody.

Brett:

No way.

Russ:

But I did demonstrate this. I went to Claude and I said, I put all my recipes in there and I said, I just asked simply how could I improve my marinara sauce

Brett:

Recipe?

Russ:

And it says to me, you're not simmering long enough. It went and looked at my recipe. It's like you're not simmering long enough. Interesting. You're using dried herbs. Try using actual fresh oregano, fresh basil in your sauce. So it went and consulted my recipe. Right.

Brett:

Did you try any of those recommendations?

Russ:

Not yet.

Brett:

Okay.

Russ:

I'm going to make the Texas beans though this weekend.

Brett:

Well, let's see if it's

Russ:

We're going to go

Brett:

Russ Original or Russ plus clawed. What does better,

Russ:

I'm going to ask it about how to improve my Texas beans risk. But the thing is, when you go and you customize an AI and you can do this so easily, this is not a technical problem at all. Like Claude has a feature called projects where you can start a new project and upload information into its knowledge base and it will consult that knowledge before it's first and then on it'll

Brett:

Consult that knowledge plus use the llm and now you've got specific insights.

Russ:

That's exactly right. And this is the thing that people are really missing the boat on. And I think a year from now, everyone's going to know this intuitively because there won't be another AI model that rolls out that doesn't allow us to customize it easier and easier. So they are going to make the user interface and the ability to customize AI and fine tune AI to our circumstances easier and easier as it goes along. I'm sure chat, the next iteration of chat GT will make it even easier. Right now what you have to do is go build a custom GPT, which is actually quite easy to do as well. It's a little bit clunky, and I think by the time we get a year from now or whatever, we'll think that the interface with both Claude and Chat GP PT was very clunky. But right now you can already go in and start to fine tune AI for your use case, which is you're a marketer. So we are going to fine tune with the typical marketing stuff like customer personas offers that are very well written out in detailed so that because if you think about it, an off the shelf ai, it's ingested the entire internet.

Brett:

And

Russ:

So it's incredibly knowledgeable, but it also knows absolutely Jack Squat about your business or any business that you're marketing. So I've created this process, I call it the AI Marketing Hub, and all it is is just if I build one in chat gt, it's a custom GPT with custom knowledge about the client's business that I'm working on, or if I'm working over in Claude, I'm going to use projects to create an AI marketing hub. And it's sort of like a central command center from which I'm going to do any AI work with that particular business. So inside that hub, there's custom knowledge about the circumstances of that business. And then I have another hub for, so it's sort of like having an MCC for a client in Google this. Every client has their own

Brett:

Place, account hub accounts, it's all kind of connected a little bit. So let's do this. So fascinating concept AI hubs. We're going to dive into that, really explore that. I know that ultimately this is more about AI is really just a tool, so it's more about what you do with it and how you use it that really matters. But you did say something that was interesting. You said, Claude maybe is your most advanced model or your favorite model. I still talk to people that I'm like, I use Claude. And they're like, who is Claude? I'm like, well, Claude is AI right by Anthropic. And I use chat GPT too. I've got chat GPTO on my phone, but for anything creative writing, I actually use Claude. I really, really, really like it. Compare and contrast the two just briefly for those that are wondering and those that are new to Claude,

Russ:

Well, I'm using Claude for the same purpose. You're using it for any sort of content creation or copy creation. It is superior, my opinion to chat GPT. The problem with Claude, right? The reason that I end up switching back over to chat GPT quite often is because it doesn't have certain features. So for example, Claude is not Web Connect. So one of the best use cases for using chat GT is to kick over there. And I might have it go look at a webpage for me. Like, Hey, go look at this landing page and evaluate X, Y, and Z on there, or pull the copy in here so I can work on it or whatever it is I'm trying to do. So Claude for whatever reason, is still not web enabled. It will be that this is one of the places we're seeing lots of development is in these kinds of user interface type things.

It can't accept video, it can't accept audio. So the multimodal capabilities of Claude are much more limited than you get over at Chat G gt. But overall, if I'm working with text, I am going to use Claude. And then the other big thing over at Claude is this concept of projects, which we talked about and is so critical to building an AI marketing hub we're talking about and sort of fine tuning a knowledge base. But the other thing that they've released that is very cool is called artifacts, which is essentially where you're able to talk to the AI on the left side of the screen and on the right side it's building something, an artifacts what they call it. So it could be a document, it could be an application, it could be interesting, it could be a set of graphs or a dashboard or something. So you start to see where AI is going to go is do you, Brett, you remember when the first CMSs started to come out like JU Law and Drupal and

Brett:

Yeah, early

Russ:

In the day,

Brett:

Absolutely.

Russ:

Before that we were using Dreamweaver or writing custom websites and stuff like that. And once those CMSs came out, we saw such an explosion in content produced on the internet. It was like we have these, what you see is what you get editors, we call it the wyg. Wizzywig is such a huge thing because it was like it took coding out of the game at least for producing, just being able to publish the internet. Well, what I'm seeing happening with artifacts and how when I started to play with artifacts, I started to feel like I had this magic wand in my hand and it was sort of what you say is what you get. So if you can describe it in the English language, I saw somebody on Twitter the other day say something like the hottest new programming language is English,

Brett:

And

Russ:

I love that you say what you want.

Brett:

Can you ask good questions and can you articulate? Which you don't even have to be all that clear, dude, that's so good. And I'm a big fan of Claude too. Obviously you got to use multiple tools. So this is not this tool versus that tool. It's important to be educated. You should probably play with and use Claude as well. But yeah, what I'm using it for a lot of times is transcripts from podcasts or transcripts from interviews or transcripts from long videos and stuff where it's like you upload that transcript and then you can just converse with Claude about that transcript. And I've used it where I have a PDF had this offer from an investor or whatever, and I looked at it and I was like, Hey, pull these numbers from this image or from this PDF, put it into a spreadsheet for me. Okay, now clean this up, change this. It's awesome dude. And I love that. Now the best programming language is English. That's really, really smart.

Russ:

It's going to head to where if you can understand how to articulate what you want. So I really think people that are in the creative world are scared that AI is coming for your job. And there is some truth to that. I mean we just had reports that Amazon, it's Q program or whatever, just replaced 4,000 years of programming time, crazy,

Brett:

Crazy

Russ:

Using ai and it's 79% accurate. And the question is of course, what are they going to do with that? How does that impact programmers at Amazon and everywhere else when we're starting to see this? So there's that. But the exciting thing about this for me is that if you're a creative person and you've ever wanted to build something but you're like, I don't know, I'd love to build that, but I'm not a videographer or I'm not a coder or I'm not, whatever it is that is that barrier, they are going to remove all of this for us. It is going to be literally a magic wand where it's what you say is what you get. And if you want to see this in the earliest infancy, go play with Claude 3.5. It's even the free version has artifacts. And just ask it to build you a landing page for example, and it will definitely activate Claude's artifact function.

It'll start to build the page on the right side and you can talk to it on the left and say, move that button down, change the color to this. And it's just going to keep rewriting that code. And I mean the website will look like it's from 1998 because it's early days, but this is just what's interesting about it is that communicating with something and just sort of casting a spell, right? Like hey, switch that over to this, switch this over to that, and it's just speaking English. That's the programming language you're using because it's going to write the code for you in the background.

Brett:

Totally. And I'm a huge believer that there will always be a need for smart humans. Of course, we've talked about this a long time in the agency space. I remember when years and years ago, Google first came out with smart bidding, with machine learning based smart bidding, and a lot of agencies were terrified and they were like, this is not good. This is not going to be good for your business. Google is evil, whatever. I choked up, hold on. I was on a rant. I got choked up. Oh geez, this is not good for your business. What we found is it really did not limit people's desire for agencies. It kind of just allowed us to do more. Now we could focus on more things rather than just bidding. And so the way I've heard it described, which I really believe is true, you're not competing against AI per se. You are competing against people who are using ai. And so it's always going to be a partnership. Smart humans plus AI is likely going to be the future. And so is then the solution. Russ, just better prompt engineering. Is that it? Do we just need to get better at asking questions we need to get better being more clear, more succinct, more whatever? Or is there something more?

Russ:

Well, so when AI came out, I told you I'd pumped the brakes and started watching it. I saw courses coming out already. Some of the big names or whatever that are always sell course exactly four days after chat toed is out, there's a course on it. But I was basically like, we need to study this, we need to watch this. And one of the things that I saw coming out a lot was, here's a hundred prompts as lead magnet kind of thing, and here's some different things around prompts and there's no question about it. Better prompting gets you better output, but it still doesn't solve the problem that ai if it's not fine tuned. In other words, if it doesn't have any information about your business and its knowledge base, it's always going to give you generic information. I mean, consider that if you hired a consultant for your company, a human consultant, and that person came in, they were very capable and very knowledgeable and very smart, but you blindfolded them, didn't tell them anything about what they were selling, who you were selling it to or anything about the business it was trying to market.

And then you ask it questions like we ask AI sometimes like, Hey, help me build a webinar series or what new products should I roll out for Black Friday? Stuff like that. And that consultant is going to give you maybe good generic advice and that's what we're getting from ai. So that consultant, if that consultant knows what they're doing, which in this case it would be the metaphor is that AI does know a lot, it's very knowledgeable, but it doesn't know your business. So a better prompt, being able to ask the AI in a more clear way is going to get you better output. It's still not going to get you there until you actually take the time to fine tune it or add to its knowledge base.

Brett:

I want to talk about the knowledge base and the AI hub because that's really a neat concept that you've kind of pioneered here and I want to dive into that. I do want to talk a little bit though about better prompts because I think this is important. Used to do a lot of customer testimonials. We used to do video interviews with customers and then create customer testimonials from that. And I promise you, the person behind the camera, which for a long time it was me, the person asking the question, the quality of those questions, the sequence of those questions in this case, because it was human IT enthusiasm, it's an art, you would get lame answers, boring answers, timid answers, afraid answers, or you could get dynamite stuff and it was based on how you asked the question. And I think the same is true if you're doing surveys and stuff like that, the way you ask the question, you want better answers, ask better questions, a little bit different because AI is not human and not emotional, but any tips on better prompts just as a baseline that I think some people will really benefit from before we dive into the Yeah,

Russ:

And that's such an important point. I mean you're right. Surveys. I mean look, we're all marketers, so we know words matter the sequence of when we say things. And so there's no different here when you're prompting. And so what I like to do is keep in mind four, this little four letter acronym, RTIF. So when you build a prompt, if you can start out with telling the AI what its role is, that's the R part and RTIF. What is the AI's role?

Brett:

Pretend you are a marketer, copywriter, director, editor, whatever.

Russ:

Exactly. And what that does, the issue is that large language models are large. They've ingested the entire internet. So that role immediately zeroes the AI down into some domain of knowledge. So if you tell it it's a copy editor, then it's immediately going to zero down because right now you start with a fresh generic out of the box, new chat, you could literally ask it anything. So it's literally accessing the entire index of everything it has, which is everything that we've ever produced and put online. So the minute you give it a roll, you're going to get better output because it's going to be like, okay, I'm zeroing in on this domain of knowledge. The second thing to do is to give it a task. So if you give it a roll, you're a copy editor, your task is to review these product descriptions. And so now you've zeroed in even further, you've given it more clues as to which domain knowledge that you want it to access inside of a sprain, and then we give it input.

That's the I. So role task input. Now input is super important. If we can spend more time with input, the better. So one thing you could do is you could say, here's the product description I want you to copy at, and that's it. So that's all the input that you give the AI is you are a copy editor, your task is to copy edit these product descriptions, here's the product description, edit that please. Then the last one is format, which is the way you want that output formatted. So it could be that you want it long, it could be you want it short. It could be that you want it in bullets. It could be you want it at a table or you want it however you want that to come back to you is the last bit of a solid prompt. But let's kick back to the idea of this input is super important.

If you tell the AI that you want it to do to assume the role of the copy editor, edit these product descriptions and then you give it the product descriptions, you want it to edit, that's basically a zero shot prompt, what we call zero shot. It means that you've given it zero shots. And really, I don't know why these engineers want to call it shots because you could just use the word example. So if I instead say you're a copy editor, your role is copy editor, your task is to copy edit these product descriptions and then here's five product descriptions that are done right now. If you've hit five product descriptions, you have a multi-shot prompt. So you're going to get way better results because you're basically saying use these as a model. And so that is a major part of prep, preparing a good prompt is whatever it is you're asking it to do, can you give it anything more and show it something that is the proper output?

Brett:

It's really great, really, really great. Okay, so let's then talk about AI hubs, AI marketing hubs. How are you building these? What do these look like? Walk us through that.

Russ:

Well, so if we start with this idea, I mean we can go back to what I just said, which is when you give it something to go off of when the AI is fed something into its knowledge base. So it's like, let me give you some more context here so that you're not just again, pulling from this enormous index of information that you have. Instead, I am giving you examples, I'm giving you other things. So yesterday I was at a conference, actually I spoke at a conference and somebody from Google was speaking there and I went to her session and in the end I raised my hand and I said, so what you're telling us is that we need to be bringing our own data to the party, right? Because she was going through some AI enabled ad stuff that's now available at Google Cloud, and in the end what she was saying is you can bring your own data into Google Cloud and we can crunch it up and segment it and do all kinds of stuff and then use it for targeting, creating segments and improving your ad output.

And so at the end of it, I was like, I feel what you're saying is that you guys are doing all of this with Max and everything that you're doing. You're doing a lot of AI stuff already. You're doing a lot of machine learning and everything you're doing there, but that now we are able to bring our data in from Salesforce or HubSpot or our email analytics or whatever and crunch stuff up and she goes, yes, I really love that idea. I'm going to start using that of you need to bring your own data the party. And so I was like, I'm going to start using it too. I really love that. And so I want you to think about an AI marketing hub as a place where you've brought your own information and data to the AI party. So you've got the AI model and it's tremendously capable, we know that. But I think if you're like me, you are probably getting some pretty frustrating output from it. So you're just like, okay, this is good, not great. And as marketers we

Brett:

Know, and maybe that means it's entertaining but not usable. It was interesting that you created it and it's kind of impressive, but it's like, but I can't use this.

Russ:

I can't really actually use this. And so when you build a hub, whether it's through a custom GPT or through a Claude project, or you could go pay Accenture or Deloitte to come in and build an entire model off over the top of llama for billions of dollars, which is what they're doing, they're coming in and charging lots of money to customize AI for all kinds of use cases for the biggest corporations in the world. But you and I, we can take our small business, our mid-size business and we can open up an off the shelf account over at Chachi PT or with Claude, and we can start a project or a custom GPT and we just add to that project's knowledge base with information that's going to be relevant to the type of marketing that we're doing. So if we start thinking about what kinds of things you might bring to the party, the first three things I recommend because there's just almost no marketing case where we don't want to know who we're trying to sell to and what we are selling. So if nothing else, you want to build an AI playground, I call it an AI marketing hub. So use a project or a custom GPT and then upload customer personas and detailed information about offers. And this is just ugly documents in PDFs that it can read. So if you're thinking

Brett:

Clearly outlined that, hey, this is what this PDF DF is, this is what this slide deck is, whatever.

Russ:

Yep. Yeah. If you were going to unblind fold that marketing consultant we talked about earlier and begin with how do we get more out of this marketing consultant we're paying, the first thing you might tell them is, here's our customer and here's what we're selling, right? Because without that, I don't know how we're doing any marketing about anything. You're absolutely working in the blind. You'll be shocked by how much better your AI performs with just a little bit data points. If you have a good well constructed persona and make sure that you've elaborated on the most important part of a persona in my opinion, which is the pain, the challenges that cut. If you were just going to give me one piece of information about somebody that I needed to sell something to, it would be what's the problem? So make sure that part of that persona is very well fleshed out, but you build a persona, however it is that you build a persona. There's no shortage of trainings out there about building personas, but you just need to get it into a readable format. PDFs are great. You write it in a Google doc, create it in a Google Doc, and then you attach it to a project or a custom GPT and you do the same thing with whatever. Take one of your bestselling products, for example, and do a nice description of that. Maybe pull the product description

Brett:

Out so it even be like the product detail page, but put into pdf DF form or something.

Russ:

Yes. Yeah. And so you say, how do we sell more of that offer this PDF to this customer and the AI will be like, let's talk about that. I got a lot of ideas and you'll be, so what I think a lot of people are doing is they're sitting there thinking, I keep getting this average output from ai, so I really am just going to kind of put this on the side until AI gets better, until chat two, PT five comes out or the next cloud. I mean it's good. It's not great and I want to produce great stuff. We have to be producing great stuff. That's what the best marketers do. We don't build average stuff.

Brett:

And

Russ:

So this is, I think a mistake because we don't need AI to get better. You need to be better at using ai. And so if chat g PT five comes out and it's smarter and faster and stronger and able to leave tall buildings with single bound, it still isn't going to know anything about your business. So you've got to learn this skill, which is basically how you fine tune an AI and you don't need to know all the fancy dancey fine tuning tricks and moving weights and all of the weird things that engineers talk about, but we need to understand is what do marketers need to do to AI to get better output from it? It starts with giving it things like customer personas and offers.

Brett:

Yeah, I love it. So I want to dig into some other details here. Maybe get a couple of examples from you too, how you've used these for different clients you work with. So personas and product details or offers. If you want to go next level though or give it more information, what else are we sending? Are we sending things like customer reviews and competitors and things like that? What else are we putting in this marketing hub if we want to go next level?

Russ:

Alright, yeah. And what want you to think about is imagine that anybody that's working on that business, whether it's a client or whether it's your company, is working from an inside of the hub. So now it becomes a game of what else could I give the AI in terms of context and information that would make it stronger and make it more useful to me? And what you find is that it can assist with everything from media buying, data analysis, content generation, product marketing, whatever it is, there's almost no scenario that it won't touch. And so other things that I have bolted in here at different times based on what it is we're working on is you might bolt things like data on there like sales data, inventory reports, website analytics, email reporting, things like that. So you can plug data into its brain and have it understand that first of all, what I like to do is give it a document and I just say simply to it, what do you see there? Because I want to see if it understands,

Brett:

Is it understanding this?

Russ:

Yeah, I'm sort of checking for its understanding. And what you'll find is that it does understand your sales reports, it understands your inventory reports, it understands these things like your website analytics. You can plug things like reviews on products in there, writing samples. If you're trying to create an AI that can more easily write like you, going back to what we talked about with multi-shot prompting, can you give it good writing samples? Here's 10 emails that I wrote already over the last month that are good model when you're writing this promotional campaign, you can put anything in there. If you've got a loyalty program for example that you're working on and you've got details around that, attach it in there and say, Hey, this is our loyalty program. Just so you know, we got a loyalty program, big market analysis, but it does become this sort of creative game, which I love Brett. It's like, what else could I give it?

Brett:

Right? That is how would that change the output and change what it actually creates better,

Russ:

Right? It's such a creative process. And so depending on what it is that you're working on, you just want to think. And if you really think of it this way, if I was working with a human who I'm asking to do X or Y or Z, what would I want them to have access to? Do they need access to my inventory reports to do this work? Do they need access to my website analytics? Because if the answer is yes, they would do a better job with these three different resources, then you ask yourself, how much effort would it be for me to get that into a format where I could bolt it into the brain of the ai? Well

Brett:

Treat it like a marketing treated like your assistant CMO or something. Treat it. I like that frame. What would a marketing employee need? Give that to the AI as well.

Russ:

And the thing is that we may not see that the next few models that come out are all that much more intelligent than the AI that we're playing with now. And we might see that they're become quite a bit more intelligent, but even if we don't, where we are going to see a lot of development in AI is going to be in its multimodality. So multimodal is an important place to be watching can the AI take in and put out, so what can it take in and what can it put out? So can it take in text image, video, audio, data code, and can it put out, so can it take in text and give you back an image? Can it take in an image and give you back a video? Can it take in audio and give you back a video? So that is where we are already obviously seeing right now, chat, GT four oh has vision. It can see things, it can talk to you. That's another modality. And when we think about how we are going to be able to give an AI context around our businesses and our customers and everything, as the modalities change, so will your ability to

Brett:

Put

Russ:

Different things inside of its brain. So if you could literally take, were we live when you were talking about the customer videos that you were

Brett:

Doing? No, I wasn't. No. I can frame that really

Russ:

Quick. Yeah, tell that story because I mean you could upload that whole video in there

Brett:

And you mentioned that Claude can't communicate with audio or video, but Claude is the tool I use most. And so had this project four client, they had five customers give these long interviews on video and ended up each one was like an hour. Now in the old days, we would sometimes pour over those. We'd watch the same hour like five times and try to piece out, pull out little nuggets of hey, this five seconds here, this 10 seconds here, we'd make for a great clip or great customer testimonial to put into a TV ad or something. So what I was able to do though with AI is I took the video, stripped out the audio, uploaded the audio to Riverside, which is a tool we're using now for the podcast. It will transcribe things. So created a free transcript. I uploaded that transcript to Claude and I said, Hey, here's the project, here's the product, here's what it does.

This is what we're trying to do, trying to create a TV commercial. And so find me five or six quotes that speak to benefits, speak to pain, speak to experience, and I need them to be about five to 15 seconds long and it spit 'em out like that. And so then I was like, okay, those are cool, but give me a few more or okay, give me a few more and make them longer or are there any quotes that talk about this? And then as I got to the ones I liked, I was like, okay, give me all of those quotes and put them in a time code spreadsheet where I can see start time, stop time and how long that quote is. And now this was all taken from tech. So Claude was like, Hey, these are going to be estimates, but we'll give you best guess. It was pretty accurate. So dude, this was something that would've taken days and could do it all in. I don't even know how long it took, but it wasn't long. It was 30 minutes. Hour or something like that. Just crazy.

Russ:

What's even crazier though is that you had to strip out all the audio you had to do. You had to actually,

Brett:

I had to do some work.

Russ:

You had to do some work to get it into a format that this tool could accept and that's all going to go away. Eventually Claude and every other model, it will accept all modalities and it'll be able to output all modalities. So in that case, you would've just uploaded the video and

Brett:

Then it probably would've spit out the video for me and then I could say, well, give me the video in this format.

Russ:

That's exactly, that's exactly right. So I think the big thing to think about here is that this whole idea of what data are you bringing to the party? Because if you're just using generic off the shelf AI and you're asking it questions like, Hey, can you make me some interesting things to post on LinkedIn about some topic? It's like, sure. One of the frustrating things actually for me about AI is that it always answers you. It will

Brett:

Always answer you, try again, try harder, dummy. That was a stupid question. Yeah, it doesn't do that.

Russ:

Yeah, it never says, I don't know. And it never, and even though perplexity one of the startup companies that's challenging, Google with ai search perplexity is starting to get a pretty cool interface in terms of trying to ask you more for more context and ask you questions and prompt you and have you select things. And that's what we'll see from all companies. They're going to try to make it so you don't have to be a prompt master, you don't have to really understand how the model works. But for now, if you ask it a question like, Hey, can you create me a LinkedIn editorial calendar? And it has no context about your business, it doesn't know who you're trying to talk to, it doesn't understand your offer, it doesn't understand anything, it will answer you. And then at the end you'll go, I guess I'll use this, and you're going to have a pretty boiled chicken bunch of stuff. Totally. Or you're going to be like, that was a big waste of my time. I'm going to have to go do this myself anyway. But then you go and you plug in all this stuff, you create a hub and the hub, the other cool thing about this hub concept is that it eliminates what I call a cold start problem.

So the cold start is, okay, every time I go to use ai, I have to reprompt give it all this information and have it understand this well, these projects and these custom GPTs have memory of their own and it's contained within that area. So it remembers things and it obviously is able to access things in its knowledge base, but it's also remembering things that are happening in those conversations so that you don't have this cold start problem, you just jump back into your hub and you start back up talking about the LinkedIn editorial or your next email promotion or whatever it is you're doing,

Brett:

Just like you would converse with your CMO or marketing associate or whatever. They've got context now you've been working on stuff together, now you can just ask a question rather than, okay, every time I got to give all this background, it already has it. So that's really cool. Can you give some examples and you practic to be careful on confidentiality and stuff, but examples cool things you've seen people do with these Well-built marketing hubs.

Russ:

So there's three roles that I like to build out and when I'm working with clients on these, I ask them, where do you see the biggest use case for these? And the big three places that I think AI can do better than the average human can is one role is I call it the scribe. So the scribe is like, it's your creative person, it's your creative role within this AI marketing hub. And so we take in, of course, if you were going to hire a writer or somebody that was going to help you with content marketing, what would you want them to have? You might want 'em to have things like editorial guidelines. You might want them to have access to writing samples that are good, maybe they aren't even yours. You go and look at a competitor or you look at some brand you admire and you say, let me pull these last 10 articles or emails or ads or whatever it is, and you say, these are good samples of what we want to be producing.

And you give that to the AI's hub, you give it to the editorial guidelines. I've built a very, very cool prompt that is able to extract the style of a writing out. I post on LinkedIn the other day, I said, and I decided to do this with an author, one of my favorite authors actually John Steinbeck from Wrote Grapes Rat and all that stuff. And I took the book of Mice and Men on PDF put it into Claude and I gave it this prompt and that prompt basically says it is a big long prompt that's like, I want you to look at this and this and this and this. Everything, the cadence, when you look at the styles, are they using idioms and metaphors and dah, dah, dah. So there's like a million things that this prompt is asking it to analyze inside of the Steinbeck piece or whatever kind of content you give, you give it. I also gave it 30. I went and just scraped 30 customer stories off of Salesforce's site and I ran the same prompt. What I'm trying to do is build a mimic, literally extract the style and the tone and everything about that data. And I went and I grabbed a paragraph from a real Steinbeck paragraph from one of his short stories, and then I actually had it write a short story Claude using this prompt that I've used to extract

Brett:

And it read of Mice and Men you gave it that

Russ:

Get rid of Mice and Men and it's also probably got a decent idea of who Steinbeck is. Anyway, but it was really this prop I think that went in there and was like, because it did the same thing when I ran it through Salesforce's stuff. And then I've created a section that's like write like Russ, so I want you to be able to write like me. And it still doesn't get me all the way there. But with the Steinbeck thing, it was crazy. I posted this on LinkedIn and I said, which one's Steinbeck and which one's clawed?

Brett:

And

Russ:

People

Brett:

Did people get it right?

Russ:

Swearing up and down, A, it's B, it's A, it's B. And I was like, no one could agree on which one it was. That's how good of a

Brett:

Mimic

Russ:

It was, right? So that's the scribed functionality. And what you want to think about is what could I give an AI to turn it into a scribe? The second functionality that AI is really good at is what I call the optimizer. So it is somebody who can evaluate assets like landing pages, ads, emails. And for this we use rubrics back in the old digital marketer days when we were building rubrics for people to say, here's what are the 10 elements of a good ad

Or here are the 10 elements of a good opt-in page, or here are the elements of a good email promotion, grade your work on this. We used to do these when I was with digital marketer. We would create, so instead of just intuitively looking at say an ad and saying, well, I see the problem with that ad, fix this. Instead, we would literally extract the knowledge of Molly Pittman, who's such a good ad person and put her knowledge into a rubric where it was like anybody could look at an ad and kind of figure out what's the problem with this ad? Where could I improve this ad? Well, if you feed an ai, a rubric like that, and then you show it a landing page or you show it an ad or you show it an email, it's able to really quickly, it scores it, it would like a professor would, but it's a professor that's got a rubrics working against for a landing page or for, and it's really, really powerful.

And so I call that role to optimizer. And then the last one's the strategist. The strategist is somebody you talk about long-term, bigger picture type marketing stuff and you feed it things like business competitor analysis, market analysis, SWOT analysis, bigger picture idea stuff and give it an idea of, and you can have these crazy long conversations with it about things like strategy. Like hey, I'm thinking of, I think I want to expand my product line in this direction. What do you think about that? What are some risks here? What are some things you can have these kinds of conversations so much context for it to be working for him. So it's pretty wild.

Brett:

That's great. So scribe optimizer and strategist, really, really great. Love that. So we talked a lot about chat, GPT and about Claude, what about Gemini, Google's Gemini? Are you using that at all? What are your thoughts? Any POVs there?

Russ:

Well, so my big point of view there is that Google, and I don't use Gemini very often very much, but Google has instant distribution. They've got AI built into everything. Just like my son, I just bought him, he's going away to college. So I got him one of the new surface laptops with copilot built in. So Microsoft has immediate distribution, Google has immediate distribution. And here's the big thing to think about that's connected to this concept of adding to a knowledge base is that Google already has all your information. So if you're using Google Workspace for example, they already plug, they can plug their AI into your Google drive,

Brett:

Legitimate right into drive

Russ:

Into your calendar, into everything. And that's where we're going with this is that the extension of the knowledge base. So you'll have Gemini, the brain, the off the shelf ai, but it will have access to, and it already kind of does. It's really early days here with that laptop that my son just got. They are not allowing certain functionality in there yet because they're saying that they're still trying to figure out the security and the privacy side of this. But that's where we're going is where you want to extend your knowledge base. We've got all your files, we have access to your Google Drive, we have access to your email, everything. So Google is a major player and while OpenAI feels like the cool kid on the block and Claude sort of feels like the guy from the village that's super like artsy and cool, that's fair. It's really going to be Microsoft and Google that. I think we'll pull ahead eventually because just

Brett:

Because of distribution and access to all the data. Yeah, it's a really good point. And Apple.

Russ:

So Apple, while they don't have a model of their own, they're just renting right now, they could easily go build one if they wanted to. And then you have Facebook, which is super interesting because they've decided to go open source and allow people to build on top of llama and it's as good as CHATT four. That's what I've heard.

Brett:

Lama's great and not played with it, but I've heard it's

Russ:

Awesome. Yeah, I mean it's amazing. It doesn't have a lot of the capabilities of this and that, but you can build on top of it and build anything you want. That's what's crazy about what Facebook decided to do is open source. Super

Brett:

Cool man. Early days. We are early innings, just like when we were first starting to use the internet, that's where we're right now. So Russ, it's people listen to this and they're like, all right, I got tap more into Russell's brain and I need to look at some of these resources. Where can people connect with you? I think you've got a mini course that people can kind of take. Where should they go

Russ:

From here? Yeah, well if you want to get your feet, well ai, you go to join the click.ai/jumpstart. That's a 60 minute course where we've got eight videos and six exercises and by the end of it you'll be like, I feel pretty good about this. So you can go to join the click.ai/jumpstart and take that little mini course and it moves fast and you'll just open up a free version of chat BT and do the exercises. And you'll feel a lot, you'll feel like, okay, I'm ready. Right? I feel like I understand this now, but I love LinkedIn. I talk to you on LinkedIn when I hit you up. So if you see me over on LinkedIn, I love connecting with people over there. So it's just Russ Henneberry over on LinkedIn. But yeah, check me out@jointheclick.ai and I'd love to, that's

Brett:

Awesome, man.

Russ:

Geek out on this stuff.

Brett:

Let's geek out a little bit. Join the click.ai I'm fired up to. I just want to take my game to the next level with ai. And so Russ is going to be my guide, so check out his resources, check him out on LinkedIn. I'll of course link to everything in the show notes, but you can find it as well just by looking. And so Russ Hena Barry, ladies and gentlemen, Russ, thanks so much, man. Ton of fun.

Russ:

It's a lot of fun. Thanks Brett.

Brett:

Absolutely. And as always, thank you for tuning in. I'd love to hear from you, what would you like to hear more of on the show? And as always, would've loved that review on iTunes if you've not done. And with that, until next time, thank you for listening.

‍

‍