C (2:32)

Hey, Steve. Hey. Thank you so much for having me. I'm so excited. Yeah, this is going to be fun. It's a topic that is really close to my heart and you're dabbling in it, so I'm excited to talk about this stuff.

B (2:45)

Yeah. Actually, before we dive into the AI part, I know that you've helped thousands of Amazon sellers to scale at PPC Ninja. In the current, current landscape on Amazon, with the rising fees, tariffs and competition, where do you think the biggest opportunities are right now for sellers to stand out?

C (3:06)

Yeah, so I think definitely the tariffs have been a big blow to a lot of businesses and they're still trying to figure out how they're going to, you know, deal with it, whether they're going to change their pricing structure or they're going to make fundamental changes to their supply chain. And of course those things take a lot, a lot of time. But I think for now the sellers that we're dealing with are extremely conservative about their ad spend and they're just being very, very cautious. That's the best I can say. It's like so cautious that they just, they're not sure. Should they change strategies, should they keep holding onto their prices or should they experiment with any kind of pricing tools, et cetera, to see what others are doing? There's a few different approach, but for the most part people are just staying where they are. That's what I've observed. No, no one's drastically changed their advertising strategy. You know. Yeah.

B (4:09)

And their prices also, or I am.

C (4:12)

Surprised, like literally less than 5% of our, you know, agency clients have changed prices. Everybody else is holding on to their prices. Yeah.

B (4:22)

Okay. Okay. And in terms of AI, what are some of the opportunities that you've been implementing with some of your clients?

C (4:32)

You know, AI gives you speed and it gives you efficiency because you can get, you know, creatives created like very, in very short periods of time and very cost effectively. So we actually help our clients with AI listing optimization. So the first part of that is Rufus optimization for if you're selling on Amazon, then you want to get your listings ready for AI. So the AI era of conversational e commerce. And so your listings need to be written in a certain way that these LLMs can actually understand and recommend your products versus, you know, some other product that's just been optimized for SEO. So SEO has, is slowly being replaced by, you know, this generative AI enabled or AI ready listing. And it's got to be structured in that way. So that's one. The second is video. Video assets, you know, you can now create so many video ass in the shortest period of time without needing any kind of fancy equipment or models or even the product. You can just superimpose layer and just very quickly and rapidly test a lot of ideas, do a B testing, and then it's a much better way to mass expansion without spending a bunch of money. And so we help with that. And I think we've seen a lot of positive response on the AI offerings that we have right now. And hopefully it's going to grow, it's going to go get better and better.

B (6:09)

Over time for the audience members. What is involved with Rufus optimization specifically?

C (6:18)

Rufus? Yeah. So Rufus currently is kind of in a primitive state, but it's only going to get better from here on. Rufus is currently giving you prompts and answers. So if you, if anyone discovers the Rufus button, which is a little hard to find, but it's kind of on the left bar and the top, top left side, if you click on that Rufus button, you can start having a conversation about the product detail page. You can ask questions of Rufus about that page. Because there's just so much like clutter, I would say a lot of, a lot of it is useful information, but a lot of it is like other sponsored ads and, and then there's reviews and Q and A and things like that. So if you actually want to see, you know, if, if you want to know if this product is actually good for you, if it meets your needs or not, you can just ask some basic questions. Is it good for my skin? I have dry skin, for example. Like without having to read the whole listing, you can have a shortcut way of knowing everything about the product that you care about in a few words. And so Rufus currently is doing some sort of pre curated list of prompts. I think I've seen 15 or so. And if you keep clicking or keep asking, it gives you more and more and more like five at a time. And then up to 15 questions and answers that are kind of already kind of pre curated for you to explore. Now what we do with Those Q&As is that we extract them. And the way to extract them is to basically use a tool. And we use a tool that we created and that extracts all the questions and all the answers and gives it to you in a easy to use format. Now what we do, my team does, is we take that and put it into our own custom GPTs and we generate content with it and sometimes we generate images that speak to certain negative aspects that Rufus is highlighting. And you don't want that. You really don't want Rufus to say something like, customers think that this product will fail on these points. You know, and so you proactively, you know, you want to proactively address those very prominently so that even if someone were to discover those negative questions and answers, they would know that, oh, this is probably because the, whoever bought it made a mistake on their end, it's a misunderstanding or whatever. So we like to kind of improve listings in a way that proactively address anything negative being flagged by Rufus. And we also like to provide comparisons with other products, similar products, because one of the questions that you can ask Rufus at this point is how does this product compare with that other product from that other brand? And if you have something ready in your listing, in your images, or anywhere in the A content, it will be picked up by Rufus. So, for example, the kind of most direct way of addressing those sorts of questions is to have an us versus them comparison image on your listing. And it can be kind of further back in the gallery because you don't want to clutter people with too much like text heavy content, like right in the, in the, the head of the, the gallery. But it can be something that you write for Rufus. And then we actually do another thing, which is we create an FAQ image. So the image itself has questions and answers and Rufus is picking up all of that. It's getting everything from your listing, including the text on your images and the, the meaning of the image itself, like the image is understood by that LLM. So those are some of the main things we do with Rufus. And then there's a whole bunch of other kind of minor and more sophisticated steps that we need in order to get it completely ready for Rufus.

B (10:21)

Do you find that the images work better or just spelling out like the FAQs in the text or do you do both?

C (10:29)

So we're limited on text space, as we know. Right. But the images tend to be more kind of forgiving of like dense, heavy text on, you know, that amount of space that we have. Whereas, yeah, a bullet point can only be, you know, certain number of characters, 500 or whatever, depending on your category. And then of course a can be used. But you want to make it pretty. Right. A lot of people focus on the aesthetics versus the detail. And yes, you need both. So it's kind of a balancing act of like, where do we put what in order for it to look pretty as well as, you know, very, very Education as well as ready for an LLM to act on.

B (11:12)

You know, what's funny is in D2C land, I've been optimizing for LLMs also. And just one example of something I did recently was on a category page, I said, we've shipped over 1 million wedding handkerchiefs, right? And then now if you ask ChatGPT, who's the leader in this area, it'll say our company like, things like that right now. And it's just Amazon. You're more restricted obviously, in what you can say, but it seems like you've gotten around that with, with images and, and answering all these questions ahead of time because you can reframe the narrative. I imagine Rufus gets a lot of these opinions from like the reviews that have been left. Right.

C (11:51)

You know, even one review, one negative review can influence Rufus, you know, to say something stupid. Like, literally one review out of thousands can be, can be the make or break. So, yeah, you've got to really pay attention to the negative ones.

B (12:06)

But you can reframe the narrative, it seems right by.

C (12:10)

Yeah, with images. Yes. Yeah. So just an example, you know, if, let's say you have a product that requires some amount of care, like for example, a necklace or a pendant, something that falls off easily, you know, Rufus might say, oh, the, you know, customers have complained that the pendant falls off or something. And so if that's the complaint, the number one complaint, then you can basically put it on them. You can put it on the shopper to take care of the product, and you basically add a care instruction or a care page that says don't take these delicate pieces to the bathroom, for example, or don't use it in the shower or keep it away from shampoo or things like that so that they know that they need to do something to, to make it last longer. It's like it's a treasure that you treat with, with care.

B (13:07)

And then Rufus will include that in the answer, hopefully. Right.

C (13:11)

You know, it, it will eventually. And the thing is, one, one point here, which is different from SEO, the traditional SEO, traditional SEO kind of has a shorter window to kind of change, whereas Rufus takes a little bit longer because it has to be proven over time. Like, customers have actually need to change their opinion over time for Rufus to, to give you different answers. So we haven't seen those answers change overnight, but over time, yes, they're going to be different. It's going to start absorbing some of the content and providing those as responses.

B (13:49)

And I know that tool that you said that scrapes like The Rufus answers. I remember, like at least at seller summit you give that tool out for free, right?

C (13:59)

Yes, yes I did. Yeah, yeah, it's a very simple tool and all it does is it just kind of simulates clicks and makes the questions. Basically it simulates a click on a question inside the Rufus panel. It opens all the questions up one by one by one and then it goes and reads those questions and answers and gives it to you as a copyable clipboard kind of text. And so that's just a simple, simple JavaScript code that just sits in the browser and just reads whatever you're doing on the browser itself.

B (14:38)

So I remember we just had a conversation before I hit record. You said you don't consider yourself a programmer. Is that something that you created with AI or.

C (14:50)

Yes, I created that with AI. I am not a programmer and you know, I mean I have kind a basic, basic fundamentals of programming nailed because I went to, you know, I did my engineering a long time ago. But it's not like I use programming on a day to day basis. If you were to say, hey, write a, write a piece of code in JavaScript, I'd be like, I don't even know where to start. Like I do not know how to, how to start. But because I have that high level understanding of, you know, this is how programming works, I can actually ask and any LLM, especially claude, which is probably my favorite for any kind of low code kind of task, so I can ask CLAUDE to generate a browser bookmarklet and then it'll do it, or I can say create an HTML page that I can download and use interactively. So I kind of know a little bit of that in order to give it instructions. But that's all you need to know. You just need to know what's possible and then the LLMs do the rest for you.

B (15:56)

So in terms of the people listening, and we're assuming that most people listening are not technical at all, what are some of your favorite AI tools right now from a non coder perspective? Just to, just to do what you want, I guess, just without any code really like what are you using it for? What are your favorite tools right this second?

C (16:18)

Yeah, so of course chatgpt for any kind of brainstorming, you know, any kind of high level planning around certain, you know, tools and things like that. I can, I can do that high level programming or planning with ChatGPT then the actual programming I do with Claude because Claude understands code better than any other, I think any other LLM at this point.

B (16:45)

It seems to be the best right now.

C (16:46)

Yeah, it seems to be the best right now. Yeah. The only trouble with Claude would be that it runs out of. I run out of credits very quickly on. On Cloud. Even though I have the paid plan, I keep hitting those limits. I'm not ready to upgrade to the 90 plan yet. So I'm kind of trying to make do with, you know, my ChatGPT subscription. And so I do it in two parts like the planning here and then the execution in. In Claude. Now I also do like perplexity. If there's a research project that requires any kind of like live web based search, perplexity seems to shine there. And then Gemini is good for. I think Gemini is. We use a lot of Gemini for our Google workspace. It's kind of integrated into Gmail and BigQuery and Looker Studio and things like that. So those are, that's where I use Gemini the most for those sorts of tasks. I haven't found Gemini. I don't know. There's something about Gemini that makes me not want to use it.

B (17:56)

Okay. Yeah, I don't use it either. I'm just curious. So.

C (18:00)

Yeah, you don't. Right. Okay, so I'll give you an example. It's like very, a very stupid basic example. When I am chatting with Chat GPT, I don't care about spellings. Like I don't care like if I make a mistake I'm not going to go back, back, back to correct it because I know it is so good with understanding what I meant. Like even if it's like completely off, it really literally looks like a cat walked on over my keyboard. I don't care. I just, my fingers are flying across the keyboard. I am like constantly in AI line. Like I have multiple tabs open and multiple things happening. While one tab is working on one operation, I open another and I do another. So I don't have time. Right. ChatGPT never makes me feel stupid about my spellings. I'm like, okay, this is great. Like I just understand what I'm saying. Just keep going. You know, my shorthand, my, my bad spelling things. Go ahead. Gemini rubs it in the. Like literally rubs me the wrong way by showing it to me that, oh, you said this. You probably meant this. Yeah, right. I spelled Amazon AM a X O N. Don't remind me of all my spellings. So I somehow, yeah, somehow Gemini rubbed me the wrong way by just reminding me of my spelling mistakes and the way I had said something and you probably meant this and. Yes, yes, yes. Right. And then Claude tends to have a different personality. Claude seems to be like this very. Gosh, these guys have personalities. Anyway, so Claude has a personality. Claude tends to be this quiet, super smart, intelligent being that will only give you as much as you need and nothing more. You know, it's not super chatty or anything like chat. GPT is chatty. Like when you do the conversation mode, it will talk to you like a human. It'll put the earth and the ahs and ums and everything. Whereas Claude doesn't do that. Claude will stick to the work and when it's done, it'll give you helpful suggestions. Do you want this, this or this? And then you can pick from one of the options that it suggested. So the conversational naturally flows to the next step with Claude. So somehow for coding, I don't know that that frame of mind where you're really thinking about how you want to proceed. You don't want chatty, you want, you want the thing to understand what you want and then without distracting you too much, give you the next couple of of, you know, best next options for you to keep.

B (20:27)

You don't like small talk. I get it, I get it.

C (20:33)

Yeah. Maybe.

B (20:35)

What I did want to ask you though is that what are some of like your most impactful automations that you're using or you've developed that just make your life so much easier?

C (20:44)

Oh my God, there's actually so many. But let me share a secret, which is that even prior to getting into AI, I was very, very automation focused. So I to use app script. So Google Apps script for all kinds of things. And chatgpt just, or OpenAI just came along and made my life easier. So it enhanced my kind of style of working by shortening the duration of, you know, creating these projects. It just takes like sometimes just an hour or two to put together an app. I'll give you an example. This is like really, really simple. But it saved me, I guess it saved my team a little bit of time. But over time, all this kind of adds up. Right? So we have this. What is it? Gosh, Walmart. Okay, so Walmart, we have a login for Walmart. And every time my team tries to log in with our email that the client has given us and given permissions to, it generates an otp, right? It says, okay, now you gotta go, now you gotta get the OTP and then put that in and then you can proceed and login. That OTP comes to me because that email is an alias of my email right now. My team sitting in somewhere in Philippines or India or wherever, they're like, oh, can you send us the otp? Because we want to log in. So I was like, oh my God, I want to fix this. So I figured out a way where as soon as the OTP is emailed to me, right, it comes into my Gmail, right, And I have this, you know, this Google app, Google Apps Email watcher. I have lots of email watchers, but this particular watcher is only watching for an email that comes from help Walmart.com or whatever that link is that addresses and then it'll scan the email and find the otp. Now it's got to do something with that. It's got to send it to, you know, the group. I want it to be sent to all the people in the, in the. Because I don't know which of my team members asked for the otp, I want to send it to them all at once in a group. So the best place for me to send that information is in Google Spaces. So spaces is this very cool. It's almost like Slack, but in the Google ecosystem. So you have spaces and we have like 45, 50 spaces right now, one for each client, plus there's other small ones that we have for internal purposes. So these spaces, I created a space just for Walmart and every time my email hits an email like that, that OTP is sent to that space and everybody who's on that space just picks it up. Whoever wants it can, can pick it up from there. This took me, I would say like 30 minutes to an hour to create the app script that does all of this. And I got stuck along the way because I needed something called a webhook which I had to generate from within the space. And then I had to give that webhook to the Apps script so that it would kind of give, give the final output to a particular place. And that was my spaces. So gosh, this, this project has been saving me time. It's been mostly saving my team frustration because I'm not awake when they're awake and stuff like that. So I would say such automations, there's like hundreds of them that we've created to save time. And I generally follow this rule of if something is happening more than three times, if you're having to do a task more than three, consider automating it. And we find these opportunities everywhere in our operations every single day, like so many of them. So I think the combination of existing automation tools like Apps Script and the power of AI to generate those JavaScript code pieces that I can just quickly plug and play is helping me so much right now. I'm saving so much time. That's my AI world, you know, that overlap of like automation and AI right now.

B (24:58)

What about people who are just afraid to touch any JavaScript or whatever? I know, like I haven't tried this yet, but I know Claude's integrated into Gmail right now for example. Right. And have you tried that?

C (25:10)

Yes.

B (25:11)

And then you can go and you can.

C (25:12)

Yeah, yeah, I, I tried the. What is it? The, the calendar connect. So it tells me when my. Yeah, it basically reads my calendar and it tells me, you know, when my meetings are or things like that. But I haven't really needed to go there because I have my own system within the Google ecosystem. I did test it out though and yeah, I think that might be a good alternative for someone who is afraid of programming in general or touching JavaScript. It's always going to be this question of build versus buy. Should you buy like a ready made, some ready made solution to your problem or should you plow through the details and have more control over it and just like build it yourself or you have this middle ground of like you want to build it but you want to do it in a low code, no code sort of way where you don't need to understand the coding piece. So yes, I'm sure there's so many different ways you can go about it that, that work for you.

B (26:14)

So it seems like you opt for the code method in general. Is that, is that accurate then or.

C (26:19)

Yeah, because of the flexibility and the control and you know, the fact that it's free. I'm already paying for the Google Workspace system and I'm familiar with Similarity is one of the most important things actually. If you're used to doing things a certain way, if it's just an extension of that, it's always easier than switching to something totally new, which gives you maybe a little bit more. But I'm actually fine with this. Comfortable at this point.

B (26:44)

Let me rephrase the question then. Just because I think the audience is probably afraid of code. I'm just making a guess or assumption here.

C (26:49)

Yes.

B (26:50)

What are some of the most impactful automations that you use that didn't involve any programming? We can talk about your creatives, any. Anything that, that you do.

C (27:02)

Sure. Yeah, no, I, I agree. So the creatives, definitely. I have like this whole, you know, creative stack that, that we use internally. Right. There's lots and lots of different apps that we use for different things. But at the end of the day, we compile everything in Canva, which is our assembly platform. So all the pieces, like the. Let's say you have like a foreground, a background and text. We basically look at things as layers, and then we put everything together in Canva and give it some basic transitions and animations. You can. You can do that in. In that final, you know. Do you do everything in Canva then?

B (27:46)

Because Canva has.

C (27:47)

At the end. Oh, just at the end, yeah, because I think so. Canva has a bunch of AI tools. Yes. We use those AI tools where we need to, for example, text anywhere. It's an AI tool within Canva. There's actually 200 or 300 AI tools within the Canva apps set that we use from time to time. We also use Canva's own kind of native apps like Magic Expand, Magic Grab, Magic Erase, things like that that can. Can help to extract foreground from background or it can help to fill in the space. Like if you have a small image and you want to actually grow the background so that you can position it correctly, you can actually say Magic Expand and it fills the whole space. So we use a bunch of those different tools in order to get the final look that we're going after. But that's. That's on the creative side, right. I think your question was also more about the low code, no code, programming side. So I've actually had success with make make.com quite a bit of it. So to give you an example, and I think I might have shared this at your summit as well, but let's say Facebook posting, right? You start with an image that you upload into a Google Drive folder. Let's say your graphics designer just finished their image, they've uploaded it to a designated drive folder. You want that to be picked up automatically, and then you want to send that to ChatGPT via an API connection and that basically ChatGPT will read and understand that image and generate a nice Facebook caption in text with emojis and hashtags and all that good stuff. And it will also post it to Facebook simultaneously. You can post it to other without.

B (29:42)

Your review or up to you.

C (29:45)

Like, if. If you feel like you want to include a review cycle, then you can have a human in the loop and break the. The automation there, or you can take it all the way to Facebook and maybe correct it later after the fact. For the most part, it's been, you know, working fine with the direct as well.

B (30:01)

Interesting. Okay. For a brand or For. For. On behalf of you.

C (30:09)

Yeah. So this is basically for my own brand. Like, I. I have a jewelry. I. Yeah, I don't operate it much these days. Like, it's not an active thing, but I use it for, like, demonstration purposes as well as for my own, you know, testing and things like that. And so that basically posts it to my live audience that. That actually views my page and stuff like that. And it's got almost perfect captions, like literally it doesn't, you know, know, default from my prompt. The prompt is most important. Right. Because if you put, if you put enough clauses in your prompt, you should get a fairly reliable response.

B (30:51)

Okay. You know, like I mentioned before we started recording, I've been working on Steve Bot and it's read in everything I've ever done and it gives replies and it sometimes times is like off the wall. I mean, it's. The. The answer is correct, but the way it answers it is like all over the place, despite my best efforts.

C (31:14)

Right. Yeah, yeah. Hallucination, you know, losing context. All of that is real. I struggle with it too. Like, especially when you've. Let's say in. In OpenAI, you've listed some custom instructions and you're expecting it to follow all of those. It often deviates, you know, from those especially like EM dash, like those dashes that get added to center. Oh my God, that is so annoying because it's a dead giveaway that this post was written.

B (31:46)

Yeah, right, right, right.

C (31:47)

Yeah, right. It's like, oh my God, you didn't write that. So, you know, so it's funny because I actually created a shorthand to speak to my LLM and every time I do a post or I ask ChatGPT to. To write something for me, I. I just say that one code word. It's an instruction to it to go back and clean it up, remove the dashes and a few other things. So it's like shorthand between me and the LLM because it does not remember, it forgets.

B (32:20)

So this is a shorthand that you've created in a custom GPT, I guess that you've.

C (32:25)

No, no, no, yeah. Just trained it over time. So I have it in the custom instructions in my user preferences sett, stuff like that. I have mentioned that this is, you know, I don't like the EM dashes and stuff like that, but then I also have it in memory, which is another area that you can keep stuff. Right. So there's those two. The combination of those two pretty much guarantees that if I say that word, it's going to do it. And it's basically gobbledygook. It's not any word that makes sense in the English dictionary because it's a made up word that only I know and I understand that when I type it. It knows to do that chat.

B (33:08)

So it's like shorthand for those of you guys listening. Actually what I've discovered is a lot of people don't know about all these extra chatgpt features. So memory, in case you guys are listening, is where it's basically all it knows about you personally. And you could add memories, you can delete memories. You. I think it's under settings and it's just something that you should look at, you know, to see what it thinks about of you. Because it has a lot all these defaults there.

C (33:37)

Yeah, yeah, yeah. You know Steve, I. In my, in the memory I also wrote, I've also given it a few people and their styles of writing as reference like I've said, because I don't like to write, you know, my style is not very like hypey or like clickbaity or I don't like that style of writing. I tend to be more like Set Gordon for example. So I've actually put Set Gordon in one of my instructions, like right in the stable and what is it? Stable and calm style of said code. Because I don't want to, you know, ever be kind of too overhypey and stuff. So yeah, you can put a bunch of stuff like that in your custom instructions and in your memory. So how do you trigger memory? The easiest way is to say remember this. That's the, the code word that gets things straight into memory and then it'll give you a little confirmation that says memory updated like right away. And if you see that you will be able to check it and it can be, you know, you can't edit a memory, unfortunately I tried to edit a memory. The only way is to delete it and start over. You can delete memories?

B (34:54)

Yeah, yeah, I've deleted plenty because like every it.

A (34:58)

I don't even know exactly how it.

B (34:59)

Ends up in there sometimes, to be honest with you.

C (35:02)

Sometimes, right. Yeah. When it's, when it thinks that you're going to use this again, then it just puts it there, you know, like if you're working on this big project and you maybe try to describe it in a lot of detail, then it assumes that this is something that's useful to you and so it's going to add it to the memory. But yeah, you can delete it do you use any.

B (35:21)

So it sounds like you're mainly in Claude and ChatGPT and then use. It sounds like you use Claude more sparingly because of the credit usage. You know, I've never ever run out in ChatGPT ever. But Claude, it happens all the time.

C (35:37)

All the time. All the time. Yeah, yeah, absolutely. Yeah, I totally agree. I don't know how that.

B (35:43)

Do you use any of the other ones or like Grok or, you know.

C (35:48)

No, I don't use Grok. Gemini I know has plenty of memory space. Like they give you I don't know how many tokens. Like I've lost track of the. A million tokens or something.

B (36:01)

I've never run out.

C (36:02)

Yeah, yeah. Gemini gives you insane amount of tokens. And I think I like to use Gemini for transcribing videos, which I don't think any other tool is doing right now. So if I, if I just upload a video, it will transcribe it for me, like every single line.

B (36:22)

Okay, yeah. So the way I do it. Well, I do it two ways. One, Riverside offers free transcription. You just upload it and it'll transcribe it for you. The other cheap way I do it, since I had to do this with like 500 videos, is OpenAI has open sourced their whisper model and so I just run it on my own computer with. As long as you have a good video card, it can go through all that because I'm, I'm cheap, I, I guess. But Gemini is free. I haven't tried. I didn't know about that. Actually.

C (36:51)

It's free. Yeah, yeah, yeah, it's free. Yeah. And you can actually upload like pretty long videos. Like I've uploaded an hour long video and it does a very good job of transcribing it. And now when it comes to YouTube videos, I don't bother. I don't download and upload and stuff like that. There's another tool that I use with YouTube videos. Gosh, I forgot the name of it. It's installed and it's. I forget the name. But anyway, that tool, as soon as you open a YouTube video on the right side, there's like a browser widget. You can just click on that and download the, the transcription.

B (37:28)

So chat, GPT, YouTube transcription or something like that. I use that too. Yeah, well, there's different ones. I'm sure that's the one I use.

C (37:36)

There's different ones? Yeah, there's a few different ones. Yeah. Yeah. So, yeah, it's fun.

B (37:43)

One thing I also wanted to ask you is. I know at Seller Summit you had some of these really cute videos that you created for brands, and I wanted to know what you're doing to create those little videos because they were all AI generated. Are you using Cling or are you using like Google's VO3 or anything like that? What. What is your method?

C (38:07)

Yeah, so initially I was using Sora because it. It's unlimited as well. Yeah, it's like kind of being a little realistic about the cost, cost of all of this. Sora does a pretty decent job. It's no longer the. The top model out there. There's so many much better ones. And then there's a bunch of services like Open Art that give you access to other models, like including some of the Chinese models, including cling. Cling's model is called Colors, spelled with a K. I think that's Colors two or something like that. And then also gives access to VO3 through that. So there's different ways to access all of these models without paying for them individually, you know, and because costs add up when it comes to AI creators.

B (39:02)

That you're just talking about, that allows you to use all of them.

C (39:05)

Yeah, it's called Open Art. I can double check Open Art quick. Yeah, it's called Open Art.

B (39:13)

So does it work? You pay for that subscription, you get access to all of them?

C (39:19)

Yeah, in fact, they have a few credits for free. So, yeah, it is Open Art. And you can. There's. Yeah, there's a bit. A huge library and then you can pick models and then you can also train your own models and stuff like that. So it's a bunch of like, flux. You have Flux, you have gosh, VO3. There's so many. And you can immediately get access to all of them, I think, including Minimax, which is China's. One of China's leading video models. And now I think Mid Journey has also come out with their own video creation. It was always just.

B (39:57)

I haven't tried it yet, actually. Yeah, yeah.

C (40:00)

I mean, either you need to pay for it. Like Mid Journey will not let you in without paying for it. Like they start you on the paid plan. So, you know, I've been debating whether.

B (40:10)

To cancel my Mid Journey lately because I've feel like chatgpt for what I use it for is actually quite good for the text and everything. That's the most important part to me. The text, the text. What is the catch with that? Open Art or that tool? Because I imagine you don't. You get limited credits for everything. Right. If. If you're paying.

C (40:31)

Okay, so yeah, I'm looking at it here. So it's $14 to get started for 4,000 credits per month, which is almost 40 videos per month. A huge number of videos. Like, even if you were to discard like 90% of them, it's still pretty good, right? And mostly, I mean, this is what happens. Like nine out of ten videos we throw away because they're like garbage. And the thing is, the prompt is something that you evolve as you're, you know, evolving the concept, like with, with the tool and you may not get it right. You know, it might end up in a particular direction. You're like, no, this is not what I, what I want. And then you end up discarding that. So often I find myself using multiple tools just to get to the final prompt that I will then use in the final good tool that I actually rely on and I think will do a good job. Like cling, for example, or VO3. Yeah.

B (41:30)

I watched this presentation of a guy who made like a five minute video on VO3, like piecing together different scenes and everything. And he showed off his prompts that he used to generate it. And it was literally like a book.

C (41:46)

Yes, I believe that. I believe that.

B (41:51)

I mean, it's almost faster to film the video than like, look at all the instructions that he gave. Well, maybe not, but I mean, to me it was like, that is overwhelming. Which just shows to do it right, you have to be explicit about so many little details.

C (42:03)

Details, yeah. And the thing is, I bet he didn't write that prompt because you use AI to generate the prompt, to generate a more detailed prompt, which then goes and generates a more detailed prompt. So it's like multi step and then also multimodal, because you will from, from time to time insert reference, images and reference, you know, any, any details that, that can be integrated into the prompt itself, like angles, lighting, style, this and that. So there's like whole frameworks around, you know, AI generated video that, that touch on all these aspects so that you get a high quality output.

B (42:45)

I did want to ask you some of these questions also. Switching gears a little bit. I'm worried about all this. Like, do you feel like AI is, is replacing, I don't know, replacing a lot of these services? Like, do you believe that a small brand now can do a lot of things themselves with all these tools and basically displace a lot of players in the entire industry?

C (43:07)

Yeah. So my thought on that is that AI is first of all going to take away a few jobs, but it's Also going to add back another few jobs because now you're doing more things with it, you're doing different things with it. So it's not like you'll suddenly lose all your people. Like people who are worried that their jobs will be gone, the nature of the job will shift. You'll probably have to use AI intern to kind of do your job better and then more. But I think what AI is unlocking is stuff that was never possible before and I think there was inefficiencies in the way we operated. Whereas now it's going to be highly efficient. You're going to be using the people for the things that people are good at and automation for the things that automation is good at. Whereas previously it was mixed, mixed up. Like people were doing stuff that should never be people jobs, like downloading files from a particular place, like those should not be people jobs at all. Those should be just streamed in or whatever. So yeah, I think as to the question of whether agencies, agencies will be wiped out, I don't think so at this point because you still need to give it business direction, you still need to give it context. You still need human oversight. Like earlier you were saying, are you sure that you can just take, you know, the make.com output and just flush it out to Facebook? Perhaps not. You probably need some spot checking mechanism at the end where someone goes in once a day maybe and looks at and maybe cleans up the posts that were posted. Like if you're doing volume work, you will need humans somewhere in the loop. So that's not going to go away. But the nature of their work will definitely shift over time. And anyone who is in the creative space, I would say embrace AI and leverage it to your benefit it rather than be afraid of it or try to oppose it because opposing it is not gonna get, get you anywhere, you know. Yeah, yeah, that's my thought.

B (45:11)

And then last question for you is if you're an E commerce brand right now and you're still kind of on the sidelines of AI, like where should people learn? Like what do you think that people should learn first about AI or what should they start with And.

C (45:27)

Right, yeah. So you know, there's a bunch of newsletters you can follow. I have one, there's a few others that are out there. I mean there's superhuman AI, there's, gosh, there's a bunch that I subscribe to. So get, first of all, get into the, the mindset of I gotta learn this thing because you can't ignore it like that. That would be foolish to, you know, close your eyes and ears and say, okay, I'm fine. Where, where I am. You've gotta learn. Right? That's the first step. So subscribe to a bunch of newsletters. I could share mine if you're interested.

A (45:59)

Please, please share yours.

B (46:00)

I think it's an excellent newsletter. And ritu, you're always so non promotional, you know, you know what I mean?

C (46:09)

Yeah. Always so modest.

B (46:10)

What is the URL of your, of your newsletter first?

C (46:13)

Yeah, yeah. So mine is ppc-nissa.com so you'll have to put a www in front so that it doesn't go to the wrong place. I need to fix that. But www.ppc-n ninja.com and that should get you the, the entire one year's worth of newsletters that I've just, I just wrapped up one year of, of writing. So all of those are there plus all the new ones that come out once every Tuesday. So that's, that's one starting point. Like get into the, the mindset of, you know, being on top of news. You don't have to know everything about everything, but at least you need to know what's possible. Once you know what's possible, then you can hire people who can help you. You first line, you, you could get people as consultants to, to help you out with like just structuring your AI automation, internal AI automation or AI efficiency. I, I've been hired by like two, you know, agencies in fact who, who want to do what PPC NINJA is doing. And that's kind of interesting because they feel the urgency. They, they know that they need to be AI savvy or else they'll, they'll lose their jobs. So I'm helping them. But then the next step from there is to get smart people who will actually do stuff and actually be the prompt engineers or the AI creatives in your organization that will take you forward to the next level. So start with the education, bring in some consultants and then start training or hiring for talent in house. And just keep a note on the fact that the, this talent is not readily available. So if you can find someone who is good, grab them, you know, early.

B (48:03)

Yeah, I think what I found at least with the students in my class is like you have to realize that almost anything is possible. Like if you want it done, like if you're doing something. I think the example you gave was if you're doing something more than three times a day, chances are it can be automated away way and the first time you do it it's probably going to be a little bit more intimidating. Right. You'll make lots of mistakes, you'll get frustrated, but if you can just break through that initial barrier, then literally, because you can use AI to make decisions for you on a lot of fronts which you couldn't do before, I think that's the biggest benefit. And you can automate practically any stupid thing.

C (48:46)

I agree. Practically any stupid thing. I totally agree. It's just a matter of. Matter of priority and also cost. I think that's, that's. Those are the only two things that, that come to mind. But other than that, you can automate anything.

B (48:59)

Yeah, yeah. I mean, I think I'd mentioned this in prior episodes, but I've been working on something where you know how you have to rotate out Facebook ads all the time? Well, you just list what you want in a spreadsheet and automatically generates, you know, the Facebook ad for you. Or, or we probably both get tons of emails. You can have ChatGPT go through like last week's worth of emails, for example, make its determination on whether you should respond or not based on whether it's promotional or whatnot, and automatically create a draft. Like, those are just some simple examples here.

C (49:35)

Yeah, yeah, yeah, Totally agree. Yeah, love it.

B (49:39)

And ritu. So if people want, you know, PPC and Amazon services, can you just kind of describe what your ideal client looks like?

C (49:50)

Yes. So we work with, you know, clients that have at least a million dollars in sales. So we start there. That's kind of our cutoff. It's not like we don't work at all with smaller sellers, but just, just understanding that, you know, we may not be the right fit at that early stage because we have our service fees and, you know, things like that, and we do, we quality job. And so we do have a slightly higher entry point for that reason. The other criteria we have is the conversion rate on their, on their listings, we generally look for 8% and above. And why is that? Conversion rate is a leading indicator of how efficient your ad performance is going to be. So for example, if you have a high conversion rate, let's say 20%, you only need five clicks to make one sale. It's just the way the math works out. And if you just take an average cost per click of let's say a dollar, that. That's basically you're spending $5 to make one sale. Now if your sale price is, let's say $50, then at $5 to acquire one sale, your ACOS is already decided, it's 10%. Right. So we know from the conversion rate what your efficiency is going to be. Now take another example of offer a horrible conversion rate of let's say 2%. Now you need 50 clicks to make one sale. And at $50, you know, $1 per click, you're spending $50 to earn $50. That's 100% ACOs. We can't help you. So basically you need to fix your conversion rate first. Get it up to a point that's decent. That's why our cutoff is like 8% ish. Now there might be some products that have a very high price point, in which case also your conversion rate could be low. And that's fine. Those cases are fine because then the ratio sale is not that bad. You'll still make a good acos even if your number of clicks is high, you know, so those are the two main ones. And yeah, we don't really have any other blockers to kind of do a free PPC audit. And then if we are a good fit, we can start working with you. We also offer like DSP services for slightly larger brands where you know, there is an element of repeatability. So there's a initial customer acquisition cost with DSP and then you can basically remarket to those people over and over again. And that way you can kind of stretch out your dollar so that it kind of has a longer lifetime value versus, you know, products that only have a one time and it lasts for two years. DSP is not for you. Right. So supplements is good or consumables or anything that needs people to come back for more. More. That's good, a good candidate for dsp. So yeah, those are the two. And then we also do Walmart in case you have a Walmart account and we can manage that as well.

B (52:58)

And I just want to say that I've known reach. I think I've known you for over a decade now, or almost a decade, I want to say.

C (53:04)

Right, almost a decade. Yeah.

B (53:06)

And what I like about you and both Bernie is, you know, you're, you're like a. You say it like how it is. And there's no, like you said earlier, you don't want anything hypey or anything like that because I feel like you're, because you're both engineers. Right? And, and you're like my people like who I relate with personally because I feel like we have similar personalities. So I just wanted to say like, I, I trust you.

C (53:32)

Oh, thank you so much. Means a lot, Steve. Yeah, that means a lot. Yeah. And you know, I trust you because your content is so good. Like I can listen to you for hours. I literally, literally have your voice, you know, it runs through my head sometimes when I remember something because I'm, you know, I've listened to some episode of, you know, something you're maybe talking to someone or just your videos are so good and it's like so logical, so methodical. You just go from point to point. Like if your text flows seamlessly, it's really amazing. So here's our YouTube fan club, guys. Just.

B (54:11)

And oh, and I do want to just plug your newsletter real quick. Like all these automations that we didn't really get a chance to go into depth. Like retube gives out a bunch of these free automations like the one that scrapes Rufus, for example, which is highly useful in helping you figure out your Amazon listing. So as soon as you stop listening to this, go sign up for the newsletter. I sign on and I learn new things every week because I feel like you are in the AI probably 10x more than I am because I feel like you just wake up and it's AI AI AI guy. Whereas I'm a little more casual about it. Like I need some for my store. So I'll, I'll go do something about it. So yeah. But Ritu, thanks a lot for coming on the show. Really appreciate it.

C (54:52)

Oh, thank you so much, Steve, for having me. Appreciate it.

A (54:57)

All right, well, I hope you enjoyed that episode. Make sure you check out ritu's AI newsletter for the latest AI tools. And for more information and resources, go over to mywife quitterjob.com and once again, tickets to Seller Summit 2026 are now on sale over at sellersummit.com if you want to hang out in person in a small intimate setting, develop real relationships with like minded entrepreneurs and learn a ton, then come to my event.

B (55:23)

Go to sellersummit.com and if you're interested.

A (55:25)

In starting your own e commerce store, head on over to mywifequitterjob.com and sign up for my free 6 day mini course. Just type in your email and ascend.

B (55:33)

The course right away. Thanks for listening.