How AI is Changing E-commerce Drastically

Clement Wan.m4a [ 00:00:00 ]Hi, my name is Clement. We're at the BDSS Market Masters Think Tank. And my name is Clement, and I have an e-commerce brand and a number of other projects. But what I wanted to talk to you about today was how the world has changed in the last two weeks. Basically, one of the things I think, we're all creators, we're all entrepreneurs, and we all have great ideas. And I think for many of us, we like to ideate. And one of our limitations in the past has been our ability to find good and cheap enough coders to build what we want. And in these last two weeks, that's changed because we can do it now ourselves. So everybody kind of knows OpenAI's ChatGPT. Yesterday, or the day before, they just released Strawberry, or O1. [ 00:01:01 ] And basically, it's a big level up on the reasoning of the AI LLMs. And what that basically means now, from what I understand, from what I've seen, is that you now have a model that has the IQ of close to 100. And that compares to some of the previous models, like Claude Sonnet was, and is, frankly, I haven't played around with O1 yet, but Claude Sonnet is one which apparently is is one where I do most of my work. But again, these things, what's what's fascinating about all these models is not only do new ones continually get released, but you have a lot more competitors will probably see additional open source competitors of the same or approaching the same functionality within the next 6 to 12 months. [ 00:02:09 ] Of course, in the meantime, you can use these and what's amazing about each of these tools is how cheap they are. ChatGPT comes at like, depending on what model you get, or if you use the API, you're spending about $20,000 on that. $20, $30 a month. If that, of course, my previous, well, current favorite, since I haven't played around with ChatGPT, is Claude and Claude Sonnet 3.5. And what these models do is, especially you can use Claude Sonnet or OpenAI's general model to start telling it what you want to build. And as somebody who wants to ideate and who has products that they want to sell, let's say you want to be able to distribute coupon codes, or if you want to distribute coupon codes in a specific way after somebody does a certain function, if you want to create QR codes that are tied to specific coupon codes, you can do all these things as long as you can describe it as if you would to, say, another tech person. [ 00:03:34 ] So with Claude or ChatGPT, you would start with what you want to build. And then the next step is you would port it into something like Replit, where you dump in your instructions. Again, Replit is about $20 a month. And what it does, is it'll build the front end and the back end of the software in addition to using the instructions you've given it. Now, most software, and one of the big challenges of people who kind of code a little, is that you need more than just building the actual functionality, something that works a little bit, because you have all this foundational work. You need the logins, you need the back end, you need the front end. But what's amazing about Replit is within a few minutes, you can give it a one-line prompt, and you can create an app. [ 00:04:33 ] And one of the downsides of it right now is that you can't port in existing projects. So what you do is you move on to something like Cursor. So you bring it into Cursor, again, around $20 a month. And here, you just pair with it. And then you can even, for both Cursor and Replit, you can use audio to just go back and forth to iterate and build a landing page functionality for a landing page. Imagine if you sold umbrellas. Imagine if you had an API for the weather. And what you could do is, whenever it rains in a certain zip code, then you could offer certain codes or be weather-specific to a specific product or what have you. [ 00:05:30 ] But again, the imagination is not limited by what a coder can do, but what you can describe in very tangible terms to your developer or AI for $20, I guess, in totality, around $60 to $80 a month, which did not exist two weeks ago. Now, the last piece of software is, like, you want to have a nice little skin to your software. And that's where we have Vercel. Vercel creates React components. Again, you just talk to it, and you tell it, let's have it generate a sticky header. Oops, okay. So you actually have a sticky header. You actually have to sign in. But again, Vercel is $20 a month. And you iterate around, oh, I want the box to be a certain size. I want the colors to be a different way. [ 00:06:38 ] I want a certain background. And it all does this for you, again, for $20 a month. I have a theory in general that there are no hacks in, especially in a space like e-commerce, that last for all that long. The e-commerce space, at least as I've known it on Amazon since I came into it in 2000, I guess, 15 or 16, has changed dramatically. But what's changed is everything runs a lot faster and has become a lot more efficient. So what we can do now with these, what we're calling agents, to build for us is that we are able to now not rely on the software. We can now rely on somewhat unreliable people sometimes. And again, we're only constrained by our willingness to describe in very tangible terms and understand what we want. And we can go back and forth with any of these agents to build exactly what we want. To turn to a real-world example, I've logged into my cloud plan and I want to build a landing page for a garlic press. [ 00:07:59 ] And it'll start generating maybe a first draft. So now we have a landing page. But you sell on Amazon. So you're trying to generate demand on Amazon. So maybe you aren't actually trying to sell this garlic press on your page. But what you want to do is you want to create the ability to drive sales to Amazon with, say, a coupon code. Well, just tell it. As a D2C expert, how would I improve the click-through rate on my page for collecting a coupon code on Amazon? Press enter. How do I get rid of this? [ 00:08:48 ] And maybe the next step. How do I make it even more persuasive or compelling? You can get the idea of how you could go on and on with just talking as if you were talking to your best tech friend. And if you don't have a tech friend, then now you do. And you keep on iterating to the point that you have something that's workable. And then if you really want it to be efficient, because this code, the goal of these tools isn't necessarily to have a production-ready app. And but we're getting there. The point is to make sure that you have something that can work and that's functional and that you can use somebody else if you actually want to sell the app or if you want to actually deploy it in a high-volume environment, that you can do that. Again, my name is Clement Wan. Thank you for watching. And if you want to get in touch, you can find me on LinkedIn. And I hope we all build great things.