AIでEコマース企業をゼロから立ち上げる方法

Hello everyone, welcome to AI Marketing. Today joining me is Nikhil Vimal and we are going to tell you or show you or he's going to, I'll chime in, how he would start an e-commerce company today using AI. So with that, before we get started, Nikhil, will you give us kind of a brief bio on yourself? Sure. Absolutely. Thanks for having me, Mark. So my name is Nikhil Vimal. I've been in technology, startups, AI, even pre the current AI boom for about 15 years now. I've done a lot of work ranging from startups to, you said it, e-commerce brands to Fortune 5 retail companies and a lot of different other industries as well. I run my own agencies. I've started my own venture-backed companies. And, you know, I've done some angel investing, advising, writing a book. So I'm an author now as well. And, yeah, just got a lot of – been always on the cutting edge of things and always looking for new ways to start some communities, right? So looking for ways to give back and, you know, kind of like the point of this podcast, come in and share my expertise wherever I can and offer value to a new audience.

So looking forward to doing exactly that. Yeah, my listeners know that I like to follow you early adopters, so you've saved me a lot of pain in trying these things out and telling us what works. Yeah. Sometimes I'm an early adopter, especially with AI agents. I go through one or two a week just to see if they're working the way the owners say they work. Hey, I would definitely recommend that. It starts with those little experiments, and eventually, to your point, you become an early adopter if you find the right thing at the right time. So that's awesome. It's just tiring, very tiring. But all right, let's move on. So you and I go into business. We're going to do an AI. We're not going to do an AI company. We're going to do an e-commerce company and we're going to leverage all the new tools that we see out there. So, you know, tell me how you would start. I mean, let's just say we have a great idea, a great product, a great service, something in e-commerce. And so the product's not the problem. It's, you know, how you build the business, fund it and then, you know, execute on it.

That's a great question. Yeah. Yeah. I think it I think it falls into a bucket. It, you know, it's actually very similar to, I guess, even other ways that I would sort of approach other sorts of businesses. But I'll make this one, you know, for the e-commerce focus where, you know, even with a killer product, right? The point is that right now, marketing attention, paid media, I mean, all these things are, we are in the attention economy. Everyone's a creator now, right? And standing out from that requires real in-depth understanding that marketing is no longer just putting money down on an app, right? And what AI is essentially doing, the real, real function and kind of the baseline I'm going to set, right, is that AI is essentially getting an entire marketing, you know, call it marketing fulfillment or marketing or growth engineering or whatever the latest cool terms are. It's agents allow you to build that team without having to actually hire a major team, right? There was a time where e-commerce brands and D2C and CPG brands were all, you know, they need like 10 people to get started, right? A supply chain, a small marketing team, maybe they need to go out in an agency, right? I think what AI agents do is they create a real opportunity to

de-risk and lower the barrier to entry by having, you know, these autonomous, you know, call them teammates, right? Working alongside you. So the first major thing that I'll sort of frame this as is that you should go into the mindset of using AI as an augmentation for human tasks and rapid, gathering of knowledge, leverage, and other opportunities. So again, if you know this product is going to work, well, you got to make sure it's in the right hands. You got to make sure that there's not certain gaps that are missing when you're setting it up. And so that's all sort of frame and start off this conversation, right? Is that like the AI part of this is not just to do an automation. It is to create something that learns from itself and continues to be essentially an assistant in the execution side, not just research is a part of it. I kind of I kind of alluded to that earlier, but then it's what it does with that research is how it tees things up. It's how it looks like competitors and all that other important steps to marketing and then creates a proper cycle that allows you to do what you do best, which is fulfillment and operations and partnerships and even running different kinds of experiments.

While things that you know are working or know that are working for your competitors or other parts of the industry have AI agents doing the work of what would be a marketing agency or a 10-person marketing team or whatever have you, right? So I think with that framing in mind, I break it up into three important buckets for like if we're starting a business, right? If we're starting from scratch, where the first, like I said, is this fundamental research and call it like, there's a couple of ways to call it monitoring. Monitoring is kind of a different part of agentic functioning too, so I might not call it that, but essentially like market analysis, right? Research and market analysis where AI can help you discover so much, right? If you enter your company and ask, you know, what my competitors are, you're already getting, you know, data that would have taken somebody quite a while, quite a lot of Googling to discover, right? But if you have something that's constantly running and keeping an eye on the types of sales, the types of promotions, the types of collaborations, the types of creators, ads. Yep, exactly. Nothing wrong. I mean, hey, you know, like there's nothing wrong with a little bit of imitation because that's the game of competition, right?

And being able to understand and find gaps in that too, right? So you can find, you know, let's say that there's a product that's, it's not exactly the same product, but it's in a similar industry. Like I'll just take like wellness, for example, a lot of different types of products. But you certainly, there's no rule against, you know, looking at how other wellness products are doing certain things to capture a similar audience. That might be, you know, again, might not be even a part of your bundle. It might not even be the same product at all. It's a wearable versus food or something. But we can learn from each other, right? And AI can help you do that without having to reach out to anybody, without having to constantly watch something. And again, without having to have one, two, three people constantly doing analysis. And to that end, I know you're in control here. One thing I would want to insert is I've seen a lot of success where you find competitors' ads that are doing very well, or maybe an agent industry that's doing very well, and you just have AI tweak it a little bit and then recreate it in C-Dance 2.5. And it's very powerful, very powerful. Extremely, extremely. Yeah Higgs field seed dance you know even even with like I use I used mid journey for product images Even honestly even even you know chat to PT has a lot of work you know to get the prompt right so it doesn look like everything else But once you kind of get it right it a pretty good you know model

I think it's sort of that sole 6.1 or whatever they have now, right? It's not bad. It's not bad. You just have to know how to prompt it. Yeah. Go ahead. Thank you. Go ahead. Yeah. Keep going. Oh, no, of course, of course. But no, I 100% agree. And that part of it is like, you know, there were products that were dedicated to this kind of work before that were like enterprise only, right? I mean, you have paid extremely, extremely, super gatekept. And now AI, you know, breaks that whole barrier down, right? So once you get the research part in motion, once you get the analysis, the competitive sort of analysis, the, you know, parallel industry analysis, all that fun stuff, right? The next step is focusing on, you know, like it's using AI in the marketing mix, right? So I'll call this like it's sort of as a form of we'll call it preparation next, like not quite execution, but preparation, right? Where, okay, I have all this material, you have all this analysis, you have all these, you have raw data, right, to work with. Now, how are you going to, to your point, discovering tools that are, you know, AI image generation, video generation, copywriting, right?

All these kind of things have to constantly be, you know, if you're going to be constantly looking at competitors and constantly working to compete, you have to have a sort of system which is constantly cycling, personalizing, and adapting to changes, to competitive moves, to industry wants and discounts and seasonality or whatever. And having AI agents in the mix that can, again, automatically generate a Halloween promotion, right? Automatically generate a Christmas promotion, you know, Boxing Day. You know, they can go hand in hand if you're selling on Prime. You know, having it understand like when those things are. You can have all that, you know, essentially automated these days, right? Or at least to some capacity, have it create something for you. And then the human in the loop is just the approval process, which, you know, I tend to champion because you never want AI to, you know, Yeah, I know you have to have a human in the loop almost for every agent just to make sure. What I say is every human's got to be assigned to an agent so there's some sign-off or responsibility because some of the things that are being put out there, and then, you know, I've covered this before where these AI agents go rogue and nobody's watching what they're doing for two months,

and then it's reached havoc. Yeah. Oh, no, it's exciting what I'm talking about is 100% where, like, I've seen it literally, you know, give cars away for free before, right? And technically, by terms of service, that brand was entitled to give away a car. So it's crazy, right? Don't go car shopping with an AI agent, right? Like, that's like the scariest part. Rather, you know, don't, like, that's not what the dealership slash the car companies want, right? But to that end, to that end, right? Yeah, the next step is, you know, having it run the marketing mix for you. And then that preparation turns into execution, right? So having it generate content calendars for you, having it generate even outreach, right? Having it properly generate, you know, follow-up emails to past customers to, you know, if you work with like creators and people who marketed your product before, user generated content, right? I've seen very intelligent systems that will literally reach out to people who tag your brand before through an Instagram DM, through a WhatsApp message, through whatever channel that they might be most active on based on the information that you're getting and based on how much are coming to your site or what they're signed up for, et cetera, et cetera. And suddenly you have, you know, people who are, they're familiar with different

variations of your product. They're familiar with new product releases and launches. And again, you literally don't have to have a person in the loop, right? And that can even, you know, sort of derive further where, you know, in marketing right now, there's a big trend called, I think you're familiar with this, Mark, like clipping, which is this act of, you know, taking like UGC content and any sort of commercial content that you have and just, you know, generating, you know, God knows how many iterations into little bite-sized pieces and then re-syndicating them on pages in, you know, the industry of the product that you're selling. And, you know, I've seen like things like meme accounts work depending on the product, right? Depending on the product, but, you know, energy is too great on sports compilation things or whatever, right? And so you're looking at something where AI is helping you scale content. Like you might have 20 pieces of content that suddenly become 100, right? And I know for a fact that these are all AI services. Like any agency you sign up for is using AI to do this clipping. I promise you it's usually not a human in the loop. I mean, a human loop approving it and maybe like setting up the template, but it's not going to be someone who's sitting there and, you know, what is it called, iMovie or something or something.

Better software than that. But you know what I mean. No one's actually doing the editing. It's AI doing that. And even brands can even scale that themselves, right? And so I think that's the – Just an agent to that because you're throwing out so much fun stuff here. Just another agent to that would be, you know, last week's Tesla stunt with the flying car. I saw so many great ads and clearly AI generated utilizing, you know, AI to kind of replicate the car ad. But instead of the car on the ramp, they had, you know, everything you could imagine. Even McDonald's chimed in. There were Domino's Pizza. There were smaller brands. And everyone took advantage of that very, very viral event. I haven't seen anything more viral in a long time come from a company. Even stuff like that could be done quickly. with AI tools. I think I saw some sort of shipping company Photoshop one of their semi-trailers into it, right? I'm thinking, man, what a boring industry to make so fun, right?

I remember another fun event that happened this year that I remember vividly everyone took advantage of because of the A, dissemination of information and a ton of AI capabilities as well as when the two people climbed the Empire State Building and waved the flag that whatever sort of a protest move slash extreme stunt or whatever but everyone put their own brand on that flag. And I remember that was getting like, you know, God, like every, every account had their, you know, a moment of truth, you know, biggest viral hit in like 30 seconds of generating an ad that just, you know, Photoshop like text on this, on this image. Right. And that's what that, that's what that first part helps with. Right. I mean, you're able to be on the pulse of so many world events and context and things that are going on. That's almost foolish not to be in the, like if you're not using this Tesla roadster to some marketing capacity, I mean, I wouldn't, I wouldn't mind seeing a burger flying around. You know what I mean? At this point, it's funny. Maybe it's not funny to some audiences, but it's probably hilarious to others. We're all thinking, what does this future look like? Well, you can take advantage of that and be a part of it, right? So I think that all these layers work together,

and all these agents are just making sure that you don't have to be the one on social media. You don have to be the one constantly scrolling and checking out every single little thing and you don have to hire anybody to do that right And that information and understanding of the market is second to none and it learns from itself right That's the other thing with agents too. So let's say you have a viral image that you put out there that doesn't do well, right? Well, AI can say, okay, maybe like it can make certain inferences and it can give you reasons why it might not be a good idea to, maybe your brand just doesn't have like the right thing or maybe it just didn't have the right, maybe it looked a little quality. Like there's all these different things that it can look at. and help you iterate as well, because it does get better with you, right? That is sort of the fun part about agentic AI and generative AI as it stands. And then that last part is that the last sort of column, so I guess it wasn't really three, it was four, with the preparation and the execution kind of working together very quickly in the middle, preparation first. The final one is going to be what I like to call, and this is like an agentic term, observability monitoring, right? So that's when the AI is like, okay, we've done all these campaigns,

we've reached out to people, we have this creator that's done super well before, bringing them back in the loop for a new product launch or a new variation, new flavor, whatever it might be, right? How is that working, right? And then that will feed right back into the marketing. It'll feed right back into every previous step, right? The execution, the marketing mix preparation, and then that research as well so that you actually have agents that are proactively making sure that, okay, the next campaign around, we're exponentially getting better at what we do, right? So you can talk about the tools all day, right? I mean, you can go from, I know people who've created some of the most complicated things I've ever seen in Zapier, in Aden, in, you know, even in Buffer, they have like workflows now and I've seen people do it in ClickUp. Like the tool, the AI tool to some extent matters, but at the same time, there is so many now that you have a lot of different options. It's understanding how agents really work. It's understanding this consistent learning. It's understanding the sources. Where are you getting this marketing, you know, research from? Where are you getting, what are you connecting it to, right? Can your tool connect to a content calendar and to an ad creator and all that kind of stuff? can you find tools that do all in one? I've seen people do it a lot of different ways and a lot of

different efficiencies. All of it goes back to one fundamental thing, Mark, that I think you'll definitely agree with is that beyond even the concept of agents, we really are in the age of systems, right? It's not about just doing a task. It's about AI learning from a routine of tasks and a routine of iteration and a routine of constantly innovating and creating more and more personalized material. But the only way you can do that is with a system and with a set of tasks that, again, mimic a human and then go beyond human capacity, right? People are just creating little automations here and there to save a little bit of time. That's fine for personal efficiency. But what you're really trying to do is you're trying to have a whole system end to end with the proper approvals of human in the loop. And that's kind of how I, I kind of forgot to start with that, but I'll end with it because it sounds more dramatic anyways. I think that I think this idea of, you know, systematizing the work that you do and not just making it into a to do list every day, right, is the real difference. It's understanding the nuance between that, right? So that's kind of the first sort of set of answers I'll give, but I'd love to know how you want me to expand on that. I can have you talk about different examples.

Yeah, I mean, if you have any examples, for sure, I would only add that once this is set up and working, and I agree with the systemization of it and the workflows and having all that as automated, as long as there's humans in the loop, remember that. Yes. But beyond that, I've seen some pretty amazing customer support AI voice reps. So, and I've been blown away and I don't get blown away very easily, but these things are trained on this product. It's trained on everything that you give it. And then it learns. It'll ask you later on, you know, to review the tape, so to speak, because it records the calls from the customers and make any corrections. And it just learns and gets smarter and smarter and smarter from it. I mean, I've paid hundreds of thousands of dollars for a bunch of customer support reps that could be handled by one AI rep. that's damn good and it works 24-7. So that beyond what you've just talked about, which I fully agree with,

that you can just continue on in building this e-commerce company. Yeah. Go ahead. No, I think we've wrapped it up in terms of what we've given a lot of information out even now. Is there any place that you would have them start? I know we talked about Astra, just getting the ideas and the research and the competitors. But is there any place that you would say, hey, you've got an idea for a product? Obviously, you've got to go create the product. But is there anything else that you would add to that to say, hey, you've had this idea, now go do it. Here's where I'd start. You know, I'm not going to lie at all. like I've seen Shopify itself, right? The kings of e-commerce have essentially become an agentic commerce company. Like they literally brand it on their homepage. It's quite funny. Like it doesn't become a superstar in e-commerce. It doesn't become a superstar in AI. So whether you love them or hate them, their resource library is incredible.

And it's a good place to get started and understand the nuance. But that being said, yeah, I would say Astra, you know, is pretty solid. I honestly, you know, Claude can run a lot of your marketing mix these days. Like they have all the connections to Instagram and Facebook and like it can kind of do the work. I do love Buffer, you know, a big content, you know, calendar person, been using it for a long time. But genuine, like even in Notion and other, you know, similar product management tools and stuff, like there's marketing in HubSpot, there's all these different things. So you can start with really any of the baseline major industry tools or look for more niche ones. I will personally say that I think the baseline tools are a lot stronger at offering resources. I think the smaller startups with the really cool products often, you know, things are a little bit buggy. Things are a lot like they're still figuring out their sea legs. Like I've used a lot of great e-commerce and marketing tools, as a matter of fact. But I found that they're like Academy is not very good. So you're like you're jumping in blind and it's by the time you're setting it up, it's like it took way longer than just, you know, firing up, you know, HubSpot marketing. Right. I wouldn't say go as far as like using like the biggest tool you can find, like a sales force is probably not going to be helpful here.

Or like what is a big cartel? I don't don't just like jump to whatever. I still think that like Shopify and the entry level tools out there for marketing and for operational management and agentics stuff is great. But I will say, you know, now that I'm thinking about it, now I'm speaking out loud, I really do think like Zapier is incredibly, incredibly powerful and probably one of the best tools to learn. And other than them, I mean, like I said like Claude genuinely does do a lot of that like part two and three that execution preparation And then perplexity is actually the best tool by a long shot for that first part of it That the research and the analysis It probably does the most impressive there of all of them. And then, yeah, let's say like Chachapiti, Claude, HubSpot, Shopify for that second part of it, Zapier as well. And then the last part of it, that sort of monitoring, you know, you can do that with a base LLM as well. I've actually had Perplexity look at my shops and, you know, work it in. Like it can do cool monitoring there as well. Zapier and stuff like that does a pretty good job too. So I could probably, maybe there's like a whole list of like tools I can send.

I can't remember them off the top of my head, but they can be in the show notes, right? There are plenty. I'm just thinking of like the ones that I've honestly seen just do the very, very best at what they are. And, you know, there's all your Marketos and other things you can use as well. It's track marketing and obviously Meta Business Center and stuff has great analytics. But I'm just forgetting the name of some of the other more niche analytics tools that can tell you some interesting gaps as well. So like I said, I should have brought up Claude because with all their API connections to different tools. It's really good. Especially with Asana. If you use Asana or most of the CRM system. Yes, Asana is great. I actually was surprised you brought up Zapier because I thought Zapier would have been finished by now because of what Claude and ChatCPT have done with being able to connect, make these connections automatically with some of these platforms that Zapier used to be the go-between. So what has Zapier done that ChatCPT and Claude can't do? You know, the more I'm thinking about it now, I think there's a lot of nuance in this part that I was even kind of talking about too

where you can track how it's actually doing and how it's improving, which I think is really awesome. Like it does give you a lot of proactive recommendations on how, you know, it can, you know, Claude can connect a tool together just by telling it. Zapier, what I like is that it actually tells you the strength and like feeds you back how good that data is and how it's transforming. I think that's very powerful. I think that's really good. Like going back to that monitoring and, you know, sort of ability to observe how good your AI is getting. I think that's one of their biggest things. And I do think, too, that they just like generally what Claude doesn't do very good or ChatGPT or even Perplexity in this case is like I think that visualization and collaboration on workflows is just infinitely powerful. You can kind of do it if you share conversations and stuff in Claude, but it's not like and you can ask it to visualize it for you. But it's not really the same thing as Zapier where you can literally drag and drop. And, OK, it goes to like this web app and like you can you can just jerry rig some really, really cool things. but you know like Zapier has a really good job of scaling these things together right versus like I could I could say and I can say at least for the time being like it's not a it's not as

visual friendly and b I think there are times where because they're updating it so much it actually kind of like misses some of these nuanced things that you probably want to be careful of so connections will randomly drop not too often with the bigger connections they have but with you know like Zapier has a good job of you can track like how MCPs are changing at other tools that you're connecting to, things like that, right? So let's call it like the nuance of workflow building. I think Zapier excels at. I, you know, welcome Chad Giveti and Claude to keep building what they have as well, and they might eventually take it out. But I think Zapier has done a pretty good job of holding their own with like a, like I said, just this visual layer and data and analytical layer. So you really, really understand how to get from AI doing a task for you to a full agentic workforce as a beginner, as a visual person, right? Not as an engineer or something like that. Yeah. Okay. Love that. Love that. Well, this has been just one of my favorite episodes in a long time. I'm having you here. This is a lot of fun. I've really given out a lot of good information, especially for those starting an e-commerce company, really any company, not just e-commerce.

But this is the basics. And then, you know, we could spend hours filling in the blanks. But these are great places to start. Learn this stuff. It'll really transform your company that you're trying to build. And, you know, if you're up for it, maybe we should have you back for when you get into that $10 to $100 million company range and how you go about, you know, transforming your current organization into an agentic organization. That might be a fun episode as well. Absolutely. Yeah, it gets very different when I work with my Fortune 500 clients versus e-commerce. Of course, right? Some things not actually that different because we're all learning this together. and some things just completely bonkers is the way to say it. It really does get crazy the larger scale. And being with some of the biggest retailers in the world as a consultant and as an executive advisor and things like that in their AI and tech departments, it's crazy, man. I'd be happy to come back and chat about it. It's a whole other conversation, a whole other set of tools and a lot of interesting principles that take what we talked about today

and stretch them beyond to a scale that's unimaginable, but AI is making it happen. So here we are. All right. Well, Nikhil, we always wrap things up. Let people know how to contact you and anything else that you want to talk about, a short commercial if you want. Of course, of course, yeah. So, you know, Nikhil Vimal on LinkedIn, a great place to reach out to me. I check that a lot, nvimal.com. I'll plug my book, The Essential Pivot. It's, you know, what my learning and innovation, or what I've been learning, you know, working in the innovation world for the last 15 years of my career, turning that into a book from a lot of amazing stories, anecdotes, client examples, all that fun stuff. It's a good book about understanding and how to navigate change, how to separate signal versus noise, all that. That's launching at the end of this year. Envamall.com, my personal website has a place to put your email in for the pre-order. And yeah, I do a lot of cool stuff on LinkedIn, though mostly in a lot of my communities there. So always happy to connect with people, chat with them about AI marketing. And you know, Mark, I'm looking forward to being back.

You got me all excited about what that next episode could be. So I hope we make it happen sometime soon, man. We are going to do it. I'll send the info for it. Please do go and connect with him on LinkedIn. Either follow him or ask for a connection request if you've got something very interesting you want to do with him. And with that, I will say thank you again. And we will do another episode probably in the next month or so. Of course. Of course. Thanks, Mark. We'll chat again soon. Thanks for having me. Thank you. Thank you.