AI-Powered Ecommerce: Real-World Hacks & Advanced Seller Tactics with Michael Michelini

In this episode, Mike Michelini takes us deep into the trenches of how he’s using AI to supercharge his Amazon and ecommerce businesses. From analyzing Sellerboard data to creating smart inventory forecasts, building live chat agents, and sourcing products with just a prompt—he shares practical hacks and honest insights from the field. A must-listen for experienced sellers looking to future-proof their business and harness the real power of AI.

• 📄 Download Slides: Ecommerce AI Tactics PDF
• 📊 Sellerboard – Profit analytics for Amazon sellers
• 🤖 Grok (X AI) – Elon Musk’s LLM on X (Twitter)
• 🧠 Claude by Anthropic – Conversational AI with strong context memory
• 🔐 Venice.ai – Privacy-focused AI platform with crypto payment support
• 💾 Pinecone – Vector database for scalable AI memory
• 🌏 Cross Border Summit – Nov 3–5, 2025, Chiang Mai: crossbordersummit.com

Thank you everybody for listening in.

• Direct Download: Right-click here and click “Save As” for a direct download
• iTunes: Listen and subscribe on iTunes for free!
• Android: Check out Android Options or Listen via Stitcher Internet Radio streaming
• If you enjoyed this episode, leave a positive review on iTunes, & more (directions here)!

(00:00) Episode 464 of the Global from Asia podcast talking about AI and e-commerce, some hacks, some real life strategies I’m using to accelerate our Amazon and e-commerce businesses. So, let’s tune in today. Welcome to the Global from Asia podcast, where the daunting process of running an international business is broken down into straight up actionable advice.
(00:27) And now, your host, Michael Molini. I hope everybody’s having a fantastic morning, afternoon, evening, wherever in the world you’re listening to this or maybe watching. I am here in my home studio. I was on the road busy as always. I mean, I was talking to Andreas, my partner, one of my businesses, and he’s saying that he thinks AI is going to make people lazier.
(00:54) It might be true for most people, but for me, I’m like I’m I’m I call it supercharging, working harder than ever, building more than ever. And I’m excited to share with you a few of the strategies that I’ve been using. I had ecom meet up here in Mai, and I figured I’d take the camera with me to record. We also have an amazing mastermind at Paya in mid July and we’re going to start to do more exclusive experienced seller meetups just because we want more experienced people and no offense to the new people but it’s like a different kind of event. So I feel
(01:31) like I got pretty advanced and some of the people there weren’t following it. So I’m going to move to more advanced meetups just FYI. Stay tuned for that. of course our amazing crossber summit which is very advanced I mean I want to learn from it and I’m an experienced seller for 20 years so it’s an event I want to learn from so anyway some I feel like advanced or maybe get some insights about ways I’m using AI for e-commerce for reporting a lot of it might not be external it’s internal but that helps you grow so let’s tune in attention
(02:08) e-commerce entrepreneurs ready to take your business to the next level while Living the VIP life in paradise. Join the GFA VIP Villa Mastermind, Paya 2025, an exclusive 4-day mastermind for established e-commerce sellers. From July 13th to 16th, you’ll be immersed in a luxury private villa with a pool, dedicated meeting spaces, and 10 handpicked highlevel entrepreneurs, all focused on growth, strategy, and big wins.
(02:38) Structured knowledge sharing sessions, actionable insights, and unmatched networking. Plus, adventure and fun in vibrant paya spots are extremely limited. Secure your place before it’s gone. So, I feel like if I use a keyword AI, more people would come. E-commerce is not as cool as it used to be, I think. Right. But I’m going to share some of the things I’ve been learning with AI for e-commerce and give you some prompts and tactics and and hacks if I can figure out I use this gamma.
(03:10) But I think one of the big ones is is personalization Alex is sharing too about AI replacing humans. But with that I think will mean websites might totally change based on the person visiting the website. I think we’re still used to seeing what everybody else sees, but it even already happens with with Google search, Amazon search.
(03:32) The results are based on the person. They’re not just based on your search keyword, but they know who you are, where you are. I think that’s going to get even more more apparent and clear. Like websites might just totally morph based on who you are. It’ll know all about you already. It will give you products and services that you want based on that.
(03:51) But what I’ve been liking a lot is inventory and and the pricing optimization. So I’ll show you exactly what I’m I’m doing. So I put a lot on websites. So this is I’m not going to put this on. If you’re not here, you’re watching this, you’re not going to see this. I’m not going to share. It’s totally totally open.
(04:09) But for those that are in the room, this was spit out by AI. What I did is I took I use Sellerboard for my Amazon and I downloaded it for this month of May, last month. And what it did is it read the top products that sold. So these are my I cannot see. Sorry for my glasses. Can you read it? What are the products? I there are various bar bar supplies.
(04:32) Bar supply. Yeah. Yeah. Like I’m in in the bar supplies category. So I’ll show you a little bit. No problem. No problem. So basically I’m pulling this from sellerboard. You can pull this directly from Amazon. You don’t have to get fancy about the data, but you would I go is 1 million.
(04:53) Why do you want 130? I didn’t 130 came because I see one million and then you you can get this from Amazon directly or Shopify or any kind of platform. But I downloaded it by by uh month. I just took one month sales. But what I also was doing is I started putting more and more data into it. So, I just use grock.com and then I just I just literally go to Grock and then I would see if I can show you my chat history.
(05:19) My wife is using Gemini and she’s using she subscribed to to So, basically what this is I’m giving it the data and it’s giving me results back. So, this is in two years of history of this of this sales. So what it breaks down is some of the stuff’s pretty basic like just total sales, but it gives my top product by sales.
(05:40) It gives me my sales growth for the top product. So it identifies which when it started to really sell, but you can see how I started really dominating like actually last month was record for me. I think that I think the tariffs were good for me and then also highlights which had high high refund rates.
(05:58) But my favorite part is the actual conclusions. So, it doesn’t just give you the data. It gives you like which ones are not doing well, which ones have bad sales, which have high refund rates, which it says sales peaks, maybe holiday growth opportunities, tells me products that maybe I should focus more on.
(06:16) I This is better than a a human. Like, I have great humans working for me. I sent this to like one of my Amazon account manager and he was like, “How did you do that?” I I showed the charts and he was pretty blown away and it just gives you like recommendations. So, inventory recommendations. I feel like that’s one of my weaknesses in e-commerce is predicting inventory sales.
(06:41) But now you can I putting in two years of data. I’m uploading all this right in in raw. Like it’s not like it’s I’m not spending a lot of time. My prompts are just simply like see attached. Catched. This is a longer one. I actually sometimes what I do is I put I transcribe a Zoom call and then I put the Zoom call transcriptions into Grog and then it summarizes it.
(07:04) But I do it all within the same thread. So my threads I call them threads are very very long. This is cuz I wanted to keep the memory. But you can see the first one was you’re a financial analyst. Help make a report showing the performance of last month’s Amazon sales taking into account which products are most profitable versus highest turnover.
(07:23) make recommendations on areas we can improve. And then I pasted it, but then I realized it wasn’t as good pasting it as it was attaching it. So, I uploaded this I uploaded it as a CSV in the second shot and it it gave me it better. Again, the stuff’s still just developing. Like, it was kind of stuck.
(07:45) I think it didn’t really fully pull it out, but I said, “Can you try again? Here’s a spreadsheet.” And then it gives me this report. Right. And then I download this report as HTML. See? And then I put this I just like to have it on the website. So I put on this it’s my private kind of like staging server.
(08:06) And I just host it there and then I just send it to people. Maybe I should be private about this because it is my my sales top products. But I’m not trying to index it in Google. But I just keep talking about it and then I use this long thread. But honestly, Mike, you were always honest about things you do.
(08:25) Like people know how much success, how much unsuccessful, you know, I don’t think there is need to copycat sellers because that’s the problem with e-commerce as soon as they they know what the takeaway is. But if I if I’m correct, you have some consultants. You also about people who wants to Yeah. I have a I have a service company where I I have a couple clients that I help on a long-term basis.
(08:53) It’s not really like a You have the service that you offer people to help them achieve success. Yeah. I mean, I work with a couple of like more established sellers that I know. It’s not like a big service I tell everybody about cuz I’ve learned with services. I don’t like to do this the sales part and I have to like explain and educate.
(09:11) So, well, you’re the coach. You’re not. Yeah. So, if I have a a friend or somebody I know is established and they like what I do and they somewhat flexible, I might work with them on like a on like a monthly retainer basis with our team. Yeah. And our skills. Yeah. But yeah, it’s not my core service. But of course, good point about copycats.
(09:33) Yeah. Yeah. Yeah, I mean that’s one of the issues, but I’m not that I mean I I I’ve shared about my brands on on my shows, but of course even with AI, you can also use it for sourcing. So you can take a picture of a product and put it into probably any LLM now, but I I like rock, but any LLM, you can put a picture and you can or maybe some text and you say, “Find me suppliers of this product.
(09:59) ” I’ve blown people away with that because they always get scared because like sometimes a seller friend will say that he has some only one factory in China can do this. I find it in like one prompt and he’s like cuz he doesn’t want to tell me like is it this this blah blah blah engine company limited like he doesn’t say anything.
(10:20) So yeah, the barrier is dropping, but at the same time, I think I think the hard part is to get noticed now, right? You have everybody spend so much money on advertising, right? So I think the moat is the distribution to distribution or the attention. I always think of Gary Vee and he says you’re trading trust and attention. That’s all we have.
(10:42) That’s why I I yeah, I really try hard to to keep trust because I think that’s the most valuable thing within a world of AI is is trust, but it’s it’s trust and it’s also distribution or attention, eyeballs, people paying attention cuz now anybody can find a product, anybody can list a product, anybody can make social media ads or make anything.
(11:04) It’s getting almost scary, right, to stand out. But let’s go back to my little simple outline. But yeah, so one one use is inventory and pricing optimization, right? So we still have our human team, our experts like Christian might be watching this, but I I still tell them use please if you don’t use AI, you’re going to be behind.
(11:26) So use this to give you superpower. So you can put this data into it and it gives you ideas of price, trends, inventory. It’s it’s mind-blowing because I’ve even I felt like I was somewhat guessing before how many to reorder and now you can kind of see historical trends, history. It’s it’s really amazing. Customer service, this is something I’m not still an expert in, but I I know this something we got to get better on in my business.
(11:55) But with with with AI now, you can have, of course, automated emails. I mean, the very basic is nobody in my team asks me for a script anymore. I used to have people ask me for a script of what to say to a customer. AI helps them with the script. So even for me, I when I have to deal with like a a a difficult situation of of a customer or a client, I I just type freestyle to LLM and it helps me like write a good response.
(12:23) So that’s a basic one, but the best are these chat bots, which maybe Alex can share in future session. Clo. Have you heard about Clo? Yeah, Claude. My firefighter loves it. Yeah. I mean, she said the all the coding is very accurate, high accuracy. Yeah. CL for the prediction for prediction model. And she said it works. It’s it’s not good as you do it by yourself.
(12:46) It’s not good as that. Let’s not kid ourself. But it’s high accuracy, high almost zero mistakes. Yeah. So, CL of course is a great LLM, but I what I mean by customer service is where I have a live chat on my website. Uh-huh. And they can chat to an AI and it almost seems like a human. I mean there I think it’s been around for 5 10 years but people always know it’s agent and they’re like talk to human talk to human.
(13:11) But I’m trying to make it where it doesn’t seem like an agent. So when I feel comfortable about that I’ll start to put like chats on my websites so that they can get an answer in real time from the AI. So that’s one one thing and it is possible and it can be done and it’s just I haven’t gotten to this yet but you can have chat bots on your website.
(13:31) So maybe even for the meetup Stefon you could have a chatbot. So I know I don’t want to I know you’re so busy talking to so many people. Imagine you could have them talk to someone that’s trained on your experience and they could get answers. Now are getting a little bit the the type of people that are coming this year have some not unusual commercial people like you or Rene or him or him.
(13:56) So a little bit unconventional. So the the request that I’m getting out of ordinary so I I’m not confident with AF because it’s true. I mean to to be effective it should have some kind of context and history of how to handle that request and that answer but I mean customer service for sure is totally going to change.
(14:19) I mean I work with a lot of people in the Philippines on our team and there’s a lot of these BPOS and those guys are in trouble. I I think they’ve already if they’re not already Yeah. their their days are numbered because you don’t need those big rooms of agents human agents like responding in chat or phone calls. That’s all moving to AI and I should get ahead.
(14:39) I have to figure there’s so much I’m learning. I’m I’ll show you some videos too. They’re fun ones I showed Alex. But customer service for sure you got to like get this going with AI. So here’s some exact prompts if you want to use customer analysis. So you can just use something like analyze purchase history to suggest five personal recommendations.
(15:00) Basically another way I look at that is product ideas. So if you need new product ideas for your Amazon or e-commerce business, you can take your current products and your sales data. Actually, I could probably go back but may I can do it right now in that same thread, right? I could probably even ask right now like can you suggest five new potential products based on this sales history above? I don’t even probably say above.
(15:30) So the cool thing is you’re training your model, right? So you notice I didn’t have to give it any data. I just prompted something because it already it already knew it already has the data. So once you start to train it, it’s it’s giving me giving me ideas right now as I as I am doing this. So one of my favorite things to do is I say give it to me in HTML CSS because then I could download it as a web page and I can like host it really fast.
(15:57) But yeah, it gave me see it gave me five product ideas. Interesting. And I might even get some ideas from this. But what’s really interesting is it’s it’s it’s getting smarter every day and it’s learning about you and your business. So the more you talk to it and train it, I think you actually have to have optin. I I I allow it to remember.
(16:18) Some people turn it off because there’s a setting in in these LLMs. You can turn this on or off. You can personalize using X. You can have See, I I like this on, but I saw somebody I watched the YouTube video about using Grock, advanced Grock. He turns this off. Personalizing Grock with your conversation.
(16:38) I leave it because I actually wanted to remember cuz I’ll reference it. I’ll say based on what I said this give me ideas. So, I’m I’m consciously knowing what I I think we should all be aware what we put in there, it’s going to remember. Even if you delete it, I don’t think it’s really deleted. So, what you say to these things is going to be tracked.
(16:57) I know, but I don’t think you need to worry about because you’re a vanilla guy. Yeah, I always say that about I need to worry about what I’m looking and what I’m searching. So, I couldn’t do this like your your your two kids and a beautiful wife. Yeah, but like your medical look up like medical things or or maybe it’s not I I don’t I know that’s something like a lot of people say about privacy.
(17:22) Oh, you have nothing to hide, but it’s not really about that. It’s like certain data you just don’t want people to to have or store about you even like like I sometimes see this is I ask medical questions to venice.ai. So Venice is I pay for it but you can get it for free. It’s also a little bit more crypto friendly and cryp it’s got even a crypto token called V.
(17:42) You can pay by crypto. You can pay by crypto. Okay. But also it’s private so it doesn’t at least they have a long post. Their CEO put a tweet out but they don’t store your data. Do they use a Kosh hosting? So, you should just be aware of what what you’re asking and where you’re asking cuz I think they just said like last week OpenAI will give their data to like the FBI if they ask something like this.
(18:08) I I think they say they store and they can use that if someone asks for it. So, just be aware of this. I’m a privacy person. Even if I don’t think I have anything to hide, I I don’t think Yeah. Just on that search you just did, how confident are you that it’s not hallucinating or that it just gives you a bias to give you a very reasonable list of recommendations? Like how confident that it’s really Well, I mean I I think about the product recommendations.
(18:36) Well, he’s a product research guy. We could ask him. But I mean, one, I know my obviously I know my my product pretty well. I’ve been doing this for a long time. I can tell it’s looking at it’s it’s not really doing totally unique. It’s doing iterations of what this is my top selling product.
(18:53) So, it’s kind of it’s making a variation of friendly interlocking sne. So, these are like liners that you put in a restaurant. They’re they’re more commercial use, but homes use it. And it puts a little bit of a gap between you and the the glass and the so you don’t you don’t put the cup here and then the water sticks inside the cup.
(19:21) It puts air so it dries. It’s clean. Yeah. So this this is one of our better better ones. So it’s it looks like it’s just giving it’s not given truly unique. I don’t know if Alex or others saw it, but I forgot it was some famous I think the Google one of the Google founders says AI gives you the best answers when you yell at it.
(19:41) Did you see that one? It was like it was just like a few days ago. I saw it on Twitter, but he says like, “Is that really the best you can do? Can’t you do any better than that?” And he they say that’s how you get and then somebody quote retweeted and says just like he does to his human staff. True. But of course, you got to be careful about hallucinating.
(20:01) But at the same time, I it’s not like I’m totally blindly trusting what it says. I I know my product and my business. But yeah, you got to definitely be careful about hallucinating. Sometimes it does. I’ve noticed when it doesn’t know the answer, it just gives you some kind of BS. But I feel like every day this stuff is getting better.
(20:19) Do you remember there’s something called the Will Smith spaghetti? That one. Yeah. Yeah. So they’re they’re almost celebrating it now, the AI people, because the there’s this anti-AI people that think this is all BS and it’s going to go away and it’s not going to change, which seems like the stupidest thing you could ever hear.
(20:36) Like it’s going to it’s not going away. It’s going to keep getting better, but they’re like, “Oh, look this Will Smith guy. Look, he’s” and the first person, I think it was 2023 and it it does look really bad, right? Like it’s totally bad, but every year they’re doing a new one and now it’s like looks like Will Smith eating spaghetti, right? Like, so I also think it’s getting better, but it’s also based on your prompt.
(20:58) When I am more serious about my prompts, my prompts are like a paragraph, sometimes even multiple paragraphs. I’ve actually think about prompts now as cuz I’ve been working with like online teams for 10 years, more than 10 years. I think 2012 I think even I think I’ve had like work at home moms in Texas in 2006 I think was the first 2006. Yeah. I don’t How old are you? 44. Okay.
(21:23) You’re young. Yeah, I know. But I read even before Tim Varys, I hired a work military wives that they there was like a website about military wives. Military wives. They were maybe hot. I had two. One, she used to be a U-Haul call center, but they basically couldn’t work a normal job because their husbands were traveling on military.
(21:44) So, they needed online jobs. So, they it was like a military wife group. And I heard about it and I hired two two women off of it for my customer support. But what I learned is I don’t want to offend people, but I feel like a prompt is almost the same as giving a human descript. It’s like I’m talking See, if you give a human a very unclear half a sentence, you’re going to get crap back.
(22:07) Yeah. But if you give a human a very clear prompt I call or directions. So at the same people complain about AI saying it’s not good because they put like one broad sentence and then they’re like, “Look, this gave me nothing.” But if you’re given a clear input, you get I call it garbage in garbage out.
(22:25) So you also should be aware of that. So sometimes it’s about your prompt. But yeah, I don’t know how about hallucinate. I haven’t noticed it as much. I don’t know. Me neither to be honest. Yeah, I think it’s getting better, but it still does happen. I mean, my suspicion is about judgment and the AI’s ability to apply judgment based on human experience versus if I wonder if you could prompt it again to produce five more recommendations and it gives you another set of five and then you ask to then rank all 10 in in in order then
(23:04) it might actually mismatch. So then your original prompt might actually be inconsistent, right? because you ask for the top five initially. So my my suspicion is that with AI it constantly the logic is always trying to give you a reasonable rational answer as opposed to applying real judgment.
(23:23) And I don’t know if AI is really there yet to to kind of get to what the humans ultimately do in sort of decision making in using judgment for decisions or recommendations. So that judgment like high cognitive ability like high level cognitive ability like give like authority to the AI which MML LLM can’t do you know yet to just like it’s not like a real real AI.
(23:53) Well, I think it was Alex said about negotiation, but there is it AGI, artificial government intelligence, right? But they’re basically saying or maybe maybe I’m talking about ASI. There’s all these things, but basically they’re saying like we’re going to just listen to the AI. They’re going to be like the government.
(24:12) They’re going to be the boss and then they’re we tell them and they give us judgment. But we are far away from that step. Like let’s not Yeah. I mean, I don’t know. I’d love to hear what you guys Um it’s not capable for such a thing to be honest. It’s dangerous to say can’t. It’s just a matter of time. But yeah, you can I think the discussion is how long you know matter of time.
(24:31) It’s going to be taking a while. I don’t know 10 years. Are you sure? So fast. I think it’s probably what’s I read an article that says it’s not about the technology, it’s about society and the humans. You know you remember graphanting graph graphing graphing that this layer single layer of carbon carbon bone together I kind of understand I don’t know what they were telling that it’s going to revolutionize material but what happened actually this is like everything will be new like in 10 years we will get so I have a really bad it’s an embarrassing one of
(25:03) me but I’ll show it this is me in the year 2000 in my university answers from different it’s very my friend, but the reporter was asking college students, do you think the internet’s going to change society? Came back home for vacation and some of them are actually used to computers even though they have no idea.
(25:25) Yeah, it looks I I was very nervous and I wasn’t clear. But the point is they thought they’re like, “Oh, this internet thing.” It was the year 2000. They’re like, “Oh, we made it through Y2K, but the internet’s not really changing anything.” What ch internet like what I can like put chat chat rooms.
(25:43) There were there were there was people doubting the internet in the year 2000 that it would change society. Look, I live in freaking Thailand. I sell in another country. I have a wife from China. I like that’s because of the internet, man. Like if there’s no internet, like I saw somebody tweet recently. I follow Twitter a lot. He says, “I wouldn’t if the internet didn’t exist, I wouldn’t exist.
(26:07) I see cuz maybe his like the life of like I think AI is same as the internet and I think people are still doubting it now like they’re like oh yeah it’s like a cool chat thing I can ask some question is talking about judgment I think we’re going to trust it to judge I think it’s it’s I think it’s a long way to come to this point I mean it’s not different in negotiation you put some parameters you say based on these parameters make a decision and it might be better than a human judge because a human judge has like emotion oh He’s so
(26:37) sweet. Look at his eyes. Like he’s younger. He didn’t he didn’t know as much as you. You’re like this mean person. I’m going. You judge people differently based on them. Even a judge, a human judge, they they look at two cases in two different data, they’ll judge differently cuz they’re a similar negotiation.
(26:55) But you give them parameters. It sounds cold and horrible, but I’m also a blockchain believer. And blockchain is also not as it’s it’s it’s finality, right? people get hacked and lose their money. There’s no reversing it, right? But it’s similar to like judgment and AI over from this blockchain thing is people because they just move to the new over I think it’s anything but over stable coins just got passed in the US there there are better technology from blockchain much more advanced much more stable much more less complex I mean that’s a long discussion
(27:30) I don’t know what which other technology she means but I think just last week Shopify announced stable coin payments with Coinbase so now you’re talking about crypto. Well, crypto is blockchain. Yeah. Yeah. But like I I’m sorry. I thought you were referring to blockchain as a technology as a whole. Yes. Mhm.
(27:50) I mean I mean you you’re IT guy. Please say something. I just marry IT I it guy. I’m it guy but I don’t know anything about informatics. The main the main point is judgment I believe is something you can program. There might be exceptions. Maybe a human might have to intervene in a certain point just like negotiation.
(28:11) Maybe it gets to a certain point and then maybe a human reviews it. But I believe you can program judgment. It’s it’s basically rule rules generally. I guess there’s humans. The big difference of a human from an AI or blockchain is they do make exceptions. No, I totally understand that crypto is based on the well not all crypto now to be honest.
(28:35) Like let’s face it but most of serious crypto is based on the blockchain just concept of blockchain technology as a whole but I understand from people who are in computer science it and data w now we move to much more simpler and much more efficient technology well you don’t need a blockchain for everything there there’s especially that’s a bitcoin saying bitcoin maxi saying but I I don’t fully agree you don’t need a blockchain for everything but there’s also you don’t need a token for everything there’s different layers of blockchain
(29:03) right so a lot of people think Blockchain means crypto token. But I feel like we’re a little bit off topic of this discussion and it’s a little outside of my expertise, but the point is yeah, you don’t need a blockchain for everything and there’s other technologies that I’m sure might be better for certain use cases.
(29:21) Yeah, definitely. I think that to address your question earlier about judgment. So under the hood, I used to be AI engineer. Under the hood, it’s all probability trees. We do the same thing as humans. We just we’re more quantum. So we do this. Whereas the probability tree is definitely binary. So what happens in terms of judgment is I run through all these trees very quickly and then the tree with the highest probability is the answer.
(29:53) Practically what I do is that I always have a board of directors. I call them board director. They’re just AI agents. And then each I agent is programmed to think like somebody I admire or I respect and I have them check each other and then the answers will be very close. You’ll be so surprised.
(30:15) So let’s say you rank more than 10 items and number three and four might switch when an advisor much much like a board company right they might switch but it’s very close. So because it ran through a probability tree and then when they kept get to a probability statement that’s very close that’s where it’s it’s go either way point toss yeah so do you do that how do you do I mean are they separate models separate you could do se model but in I don’t use drop but let’s say use AI you can actually have a person let’s say I like Tim Ferrris I like Gandhi I have
(30:55) somebody here and then each person just like a board will fair the answer from their viewpoint and then you have a goal of course that your goal is corporations they have to hit these metrics hit these objectives and they they all know that and then they’ll evaluate the answer and give you their perspective and they might tweak a few things but in the end don’t forget that their data is the same either through a vector database or That’s it.
(31:30) So, so that’s but you still need to have assumption these people are born by your board who they are. You need to like create that by yourself. So, it’s not objective, it’s subjective. You you could create them or you can get a person from your life. Bill Gates has a plethora of data out there that your LMA Mike was using.
(31:54) Yeah, this is this is one of mine. Hey there, my name is Katrina and I’m thrilled to represent New York Bar Store. As a bartender, I know quality tools and ingredients make all the difference. New York avatars stickers to syrups. They’re my go-to. Head over to new yorkbarstore.com to see their amazing selection. You you you don’t of course she’s attractive, but it’s not just like the visual.
(32:17) You have a background. So, you you give them a story and then they have a whole persona. So, they’re becoming a person. So you create these independent I don’t know what the word is like models or or agents and then they they’re just like a person that has history and perspective. So is that through custom GPT like each each board member you created a custom GPT on yeah but I think it’s subjective but you could of course everything because you want there you want the subjectivity there’s no such thing as is true so you want that Mike was was
(32:58) trying to store things it’s good but there’s a limit on on that database if that database I don’t know what it is for 500k for open AI at 500k you’re screwed So you really can’t grow that any longer. So So another way to do that is to API into block but your own vector database and that’s easy to do. Go to pine cone per month.
(33:23) That vector database is a true database to where you can as much memory as you want and drop you you can drop in rock you can drop paper to AI. You can drop in lava whatever you want. Okay. But that database is you. That’s what you are. And in that database, you can store your personalities or not. But I’m telling you, it’s it’s pretty damn accurate.
(33:48) When you say vector database, is that a a website that you just like what’s the vector data? What’s the vector data? Um, just to keep it super simple. So AI doesn’t use a relational database. You know, relational databases like Oracle, SQL. Okay. Okay. So this table, this row, it doesn’t doesn’t need that kind of database.
(34:11) It uses a database, a form factor called a vector. Okay. And a vector has associations with many other objects rather than just X and Y and Z just like your your tables and rows. How can we access this thing? I show one. So, so you API, you you build your your knowledge here and then you API into the LLM. Think of LLM as the CPU. That’s what it is.
(34:44) Brock, meta, open, they’re just CPUs. You get an analogy, right? So, you still need a database. Right? Now, people are using the CPU as a database. It’s like using their RAM on a computer. How much RAM do you have? 256, right? That’s all you have. Eventually, it’s going to have to drop on hard drive. I’m sorry. See, that’s what Sure.
(35:05) So, so, so now you got to start thinking about hard drives. Pine cone vector database is a hard drive. So, this is this factor database. Vector vector. Yeah. Vector vector vecctor. Vector. Yeah. Here native English speaker. Vector data. Okay. Wait. Yeah. This is a better picture. Yeah. Yeah. Because the halidation happens because the RAM I’m using that’s not the real thing.
(35:34) The RAM on rock or the RAM open AI is full. Okay. So it’s going to miss a lot of pieces that are posttop because the probability training right. So when you train a model you train all that data into pine cone or into into a vector. Okay. And that that’s the brains of the whole memory. memory. It’s the memory out. AI are all about memory.
(35:58) They either get memory from the web or they get it from you, but they have to store it somewhere. And if you store own whatever, you’re stuck with 256 or 64K or whatever they have because they can’t give you they can’t give you unlimited memory. They can’t they don’t have the capacity, but you can buy your memory, you know, or cool. Yeah.
(36:21) Just some other prompts. Of course, we know how to like create a blog post for for SEO or for websites, customer service, chatbot scripts. Just the tip of the iceberg, but I’ I’d love to hear how some of you guys are using AI, especially for e-commerce or just in general. And then then we can maybe just start to wrap up. And you want to we could do this discussion all night. All night.
(36:47) Yeah, that’s it. But we I mean I think the most important is like I think if you’re still doubting it it’s you have to use it right we’re there’s no I I don’t know I just think it’s a matter of time before you have to use it and no my mother get to get great result for her prediction thing and the the the database she used was was you need to pay no yeah this is free but you know it’s free there’s a limited version the database you use some European databases.
(37:20) They were they were the the the format the file formats were strange. So she needed to to converted in the more accessible format. Yeah. But she claims there are abundance of these databases just in Europe you you have this weird law that you need to do this specific format for this public which doesn’t have sense right now in 21st century to use it anymore.
(37:47) That’s what’s was what’s the biggest issue for her to convert and then she put it in the cloud and get really good prediction model. Yeah. Got it. Yeah. All right. Again, it’s not as it’s good. It’s work, but it’s not as good as real people are doing to be honest. I don’t agree. I remember just do you enjoy the podcast global? Do you enjoy all the community events that we do? The best way to support is coming out to our annual crossborder summit.
(38:18) com in Mai this November 3rd, 4th, and 5th, 2025. I have amazing speakers, amazing people getting together. We’re pushing the limits. We’re making things happen. This is where the movers and shakers and the deal makers come. It’s a limited supply of attendees and tickets, and we have been selling out every year.
(38:38) So, I recommend checking out earlier cross border. summit.com. Thanks everybody for watching. Give me some feedback if you like me taking the camera with me at these events, keeping it real. Some stuff I try not to share. Alvin, my editor, please. I talk a little bit sensitive stuff. I might not share it or show it.
(38:58) I also made some slides. You can download those if you’d like to download slides on the show notes. I was using as an outline but experienced I think that’s always what we talked about global major even since the beginning we were talking to people already doing business want to go global wanting to go to the next level of their business and we’re continuing that now going more advanced and more experienced so thanks so much have a great day take care to get more info about running an international business please visit our website at wwlobalfroasia.com
(39:34) That’s ww.globalfroasia.com. Also, be sure to subscribe to our iTunes feed. Thanks for tuning in.