I Started an E-Commerce Business with a Satellite Dish & Zero Customers - Asgar Dungarwalla, GiftsOnline4U

In 2007, Asgar Dungarwalla wanted to put his drinks wholesale business online. The problem? He lived in the English countryside, and there was no broadband. None. Zero. Not even copper wire.

So he did what any reasonable person would do. He looked out his window, spotted the post office tower in the distance, and physically mounted a satellite dish on his property to bounce a signal off it. Two megabytes synchronized, both upload and download. Pretty futuristic for the time. The only catch? The signal would drop every time it rained.

Today's guest spent 10 years in the RAF as an aeronautical engineering instructor, teaching propulsion systems on Phantom jet fighters. Then he ran a precision engineering education at Cranfield, then a drinks wholesale business that scaled to millions in revenue, and then somehow ended up here, running GiftsOnline4U, a personalized gifts company with thousands of products, custom watches, engraved champagne, and a brand new AI workflow built on his own local server.

This conversation is about what happens when you've actually lived through every major shift in commerce. Valves to transistors. The first moon landing. Computers. The internet. Mobile. AI. Asgar has built through all of it. He's lost his rankings overnight to a Google update, watched delivery expectations collapse from 7 days to "I want it now", and is currently training his own LoRA models in ComfyUI to make caricatures of cricketers because the big language models can't get brown faces right.

If you're building anything in commerce right now, especially with AI, this one is worth slowing down for.

Key Takeaways: 18 Years of Commerce, Compressed

The Infrastructure Reality:

• In 2007 there was no individual email, you had to email a company and ask them to forward it to a person
• Asgar physically mounted a satellite dish to get 2MB synchronized broadband, signal would drop in the rain
• His first website was built in frames, which bots couldn't scan, his team had to strip it out manually

Customer Expectation Collapse:

• 18 years ago, customers were happy to wait a week for delivery and would say "yay, I got a parcel"
• Today, even with personalization, GiftsOnline4U has had to build same day and next day delivery options
• 80% of GiftsOnline4U traffic is now mobile, 18 years ago it was 100% desktop

The AI Overviews Playbook:

• SEO is now just the baseline, AI overviews need authority signals from Quora, Reddit, LinkedIn, Instagram
• Google takes data from real human-moderated platforms because that's where actual expertise lives
• One Penguin update killed his ranking overnight, which is exactly why over-relying on any single platform is dangerous

Q: Take us back to 2007. What was actually possible in e-commerce back then?

Asgar: To be fair, 2008 was a massive crash and there was a bubble coming, but there was also huge investment in what the internet could do. For us it was finding the technical help to create your website. We found some people who spoke a good game, and they developed a full e-commerce website where people could personalize stuff, but they built it in frames.

Today you'd say absolutely no way you'd build it in frames. But nobody knew. The bots couldn't scan the website, so important things just couldn't be indexed. We had some technical people in our business who came in later and removed the frames so you could actually scan it.

And at that time, we didn't even have proper broadband because we live in the countryside and our businesses are based at the back of our property. We were lucky where we lived, you could actually see the post office tower. So we got an opportunity to put a satellite dish which connected from us to the post office tower. Even at that time, we got two megabytes synchronized. The only problem was when it rained, the signal got interrupted. But you work through it.

🔥 ChaiNet's Hot Take: Founders today complain about Shopify themes. Asgar literally bounced his internet off a government tower so his business could exist. The bar for "infrastructure problems" has dropped about a thousand-fold, but the founder mindset that builds anyway hasn't changed.

Q: Tell us about the first order. What did that feel like?

Asgar: The first website went live, and instead of pushing the website, we were a manufacturer, so we wanted to develop products. We were doing personalized champagne flutes, we bought the machines to engrave them. We were making lots and lots of money, enormous amounts of money, because nobody else had it.

Then the first order came in. It was a bottle of wine. The system delivered the first order and it was true e-commerce, full e-commerce. What we didn't have at that moment was you couldn't see the personalization. So people fill in the boxes and they just assume that we put the right information in the right place. We shipped the first product out, rang the customer up and said what do you think.

What was interesting, when the internet came along and you wanted to contact somebody, you'd say "could I have your business email address?" There wasn't one for an individual. You'd have to ask for a company email, then email the company and say can you forward that to whoever. Today everybody's got an email. All those years ago, there was no direct email to a person.

Then we had to figure out how we were going to package it so it doesn't break in the post. We figured all that out after the first order came. We had to go to the post office to get the stamps to ship the order, because couriers existed but they didn't collect from the business yet.

🔥 ChaiNet's Hot Take: Notice the order of operations. Build the product, take the order, then figure out shipping. Modern founders try to build the entire stack before launching. Asgar's playbook still works. Get a paying customer first, the boring problems get solved very fast when money is on the table.

Q: You lost your rankings overnight to a Google Penguin update. What happened?

Asgar: We were on position one for personalized wine, personalized champagne, position one. At that time, Google kept saying you've got to get backlinks, you've got to get thousands and thousands of backlinks. So we focused on getting backlinks. Then they said yeah, but the quality of the backlinks is very poor. Yeah, but you said get backlinks.

I think it was October time, I can't remember the exact date. It literally killed the business. Your traffic was up there, and the next day your traffic went down to zero. It was a dramatic step change because everybody is following the path Google said. Get backlinks, get backlinks. Even today, if you're relying on Google for your business, it wants to give the customer the best experience it can. That's what Google's there for.

But the marketplace is changing completely in terms of how internet search is done. SEO is important, but SEO is just a baseline, and everything else comes on top of that now if you want to be found on search through AI.

🔥 ChaiNet's Hot Take: The platforms tell you what to optimize for. Then they change the rules without warning. Asgar built backlinks because Google said so, and Google nuked him for it. If you're building on rented land, expect the rent to change. The only real moat is direct customer relationships.

Q: Customer expectations have completely changed. How do you keep up when you're a personalization business that physically can't ship next day?

Asgar: If you take Amazon as a giant, you place an order and you expect it to turn up next day. All e-com businesses, unless you have that level of service, it doesn't work. People expect next day delivery as standard, but we're in the personalization business and honestly, we can't do next day on certain products.

We have to physically make the products. The skilled workforce has to make them. We do have a next day option, there's a premium for it in terms of delivery. The cost is the same, but there's a cost for delivery time. We don't make any money out of it. It's just that if a customer wants it, they can have it on the next day service.

18 years ago, people got it a week later and they'd go yay, I've got a parcel. They were more than happy to wait. But today, will customers wait? Everything's now. Amazon is driving that change in the marketplace.

And it's the dopamine rush. You sit at the dinner table and people are scrolling their phones, your attention span has narrowed. Even if your website doesn't load instantly, people will scroll away. We had to make massive investments once mobile started coming in. 80% of our traffic comes through mobile now. 18 years ago, 100% came in from the desktop. So if you're not optimized for mobile, it ain't gonna work.

🔥 ChaiNet's Hot Take: Amazon didn't just change shipping speeds. It rewired customer psychology across every category. Even a personalization business has to bend to next-day expectations. The lesson is brutal: your category's standards are set by the giant in the room, not by what's reasonable for your craft.

Q: You've moved a lot of your AI work to a local server. Why?

Asgar: Three and a half years ago we were one of the first people to start using ChatGPT. When it first came out, it hallucinated, but for us as a workflow, we use it constantly now. For ideation, for descriptions, for image generation. We had a photographer and videographer in our business. He left about 18 months ago and we haven't replaced him because we can now do 90% of the stuff he was doing through AI. And faster.

But because of GDPR and security concerns, one of the products we developed was caricatures. Customers can upload their image and our system generates a caricature. About two and a half years ago, you used to have eight fingers or eyes in the wrong place. That has gotten better, but we still don't have the confidence of using the large language models out there to give consistency.

So we decided to develop a local server. We took a base model that can do images, we've been training through ComfyUI and LoRA to manipulate the image to the standard we want. Now we can produce a caricature consistently of any face, a child, a female, ethnicity. And we're just about to download it onto our own servers, so we control the whole ecosystem. We control all the data, so we don't have a problem with GDPR.

The other thing, we can actually rent that particular technology out to other e-com businesses. So they can do caricatures over it. We started with F1, then chefs, and cricket is unbelievable. The language models are not trained on, can I say brown faces? You don't get a great likeness. But because we trained it on our own system using LoRA fine tuning, we make great caricatures of Indian people. Cricket is perfect.

🔥 ChaiNet's Hot Take: Big language models are trained for the average user. Your business isn't average. Asgar identified a gap, brown faces in caricature generation, and is now planning to rent his fine-tuned model to other e-com businesses. The next moat in AI isn't model size, it's your fine tunes on data nobody else has.

Q: There's a great story about your mom and Gujarati and AI. Tell us about it.

Asgar: My mum makes absolutely delicious cakes, but her recipes aren't written down. I said mum, I'm going to write your recipes down, but she was never going to write it down. So I said, OK mum, let's make a cake. She speaks fluent Gujarati. Default tongue is Gujarati. She got all the ingredients out, and I put a chat bot on the side. I said listen to this in Gujarati and translate for me.

Mum said this is a teaspoon of this and this. The AI was listening in the back. I told her if you need to ask questions, ask in Gujarati. Mum made the whole cake, and dad got involved with pushing the stuff in the oven. Mum was saying get out of my kitchen and all the things that happen in domestic kitchens.

In the end I said, I want to take all the information my mum's given to the bar, can you translate it into Gujarati so she can read it? And it translated and said, a bit of this, two spoons of that. I said mum, can you read that? And mum read it and said that's spot on. I then took that recipe and did it in English and put it on Reddit. People emailed me on Reddit asking can I have your mum's recipe?

I tell people, start using AI and the answers will come. Don't think about productivity, don't think about this or that. Just use it. By using it, all these different applications come up. The caricature business came about because we just started using it.

🔥 ChaiNet's Hot Take: Everyone on LinkedIn talks about AI productivity gains. Asgar talks about getting his mum's secret recipes captured before they disappear. The most powerful AI use cases aren't going to come from corporate dashboards. They're coming from people just using it for actual life, then realizing what they've built.

Q: Walk us through how you optimize for AI Overviews now.

Asgar: OpenAI signed a contract with Reddit, and Google takes a lot of signals from human-moderated places. Why is Reddit data so good? Because your answers and questions are driven by the community. So if you talk nonsense, when I was talking AI art on Reddit, I got absolutely shredded. People had strong views, and that's great.

Take Quora. Quora is a really hard platform. We got banned a couple of times because they said we talk about our business. So now on Quora I don't talk about my business. I talk about my engineering experience, the jet engines I worked on, the Phantom, the Victor, the Jet Provost. That's experience I physically have, precision engineering. So when I say, well, I make watches and I use the same skill set, I don't say I make watches. I say when we look at watches, we're looking from a light of precision engineering, because I understand that technology.

You become an expert on Quora. Where does Google get this information from? Quora, Reddit, those kind of places. LinkedIn, Instagram, YouTube, TikTok. They've all signed agreements. Now when AI overviews come in, it looks at all of those signals. Because they're human moderated.

You as a business have to do all of those things to become an expert. Then there are also reviews. Trust pilot reviews become very important because that's a trust factor. People don't want a list of sites. They want the most trusted source. AI takes them there.

And here's a question. I remember the Penguin update killed my business. The focus today is on AI overviews. If they change the goalposts, which they can because they're in control of it, what happens to your business? So if you're in this industry, the goalpost will keep changing. What I say to people is, invest in yourself. As technology moves, you're moving with it.

🔥 ChaiNet's Hot Take: AI overviews aren't a new SEO tactic, they're a new game entirely. Backlinks are out, expertise signals from human-moderated platforms are in. Build authority on Reddit, Quora, LinkedIn, not just your website. And remember Asgar's warning: whatever the goalposts are today, they're moving in five years.

Q: Will the future of gifting be AI-generated art or human-crafted products?

Asgar: There's an opportunity for both. I personally believe human creativity is there because everything that's out there in AI is taught by humans. AI cannot think creatively, it can give you ideas, but those are ideas taken from different places. So there will be an opportunity for both.

For example, on our website we have bird flower watches, daffodils, lilies. The artwork is generated using AI. I could have got somebody to draw me these, but how long would that have taken? AI scales it up and gives you new ideas. You can be honest about it, this is AI, this is human. Why not give the consumer a choice? But if you want scalability, how many caricature images is a human going to create in an hour?

I had this conversation on Reddit and it was unbelievable how strong the views were. I'm saying look, I'm not against creativity. AI scales it up and gives you new ideas. Can you imagine if somebody had drawn that for you for hours and hours, you'd say do you know what, change this, change that. There's a trade-off in cost.

🔥 ChaiNet's Hot Take: It's not AI versus human, it's AI plus human. The boring framing is generation versus craft. The interesting framing is what does each do best, and how do you build a workflow that uses both. Asgar isn't replacing artists, he's scaling output by 100x and being honest about which is which.

Q: In 18 years of e-commerce, what has stayed the same?

Asgar: Nothing. Everything's changed. If you take a customer, when I first got the first order, they were happy to wait a week, 10 days. Now they want it now. If you look at platforms, social media, people want great storytelling for your brand to show up. That's why we're making YouTube videos and other bits and pieces. You have to tell a story with every single piece of stuff that you do.

I remember when we did our first YouTube videos, it was about making little cocktail glasses with personalization. We got great views, but the quality was very poor. Today nobody would watch that, no way. So people's experience on social has changed dramatically. Your experience of the website has changed completely too.

Even the route to entry into e-com has changed. With AI, you can develop a website just like that. You can have a website running in less than 24 hours, fully customized, with e-com stuff in the background. You can become a drop shipper. So in order for you to be in the marketplace, you have to develop great products or keep developing new products. That's where our business is changing to the world of AI.

🔥 ChaiNet's Hot Take: When the cost to launch a store drops to zero, the cost to stand out goes through the roof. AI has democratized the storefront. Now the only moat is product, story, and consistent iteration. Every founder building today is competing not just with brands, but with anyone who can fire up a Lovable site overnight.

Q: People talk a lot about energy use in AI. How does that fit into your local server decision?

Asgar: We do believe in green. We don't talk about it, we actually preach green. We've got solar panels in our business, LED lights. There's an interesting bit Sam Altman said, how much energy they use up in the LLMs when people say thank you and please.

Look at the energy that we're burning. Even our simple question is going through all that churning. Does it need it? No. So that's one of the reasons why, from an energy efficient point of view, having something on your local server makes a lot of sense.

You watch in the next probably 12 months, people will be going down the route to local servers because they'll have much more control in their businesses, and all the data belongs to you. If your data is in America on an American server, the American government can access it. In Britain, they can't. People are very sensitive about data, so having stuff on your local server makes complete sense.

But we still need the large models. When we were training the LoRA models on ComfyUI, to get the prompting right we had to use the large language models. Once we figured that out, we didn't need it. So there will be a need for general purpose models, but you don't need everything that goes with it to run a business.

🔥 ChaiNet's Hot Take: Local-first AI isn't just a privacy play, it's an energy play and a sovereignty play. Asgar's logic is straightforward. Use the big models to figure out your prompts, then move the production workload to your own metal. Cheaper, greener, and the data never leaves your country.

Q: What about hallucinations? You've been burned, right?

Asgar: I'll give you an example. I was running something through ChatGPT. I like to fact check across different models, so I put it on perplexity or Gemini. He ran it and said I've checked the facts through ChatGPT and I agree with the conclusion. Then I said, but you can't talk to ChatGPT. And it literally said, I'm sorry but I lied. Those are not the exact words but I've got the transcript. It just made it up.

How many times does it make things up? Whereas if you have control of your own system, the chances of hallucination are very small and your teams will pick it up because they understand your business. I'm not saying don't use LLMs, I'm saying use it but you need the human interface. Got to fact check it.

🔥 ChaiNet's Hot Take: "I'm sorry but I lied." Those are five of the most important words for any founder using AI in 2026. Trust the machine to draft, never trust it to verify itself. Always have a human in the loop, especially when AI claims to have done something it didn't actually do.

Q: Final advice for founders building right now?

Asgar: Don't forget my generation. We saw things progressing from valves to transistors, the first moon landing in 1969, then the first computers came out and we had to learn about them. Then the internet, then mobile, now AI. My generation has seen change.

I remember when fax machines came along, people would say I'll just fax it to you. The faxes became automated and you got reams of fax coming through, thermal rolls. Now I'll just email it to you. If you look at your email box now, as consumers we're getting overload with information.

I think the key thing is this. If you're in this industry, any industry today, the goalposts will keep changing. What I say to people, you have to invest in yourself in order to be right at the forefront. So if you were investing money, I would say invest in yourself, because as technology moves, you're moving with it.

🔥 ChaiNet's Hot Take: The single best investment in 2026 isn't a SaaS subscription or a course. It's whatever it takes to keep you adaptive. Asgar has lived through every commerce shift since the moon landing. The pattern is clear: the people who keep learning win, and the ones who got comfortable got replaced.

Final Thoughts: 18 Years, Five Eras of Tech, One Thing That Worked

Asgar's parting wisdom: "If you're in this industry today, the goalpost will keep changing. Invest in yourself. As technology moves, you're moving with it."

The bottom line: Asgar's story is the entire arc of internet commerce told through one founder. Satellite dish broadband. The Penguin update that killed his rankings. The shift from desktop to 80% mobile. ChatGPT showing up and slowly replacing his videographer. Local AI servers becoming a real option. Each transition required relearning from scratch. The founders who didn't keep up didn't make it.

What's striking is how he treats AI. Not as a threat, not as hype, but as just another tool in a long line of tools. He's training his own LoRA models because the big ones don't get brown faces right. He's planning to rent that fine tune to other e-commerce businesses. He's already thinking about what happens when AI overviews change the rules in five years.

If there's one thing to take from this conversation, it's the layered approach. Use big models for ideation and prompt engineering. Move production workloads to local servers for cost, control, and GDPR. Build expertise on human-moderated platforms like Reddit and Quora because that's where AI overviews pull authority signals from. And always, always, keep a human in the loop to catch the moment your AI lies to you and tells you it didn't.

The next decade of commerce will look nothing like the last. Asgar has been here before. The pattern keeps repeating, and the founders who survive are the ones who keep investing in themselves.

Q: How can people connect with you and learn more about GiftsOnline4U?

Asgar: I'm pretty active on LinkedIn, sharing what we're building day to day. If anyone's interested in what we're doing with caricatures, with our local server setup, or just wants to chat about how AI is changing personalized gifting, do reach out. We're also putting up YouTube videos now talking about why people buy gifts, the memory in time concept, and the engineering behind making something like a personalized watch.

Final words: Asgar's story is a reminder that the best founders aren't the ones with the cleanest playbooks, they're the ones who keep showing up across every era. Satellite dishes, frames-based websites, Penguin updates, mobile commerce, AI overviews, local LoRA training. All in one career. The technology keeps changing. The willingness to learn is what makes a builder. If you're staring at AI right now feeling overwhelmed, take it from someone who survived dial-up to local LLMs: just start using it, the use cases find you.

---