Can AI Help You Run a Café?

I recently caught up by phone with two people: an acquaintance I've known for years who has spent three years building a product at their own startup, and a friend posted overseas by their company. We hadn't set a topic, but as we talked, the conversation naturally turned to business in the age of AI.

What struck me again was how much of the world AI still hasn't touched. On the surface, it looks like AI has already turned everything upside down. In the real world, though, it may have only put a small crack in things. It was a reality check, and I think I felt it more because both of them had worked in tech.

The founder's own experience beats everything

Kibaek Ryu, who founded Fountain, an HR service for blue-collar jobs in the US that reached a valuation of about 1 trillion won (roughly $700 million), left the company and recently started an "AI venture studio." Put simply, it launches a new AI service every week, the way a K-pop agency keeps debuting new idol groups. Some will flop, and some might be huge hits.

A YC alum, he said he felt the old approach of pouring years into a single piece of software no longer works in the age of AI, and that's why he's trying something new.

My acquaintance had also spent three years polishing one product with a craftsman's care. Then one day their parents were looking for a place in Seoul and working with a real estate agent. The agent made a mistake, and while sorting it out together, my acquaintance stumbled onto a business idea they never expected. They did a few user interviews on the spot, worked all weekend, built the service, delivered it, and made their first revenue right away.

That's how they built a service for three B2B clients in just two weeks. Product planning, development, design and sales, all of it. Their previous business had been growing only so-so despite a lot of marketing, so seeing money come in from a completely unexpected field made them wonder whether to move all six remaining team members onto the new idea. Ironically, that's exactly where problems started.

I put the whole team on it. But since they had no firsthand experience, what came out was limited. It stopped at benchmarking American services.

When I asked what the takeaway was, the answer was one sentence: the founder's own experience beats everything. This idea didn't come out of a meeting either. It came from watching a process up close during a personal errand.

Digging into the industry brought another shift. None of the proposals they had worked hard to prepare sold. Instead, the person across the table brought up a different problem first. They even said, "Solve this and I'll pay you every month." So the team built it, brought it back, and they really did start paying every month.

My acquaintance said they were a little stunned by that part. Nobody paid for the ideas they'd squeezed out over weeks locked in a meeting room, but one sentence they overheard while walking around on the ground turned straight into a contract.

There's no moat in building anymore

I asked whether they could now just tell AI to build things on its own. My acquaintance said the order is completely the other way around. These days the bottleneck is less on the building side and more on deciding which of the countless outputs to pick.

If you find the problem, building it is a snap.

So they decided to spend less time in meeting rooms and more time out in the field. One detail stuck with me. They built the first version in two weeks but told the client it took three months. When I asked why, they said that if you say you built it too fast, people may take it less seriously.

On the flip side, being able to build that fast means others can copy it just as fast. So my acquaintance made it a fixed assumption that "if someone comes in, they can build something similar in a day," and decided to focus only on how much faster and wider they could dig in. They've completely shifted direction and are now going all in on sales and expansion rather than building product.

A few days after the call, I met them in person and heard more of the backstory. What I was most curious about was those two weeks. The answer wasn't that AI was fast. It was that they had built up a lot before. They reused the design components from their previous product as is, so they barely had to draw any new screens.

The faster building gets, the more your foundation from before shows. Give two teams the same tools, and the one with assets it already built starts from a different place than the one without.

A new problem came up too. A developer on the team started being unable to explain the code they had written, because the team had come to lean on AI too much. My acquaintance saw this as a domain problem rather than a development problem. So they wrote up a Notion doc on the industry's terms and procedures and had the team study it. They chose to make the team understand the business instead of the code.

The way they work changed too. They run different models on the front end and back end at the same time, because one was better at design and the other at logic.

Instead of making software with AI, let's become a team that's good at using AI and try anything.

They added that raising money to grow bigger didn't seem to fit their team right now, so they shifted to being self-sustaining. Their judgment was that being a team that finds problems fast and builds and tests solutions fast beats being a company that makes products with AI features.

So many legacy industries still haven't even gone digital

Those businesses tend to be old and stable, they said, so they actually have plenty of money to pay.

In these industries, things we take for granted look almost like magic. My acquaintance's product is nothing special technically, but the clients have never used tools like it, so they're amazed just watching the screen move on its own. The exact same product can look completely ordinary in one market and like a brand-new innovation in another.

It's like someone who trained to debut as a K-pop idol, failed and got crushed, then moved to the trot market (an older style of Korean pop popular with middle-aged and older listeners) and immediately took off. Same skills. They just changed stages.

When I asked how they sell, the answer was door-to-door and flyers. They print flyers themselves, walk into offices, say they make software, ask if the person has a minute, and look for whoever's in charge.

The more interesting part was where the rejections came from. The staff in charge didn't welcome the tool, and the reason wasn't that the tool was bad. It was that it would shrink their work. For someone whose time spent working is their income, something that cuts that time isn't good news. So my acquaintance changed the sales approach and started meeting owners instead of staff.

Some things you only learn by being on the ground. Real estate offices on the first floor don't advertise, because walk-ins from people passing by are already their advertising. The ones that need ads are on the second floor. I had never seen in any report that the same business on a different floor has different customers.

When I asked whether the new products they build need to connect to their past products at all, my acquaintance said that's actually where they got a big boost of confidence. What they're doing now has nothing to do with what they did before, and yet it's all running well. Then they added something that stuck with me the most.

I pretended to like these productivity tools, but it turns out I get more excited doing this kind of thing. I realized that too.

I think it means that until you try a lot of things yourself, you don't even know what you really like. As a counterexample, my acquaintance told me about a founder whose first startup failed. One idea the team brainstormed happened to take off and bring in revenue, and the founder has been holding onto just that idea ever since. On the surface it looks like "good thing it's working." But looking at that founder, my acquaintance wondered whether they're only doing it because it makes money, and whether there's another idea they'd be better at and would have more fun with. Getting stuck in something that works hurts less than failing, but it definitely happens.

Throughout the call, my acquaintance kept saying the answers are in the field, because the ideas that never showed up at a desk only popped out when meeting people and running into things.

What running an offline business taught me the hard way

Lately I've been toying with opening a Bloom café in Seoul, something like Corgi café in San Francisco. A designer on our team tossed out the idea casually, and it naturally turned into, "If people are going to gather in a space anyway, what if we owned the space?" My acquaintance had just started a franchise café with their spouse as a side project, so I decided to ask in detail what that was like.

With offline, you don't know your revenue until you open. Not being able to predict it ahead of time, that's the hardest part.

With software, you can build a demo and land a contract first. Offline, you have to open the doors to find out, and you only get the answer after you've spent all the money. So the only option is to research as much as you can up front. They went out to count foot traffic themselves, standing in the same spot every day watching how many people passed by, and thought:

I just wanted to stand a robot there and have it measure the data for me.

Someone who runs their day on AI in software was out on the street, counting passersby by hand. That scene seemed to show pretty precisely where AI stands right now. AI has already gone deep into the digital world, but in the physical world, which makes up 99% of the reality we live in, it still doesn't really work.

I heard what happened after opening when I visited the shop in person. There's a saying in the industry: the "opening diet." You lose 6 kg (about 13 pounds) when you open a shop, because early on there are no systems and no staff, so the owner handles everything with their own body.

What stood out most about the business district was the sense of distance. Cross one street, and people stop coming. If they have no reason to cross, they don't, and creating that reason was everything.

Just get them to drink it once. Give them a "wow" on the first try.

It was no different from what we said in software.

Hearing how the franchise headquarters works made me think harder. Mega Coffee (a big Korean budget coffee chain) apparently has headquarters buy buildings, put franchise stores in them and collect rent. That's closer to a real estate company than a coffee company, the same model McDonald's has used for a long time.

My acquaintance also said reading a business district is the same as at a startup. Total coffee shop revenue for the whole neighborhood shows up in the data, and every shop splits that pie, so it's no different from market sizing. However good your concept is, you can't beat the market. The total cap is fixed, so checking that cap before going in is the same work they did at startups.

Then my acquaintance pointed to an empty spot. People burn 100 million won (about $70,000) to open a shop, yet owners have so little information that they decide by gut how much that money will bring back.

When you're a small business owner doing a franchise, you burn 100 million won. Spending a few hundred thousand won on top of that is nothing.

There are definitely people who would gladly pay for information that works like insurance, they said, and that space is still empty. But they added that making everything offline efficient isn't the answer. Some people deliberately avoid restaurants that use reservation apps, and some fine dining places with expensive service staff don't have a reservation system at all. For some people, that inconvenience is itself a sign that they're being treated well.

Neighborhood businesses are the same. Things like the owner's friendliness show up directly in sales, so some areas will always be hard to automate. I thought this kind of subtle sense is something you only develop through real experience, bumping into things with your body, not sitting at a desk.

Americans are more negative about AI, and Koreans are far more accepting

My friend working in the US for their company said the backlash against AI they feel there is much stronger than in Korea. Korea, on the other hand, tends to welcome any new trend at first. When I mentioned teaching AI to grandmas and grandpas at a local community center, my friend was surprised and said not many countries do that kind of thing at a national level.

The biggest change my friend felt after moving to the US was having room to breathe. There are so few "right answers" there. Even at the grocery store, everyone's cart looks different, and there's no unspoken rulebook for how things are done. In Korea, time always felt like it was flying by, they said, but not there.

Honestly, I really don't have worries. I used to have a lot. But since coming here, they've gone away.

When we traced the reason together, it came down to distance. In Korea there are lots of people your age around you, and you can see far too clearly what everyone is doing, so it's hard to get out of that headspace. In the US, that information is hard to see. The total amount of worry hadn't gone down. There was simply nobody to compare yourself to anymore. I felt like this might be the reason people working in Korea are so tired.

I don't follow AI trends anymore

My friend said something similar about their own side. These days a new model coming out doesn't really register. They just let it pass with "there'll be another one." But they haven't stopped caring entirely.

But if we're talking about how to make money, that I'm interested in.

We also talked about a friend who is an accountant and started their own firm. They don't know how to use Claude, so they don't use it at all, even though it could probably help them a lot right now. So the conversation naturally moved away from the latest models and toward an audience focused on who is making money, where and how.

Our conclusion was that you don't have to stay locked inside the tech startup world. There are already plenty of stories out there. People who aren't developers, PMs or C-level executives. For example, an older friend who took over their father's lettuce and perilla leaf farm, runs it as a smart farm, and uses AI like crazy. But my friend added a caveat.

"Made 100 million won with one click," "great at some AI thing," that stuff just annoys me now. I don't want to see it. There has to be some struggle.

When it's nothing but success stories, you actually want to drag them down. I nodded at that. People only really enjoy a story when they see the struggle behind it, and the moment someone says "I made easy money with one click," you lose interest.

AI does seem to be changing the world in a big way, but my wife always says this: the news says almost every day that AI will threaten humanity, yet nothing has changed in her daily life of raising our kid, doing the dishes and doing the laundry.

At one point I wanted to escape the painful burnout of working in IT, something you could call "AI Blue." But running around in the middle of that blue fire, I started to wonder. Beyond the overheating and fatigue of the AI industry we're in, how is the real world of ordinary people, far from the sense of crisis in the news, taking in this change? So these days, even while I throw myself around in the middle of the AI industry, my eyes keep drifting to real places like cafés, real estate offices and community centers.

Originally written in Korean on Brunch, September 21, 2026.