Why a Five-Year Founder Chose to Build

Last week I had a coffee chat with Jeong Jaeho, a founder, and heard their take on AX (AI transformation). From how to pick a market to how to win over the owners of legacy businesses, these were lines a fifth-year founder had worked out by running into things headfirst, and they stayed with me. I want to start by writing down the two that struck me most.

1. Avoid markets your client could monopolize

Say your client is a rental car company. Even if you do AX really well and push their revenue and operations as far as they can go, that company isn't going to grow a hundredfold next year. They can't suddenly buy a hundred times more cars or a hundred times more land. Lodging is the same. No matter how well they do, the structure doesn't let them take the whole market.

That's actually a good thing. If you get results, you can keep working with that client, and above all you can copy the playbook you built there and paste it into other companies in the same domain to expand. FOMO kicks in on its own too: everyone else is doing it, so you lose out if you don't.

Now take an industry where the fight is won on brainpower. A law firm, for example. Say you help that firm like crazy, it becomes number one in Korea, and because of that every competing firm collapses. That's the worst-case scenario. You're left with exactly one client, and you're completely at its mercy. You might keep the relationship going, but being the weaker party in a market with no competitors is miserable.

Even within manufacturing, things differ a bit. Markets where it's relatively easy to ramp up production, like cosmetics or gim (Korean seasoned seaweed), might actually be closer to the law firm case, since the number one player can take the whole market. Of course, that's not to say becoming number one is easy.

2. Appeal to the owner's childhood dream

Owners of legacy businesses are often more tightly bonded than you'd think with employees who've been through thick and thin with them for decades. They're practically family. Pitching AX to save a few million or tens of millions of won a month, from the owner's side, means asking them to cut family loose. Ending relationships with people you've spent your whole life with, just to save that money, is too painful.

So the idea came up to appeal to the owner's childhood dream instead. "Why not let AI take some of the load off your current business, and go chase the dream you never got to pursue?" A new business has no tangled interests, and they already have a core business that runs well, so they're much more likely to open their wallet.

For me this was a pretty new way of thinking, and a line I wanted to test myself.

That coffee chat was so good that as soon as the weekend was over, I met them again on Monday. Jeong Jaeho of ORCAR Studio, the fourth guest on the Bloom podcast, runs a service that rents cars to foreigners as the main business, and also provides rental car companies with the customer support, identity verification and payment know-how they've built up along the way. On camera, I asked again where those two ideas came from and what five years came before them.

"I started a company because I was lonely"

When I asked why they started a company, they said they'd come to two conclusions on their own, and one of them was loneliness. There were people they liked, people who weren't bad at heart, who were fun to talk to and who could help each other grow, and they wanted to get close to them. It wasn't working.

Before I went into the military I kind of lived like trash. I'd spend like 1.5 million won a month at PC bangs.

PC bangs are Korean internet cafes, and they added with a laugh that if you game for 24 hours and eat there too, you can spend that much. During that time they often felt they were falling way behind the people around them. They kept asking themself, "How can I talk with those people at the same level and keep good relationships with them for a long time?" and the answer they came up with was starting a company. When you start a company together, you're stuck together all day, and the relationship becomes more than just friends.

The idea was rental cars from the start. Their family started a rental car business on Jeju Island almost 30 years ago, and they grew up watching it. After finishing military service, COVID kept them from going back to school, and while working part-time at a car wash on Jeju, two thoughts hit at the same time: "I need to start a company," and "I want to try this field." I asked whether that meant competing with their parents. "They're a client," they answered right away.

What they saw within a week on Jeju

They had been providing rental car companies with almost everything needed to run the business: ERP, their own online booking sites, CRM, vehicle data analysis, even used car sales. At some point it started to feel stifling. No matter how good the technology was, the organization of a rental car company didn't change much, and they felt technology could only solve a small part of the whole.

That was last spring, when Y Combinator was talking about building AI-native companies to wipe out middlemen. So the whole team moved down to Jeju, leased 100 rental cars for 100 million won a month, and ran them themselves. If revenue beat that, they kept the difference. If not, they took the loss.

Once we actually went and did it, there really wasn't much for AI to do. Not for Korean customers, anyway.

Korean customers had tons of information online and plenty of experience already, so they didn't really need help. Nothing was inconvenient for them. But within a week of arriving, they saw a different scene. Foreigners were showing up to rent cars and getting turned away.

There was no staff on site who spoke English or Chinese, and the foreigners had somehow made reservations on websites with a translator running. Neither side knew which international driving permits were accepted. For the rental car companies, if there was an accident they couldn't handle the communication, and if a traffic fine came in later, it was hard to collect payment. So rather than take one booking and risk losing money, they just stopped accepting reservations from foreigners.

As a result, only a tiny handful of companies supplied cars to foreigners, and prices were five to ten times what Koreans paid. In Jeju's off-season, a car that rents for 20,000 won a day was costing foreigners 100,000 to 250,000 won for a Kia Morning, one of the smallest cars on the road. That same day, they posted in a Facebook group asking if anyone needed someone to book a rental car for them. Five people reached out, and from that they decided this business would work.

They went in by taking the client's risk to zero

For a company that never took foreign customers, taking them on is a business expansion. There's a lot to learn, a lot of risk to absorb, and early demand is hard to secure. Of all that, risk is what makes them most hesitant to start. Hiring staff who speak foreign languages costs a baseline of 10 million won a month, and you have to build systems to handle accidents and fines. If you make that investment up front and the demand isn't there, hundreds of millions of won a year just vanish.

So they reframed the offer: "You don't have to hire anyone. We'll handle it ourselves first. When customers come, we'll deal with all of them, verify their identity and send them over. If there's an accident, our company pays." I asked if they'd basically gone in like an insurance company. They said yes, and added that from the client's side, there was no longer any reason not to do it.

Of course, at first they lost money on labor costs alone. But as bookings came in one by one, the numbers started adding up, and they could see the structure would leave plenty of profit. They've kept pushing ever since. Today, ORCAR Studio's position in the foreign customer market is "service better than the big companies, at a lower price than the big companies," and they're especially well known in Singapore.

Bed at 2 a.m., work at 6

When I asked when they started using AI in earnest, the first thing they said was that until early this year, they didn't even have time to think about doing something with AI. About a year and a half ago they had only six months of runway left. If they didn't make revenue right away, everything would be gone, and every day the only question was "How do we make another 10,000 or 20,000 won today?"

Along the way, AI didn't do much to raise revenue. What was enormous was the operational load of serving foreign customers. Ten people woke up at 6 a.m. and ran operations until 10 p.m. Because of time differences they had to answer customers in the middle of the night too. At night they fixed whatever was broken in the product, went to bed at 2 or 3 a.m., and were back at work at 6.

Last summer, while holding on through all that, they spent two months automating a lot of the foreign customer support. They had AI handle about half of the inquiries, added self check-in, and used OCR to verify identity. Even then, they said, they were using AI for operations. They hadn't gotten as far as thinking "we'll make revenue with AI."

Others were calling agents while we were still writing documents

The turning point was this January. They described it as "AI FOMO" hitting the company. Other teams were calling up agents in their Slack channels to get everything done, and they were still writing documents. So they decided internally to try turning into an AI-native company. They built a product called "Orca Brain," a kind of company brain, and practiced a lot at getting agents to do work inside the company.

They also went back over things they'd already automated. With AI support or OCR, if the success rate stalls at 50 or 90 percent, operations still take a lot of work. Internally they pushed chat up to 80 percent and OCR up to 99 percent. Once they got there, it became "a product worth selling," and they started doing B2B sales again.

Along the way, they had more than 30 coffee chats in one month. They half agreed with the idea that we're in "a world where AI can do anything," and half thought it was an illusion. That world had arrived, and yet they had only ever moved within the rental car market. They were aware of how narrow their view was. If they couldn't explore what opportunities other industries held, they thought both they and the company would fall behind.

After those 30 conversations, their thinking changed again.

Won't the formula still hold, that the company that can give customers the service they really need, when they need it, at the lowest price, wins?

A company working as an AI agency still has to raise its clients' revenue and cut their costs, and still has to keep its own resources going into that to a minimum. The formula holds just the same. Right now, they said, they're thinking about how to achieve it.

What running rental cars themselves taught them

The first of the two ideas I wrote down at the top came from this part.

Paradoxically, if it's an industry where AI can do everything, it seems a bit harder to go the long haul with your clients.

Their evidence was what they saw running rental cars on Jeju last year: AI's impact on the industry only went so far. To run a rental car business you buy land, own hundreds of cars, hire dozens of employees and keep hundreds of millions of won in cash moving, and AI can't do any of that. So even when a client they help gets results, doubling the company because revenue doubled means buying twice the land and twice the cars. The risk is too big, so most stop at improving their operating margin.

In B2B, ending up with only one client is honestly the worst. The power dynamic gets way too lopsided.

How deeply they're embedded with clients is unusual too. With rental car companies, they're in at the level of "we'll run this part of the business for you," so for the foreign customer side, the rental car company doesn't have to do anything.

Honestly, all they have to do is bring the car where it's needed.

That's because ORCAR Studio handles everything from marketing to settlement, risk management and customer service. Lately they're expanding to clients in other industries like clothing rental and education. They don't have domain know-how there yet, so they go in with two weapons: running their own business, they ended up using AI a bit better than those companies, and coding costs have dropped dramatically compared with traditional IT outsourcing.

"If I were doing B2C, I'd do games"

I also asked why they expanded from B2C into B2B. When they tried again to grow the business, the bottleneck turned out to be offline, more than they'd expected: meeting people, persuading companies, adding suppliers. To offer rental cars to foreigners, you first have to know which rental car companies can rent to foreigners. If none can, you have to start by persuading a company. If even that fails, you have to set up a company yourself.

On top of that, they really wanted to make good use of AI, but they couldn't see much of a B2C business AI could do. They felt the only working business model AI has produced in B2C so far is character chat. Flip the perspective and look at what AI is producing, and it's driving coding costs to the floor and producing work that used to take about five years of expertise: design, marketing, data analysis and consulting. The place to make the most of that was naturally B2B.

They also think the moat of expertise that B2B used to require has completely collapsed. If you're good at sales, one person can now do development, marketing, branding and consulting, so there are more opportunities in B2B. On the other hand, if they were going to do B2C, they'd do "games," because games are close to the only ideal B2C business with no offline bottleneck.

The resources went down to one-hundredth, but the return is still around 30 to 40 percent, so it's like a game where efficiency got about 30 times better.

When the cost of supply drops to almost nothing, supply goes up while demand stays fixed, so things will balance out eventually. But right now, by their math, "it's a moment where the market gets 10,000 games, 100 of them succeed, and I can add 30,000 more." Soon it'll be 100 million and won't mean anything, they said, but for now it's worth doing.

In fact, four of them recently entered a hackathon, starting the day before. Jaeho only had 30 minutes, so they submitted a Hearthstone clone. Their co-founder made 12 games and submitted 2 of them, and one placed fifth. Of the 40 people who made the finals, they were the only team not working at a game company or game agency.

Honestly, we didn't make any of it. Codex wrote all of it.

Their conclusion was that if you can compete on similar ground with people who've been making games well for a long time, it's worth a shot. But they see this window of opportunity as lasting six months to a year. So in B2B, the only way to last is to lock in your clients properly, lock in several of them across different fields, and make those companies trust only you.

"I want to own a planet"

When I asked which part of my Brunch writing had stuck with them, they said "the view on nations." Being born in Korea and carrying the tag "Korean" doesn't make it feel quite natural to them to say "let's do work only for Korea, let's make Korea stronger." Regardless of all that, they said, isn't it right to think on the scale of the whole planet, whatever country you were born in? Even if you were born in the US, they think it's best to work in whatever form contributes most to the world, among what the government wants, what companies want and what people want.

When I asked where that thinking came from, a personal dream came up. These days, when people ask about their dream, they say "I'd like to own about one planet."

The size isn't really the point. I think what I hate is being in a situation where I have to go along with organizational rules I can't understand and find unreasonable, or with the view most people hold.

In every country, there are lots of people in organizations and in the nation who think differently, and often there are moments when that direction doesn't seem to help overall productivity. So if what they think is right, and what the people they like think is right, doesn't harm anyone else, they want to be in a position to decide that freely.

When the conversation turned to the moment they find most frustrating, the answer was short.

I see lots of people around me trying something out of a pure heart, and then I see others saying, "Why is that guy putting himself out there like that? Why is he showing off?"

In Korea the formula for success is relatively fixed, so when there's a path everyone agrees "works if you do it this way," taking a different one easily draws looks of "why are you doing that?" In places where there are many ways to succeed, it's hard to even say that the path you're on is wrong.

The conclusion this 29-year-old founder, five years into running a company, drew from the very physical market of rental cars came down to how far you can stand from the formula others have set. And it seemed to me they plan to keep doing it the startup way for a while yet: stepping off that formula and building the gravity of a planet of their own.

Originally written in Korean on Brunch, September 22, 2026.