Fieldguide × Bloom: A Goldman-Backed US Unicorn Comes to Seoul

More than 170 people came to today's Bloom community event. It was raining, and I worried turnout would drop, but it may have been close to our biggest crowd ever. (With more than 350 sign-ups, we had no choice but to screen applicants.)

I think so many people showed up for two reasons: the latest Silicon Valley trends, and how to survive in the AI era. Those are probably what all of us care about.

Fieldguide is a five-year-old American unicorn headquartered in San Francisco and backed by Goldman Sachs, a16z and YC. It's a classic vertical SaaS company focused on a very niche domain: financial audits. With people talking about a "SaaS apocalypse" and frontier companies looking ready to swallow everything, Fieldguide is still putting up impressive results. Today its CEO, Jin, kindly flew in from the US to Korea. (He was also stopping in Japan, Singapore and other countries on the same trip.)

To match the theme, for the opening session we invited Kibeom Kim, who spent nine years as a software engineer at Google headquarters in the US and was CTO of the New York startup Terminal X. Their areas overlapped naturally, and putting them on the same stage turned out to be a great call.

20 events in three months, and Singapore

Last week we skipped an event. When I counted, we'd held more than 20 Bloom community events in three months. At the busiest point we were running three a week, full time, so it added up fast. I keep running these partly to find the hidden experts in the AI industry and introduce them to you, but honestly, what I'm most curious about is what's happening outside Korea. So whenever I get the chance, I want to bring in people from overseas and introduce them to you.

We took last week off because I was in Singapore. Bloom started in April with the Claude community, and we went to Singapore with the same strategy, teaming up with the local Claude community there. About 1,500 people signed up and around 200 came. What struck me was the response. I really wanted to know whether the simple story of moving from Blue to Bloom would work abroad, and whether people there felt the same way. Locally, I heard over and over that we shouldn't do roundtables, because Singaporean culture might make people shy about gathering to share how they really feel. I went ahead anyway and did it the way we do in Korea. The response was intense. That's when I became sure this is the same everywhere in the world.

That's also why we invited Fieldguide. It's great when we go to Silicon Valley to hear their stories, but it's even better when they come to us and tell them in person. And people are talking about a SaaS apocalypse these days. My day job is at Cliwant, an AI startup, and I'd been feeling pretty blue, wondering whether SaaS was going to die and whether any company besides Anthropic and OpenAI would survive. But there are definitely companies getting through this, and some are even using it to their advantage. What impressed me about Fieldguide is that it's only five years old. Companies under seven years old are usually called startups, so it's a very young company, yet it's raised from Goldman Sachs, reached Series C, and become a unicorn in the extremely niche domain of financial audits.

10x is easier than 10 percent

The opening session wasn't run by me. It was moderated by Yeryeon Seo, an AI engineer at our company. Kibeom is an engineer, so I figured there would be a lot of development talk and asked them to take it. It was only their second time, and they did much better than I would have, so I think I'll be handing it over more often. Haha.

Kibeom grew up in Korea, went to the US to study in his third year of college, finished his bachelor's and master's on the East Coast, then went straight to Google, where he worked for nine years. He worked on Chrome, Search, TensorFlow and Ads. The most meaningful was TensorFlow, and what he remembers most was setting up a team to prototype a next-generation API. That project was canceled, though. "So, everyone, you can stop waiting for TensorFlow 3. It's not coming." The room burst out laughing.

After that, New York Jumin (a Korean YouTuber whose name means "New York resident") asked him to co-found a company, and they started the fintech startup Terminal X. New York Jumin spent many years at hedge funds in New York and is now also well known as a YouTuber. "I felt strongly that if I was going to do a startup, it should be with someone like him. I'd never seen anyone work that hard in my life." The product they launched early last year took off, they closed a Series A this April, and now he's in Korea working on his second startup. His first employee was starting the day after the interview.

Yeryeon asked what was different about Silicon Valley's engineering culture compared to Korea's. That's what Koreans most want to know. We're plenty smart, so why do they get results hundreds of times bigger over there?

His answer was clear. There are a lot of great engineers there, but he doesn't think Korean engineers are behind at all. Average IQ isn't different either. Yet the results companies produce are worlds apart, and the reason, he said, comes down to mindset.

In Silicon Valley, people always dream big. Ridiculously, absurdly big.

He gave examples. When Internet Explorer held more than 90 percent of the market, it was far from obvious that a search company should build a browser. It was a huge challenge, and it worked. In the early days of mobile, putting that browser into Android wholesale wasn't obvious either. Neither was building TPUs, a dedicated chip, in the early 2010s when people were just starting to use GPUs. On a bigger scale, things like building electric cars are the same.

The most important part of the culture he saw at Google was what the founders told the team at the weekly all-hands. He believes that's where company culture gets set. What they stressed most was 10x thinking. He shared a scene he remembers. When the USB-C standard was being created, a Google team presented, and one of the things they were proud of was that charging was several times faster than the previous generation. Larry Page asked back, almost instantly:

Can't it be 100x?

Preparing for this talk, Kibeom looked it up: fast charging back then was 20 watts, now USB charges at up to 100 watts, and a 200-watt standard is coming next year.

He said there are strategic reasons for 10x thinking. It's easy to assume that aiming for 10x makes things ten times harder, but it actually flips. First, it's easier to attract talent. If one company says, "We're going to optimize this and make it 10 percent faster," and another says, "We're going to build something completely new and make it ten times better," or "We're going to Mars," or "We're going to build AGI," job seekers are naturally drawn to the second. Talent is more than half of any startup or company, and when you aim for 10x, the best people come to you.

Second, it really isn't that much harder. Compare Elon Musk building and succeeding with a gas car company to building and succeeding with Tesla. Relative to the return, the difference in difficulty wasn't that big. Combustion cars were a red ocean: lots of companies already squeezing the most out of a fixed form factor and fighting over 1 percent. Nobody was doing electric cars. Relatively few people think crazy thoughts, so there's less competition. Google's project to launch hundreds of balloons to bring internet access to Africa and the rest of the world failed in the end, but it was one of those attempts.

This applies directly to individuals and startups too, he said. If a company making 1 billion won (about $700,000) a year wants to make 200 million won more, it thinks about improving the product or getting more customers. But the moment you think you need to make ten times more, the direction you try changes completely. Maybe build a different product altogether, or expand to the US. The US market is about twenty times bigger, so those options start to show up. Individuals are the same. To grow your assets 10 percent, you can switch jobs or sell better. To grow them tenfold, that won't work, so you look for other ways.

Zero to one, over and over, and a product that sells

Leaving a big team to start Terminal X from scratch meant two years of going from zero to one, over and over. Asked what was hardest, he gave an unexpected answer. He'd been ready for ideas to fail. He didn't start with an idea he was sure of. He went in expecting to try many times and fail many times, so the failures themselves didn't hit him that hard personally.

The hard part was the team. When you fail, the team takes the hit. And after a failure, there are a lot of opinions about what to do next. When something succeeds, you can see what to do more of. When it fails, there are tens of thousands of things you could do next, and time and resources are limited. In fact, the team changed a lot with each failure, and losing teammates was the hardest part. In the end, a startup depends on the founder's will, he added, and in a sense, dying is also the founder's will.

Then Terminal X, launched early last year, took off from the start. A Japanese securities firm they signed right after launch expanded from 30 accounts to 100 a year later, and was happy enough with the product to invest in the Series A. Asked what was different, he named one thing. It was the exact product New York Jumin, the CEO, had wished existed back when he worked at a hedge fund.

What they'd built before was in the same finance domain, but none of it was something he'd actually use at work. One of the most common reasons startups fail is building something customers don't need, and when the founder is the customer, that friction drops to almost zero. You can set off in the right direction from day one. Of course, a great PM or CEO can explore things they don't use themselves through user research, but that's not easy, so starting from what you need and what you know best is a good approach.

Do open source, no matter what

Asked what developers should prepare first to compete in the global talent pool in the AI era, his answer was firm. Practical tips like how to apply would need a separate hour-long talk, so he'd skip that today. Before any of that, the best thing you can do from Korea is open-source work.

He gave three reasons. First, the biggest reason people want to go to the US or abroad is to work with those kinds of people, and with open source you can start doing that tonight. Your PRs get reviewed and merged by those people, and if you keep at it, you might become a maintainer. The reason you wanted to go abroad is already in there.

Second, you get better. You can sharpen your skills at work, but nothing beats getting reviewed by truly top engineers. Third, you get to know people, and that naturally leads to job opportunities. Someone he met two days earlier at a talk about getting hired in the US had done open source in college and became a maintainer. Without even applying, they got an interview offer through someone they knew, and got the job.

This was probably the most noted-down advice of the night. Whether you want a job abroad, a co-founder or a business, it's the most important skill for getting yourself known.

Choices he regrets, and choices he doesn't

The last question was whether he had a rule that helped him make career choices without regret. He answered honestly first.

I have a lot of regrets.

The things he regrets looking back are things he had no idea about at the time. No matter how good the team, how good the project or how decent the pay, if you're not much different after those few years than before, that time is a real waste. He was on the Chrome team for quite a while, and for the last two years he just kept doing what he'd been doing, comfortable in the codebase. Looking back later, he couldn't believe how much of a waste that time was.

What's left isn't the money, the team or the company. In the end, it's yourself.

He had the opposite experience too. After moving to the Search team, he applied to a product manager rotation program and worked as a PM for nine months. When he started, he thought it was career suicide. He was going to stay an engineer anyway, so that stretch would leave a gap in the profile he'd be judged on as an engineer and slow down his promotion. He figured it was a loss, and it actually was. But afterward, what stayed with him were the soft skills, the people, the hard experiences and empathy for the PM role. "Looking back, it was a great decision. I don't think you lose anything in life by continuing to challenge yourself and doing things you're not comfortable with."

We took one audience question: he could have gone to the US or anywhere, so why build a team in Korea? His answer was candid. Having run two offices, he said that for the same skill level, salaries differ by three to five times. Other Korean founders say the same thing. And he was confident that by drawing on his own career, he could hire really good people. "In the end it's about talent. Realistically, I hired better people than I could have hired in the US."

A $3 trillion market nobody knows about

The main session was Fieldguide. CEO Jin W. Chang, a Korean American whose Korean name is Jinwoo Chang, took the stage and immediately asked: raise your hand if you know what an accounting firm or an auditor does. Only a few hands went up. "That's somewhat by design."

The conditions for a Silicon Valley startup he laid out matched Kibeom's exactly. You usually need a secret insight of your own that spots a market or opportunity before others do. If something is already seen as a good idea, you're too late. "When we founded Fieldguide five years ago, it wasn't considered a good idea. Even my investors said it was a bad idea. Now, five years later, a lot of Silicon Valley investors think it's a good idea and are trying to fund a bunch of copycats."

His career started as a CPA at EY, one of the Big Four accounting firms. Then, doing an MBA at Berkeley, he looked for the startup he wanted to build, and that eventually became Fieldguide. The reason he started it was simple: he wanted to build the platform he wished he'd had.

He said the auditing profession has a branding problem. Nobody knows what auditors do, but we all consume their output. When you read an income statement, balance sheet and cash flow statement before investing in a stock, the reason you can trust those numbers is that an independent accounting firm verified them. Fieldguide's north star is to have every audit and accounting engagement running on its platform by 2030, and that market is worth about $3 trillion worldwide. It's a huge market, and nobody knows what these people do. Bringing massive efficiency to a forgotten industry was his second insight.

Another insight was the market trend. The market is big and the opportunity for accounting firms keeps growing, but the number of people doing this work is falling off a cliff. Simple math says each person has to handle more work. The problem was that the software in this industry was terrible. The competing products he put on screen weren't even cloud-based. They were on-premise and looked like Windows 95, and they came from big public companies like Thomson Reuters and Wolters Kluwer. On top of that, because there had been almost no innovation, small startups had attached themselves around that bad software one by one, fragmenting the stack and breaking up workflows.

So they built what he called the only modern, end-to-end platform that is both cloud native and AI native. They added AI tuned to the domain, designed it to close the staffing gap, and layered on AI assistants and agents that work like members of the accounting firm's team. Auditors end up with human teammates and AI agent teammates.

He shared numbers too. They've raised $125 million so far. The latest round was led by Goldman Sachs: $75 million at a $700 million valuation. Before that came Bessemer Venture Partners, which he called the best vertical investor, before that 8VC, founded by members of Palantir's founding crew, and it all started with YC. KPMG, one of the Big Four, and its Korean member firm Samjong are also investors. The company has about 250 people, and more than 50 of the top 100 accounting firms in the US are customers, including some of the Big Four.

Seoul as an AI product hub

Next, Junsoo Jang, the Korea hub lead and a senior ML engineer, introduced Fieldguide Seoul.

He started with something personal. Two or three years ago he suddenly left Korea for Silicon Valley to join Fieldguide's AI team. He emailed Jin and co-founder and CTO Chris: here's who I am, here are my goals, and I want to work at your company. He still remembers what Jin said on their last call. "We're Silicon Valley investors and founders, and like Silicon Valley people, we love betting on ambitious people like you. And I'm confident I'm good at it." He believed that, got on a plane, and has now brought Fieldguide to his own country.

He studied at POSTECH (Pohang University of Science and Technology), worked as a machine learning engineer, and also worked as a DRAM hardware engineer at SK hynix before joining Fieldguide's AI team as an early member and building several foundational AI capabilities. Now he's building an AI product team in Korea from scratch.

What he emphasized most was the heart of the night. Fieldguide Seoul is the company's first global office outside the US, and he wanted to be clear about one thing: it's a global AI product hub, and absolutely not a support office or satellite office that processes tickets for headquarters. Fieldguide designated Seoul as its own global AI product hub, which is why it opened its first global office here and is investing heavily.

It officially launched this May, it has two people right now, and the office is in Yeouido (Seoul's financial district). The goal is easy to say and hard to do: build a world-class AI product team in Korea. They set three milestones. First, assemble a very small but exceptional team of six founding engineers. Then, after proving success, expand with product designers and owners to take end-to-end product ownership. Long term, become Fieldguide's APAC hub. The business is already expanding into Canada, Europe, Singapore and Australia, so APAC customers will grow, and the plan is for Seoul to be the core base for that expansion.

People who join would build and scale production-level agentic AI systems, work on AI evaluation and quality (especially important in a regulated domain), and do platform engineering to support every feature. The challenge he pointed out stuck with me. "You all use Claude Code, right? Think of us as being in a Claude Code-like position in the audit and advisory domain. So there's no reference product at all to look at when we build." It's technically hard, and even they don't know if they're heading the right way, so they move with extremely fast iteration.

He was clear about who they're looking for. First, very high ownership and initiative: engineers who can take on everything from defining the problem to owning it and driving it end to end. Second, AI native. Writing code by hand is now a very inefficient thing to do. Third, people who get absorbed in hard problems. And finally, he stressed that they don't put much weight on years of experience or past titles on a résumé. They focus on how ambitious you are and how much you want to solve this problem.

He gave three reasons to join. First, from day one you work with world-class talent from MIT, Caltech, Stanford, Berkeley, POSTECH and KAIST. Second, the experience of building a frontier-level AI product. What he said here changed the air in the room. Before the AI era, up until four or five years ago, engineers used to say: if only I'd been born during the smartphone boom, I would have built an app like Uber and had a hit. "I think the next wave is obviously this." Third, global-scale impact. More than half of the top CPA firms in the US already use it and customers are spread around the world, so a PR you build and deploy today ships as a feature global customers use today.

Why a kid who grew up in LA came to Seoul

I moderated the fireside chat that followed. I'm not a developer, but listening up close got my heart racing. I've brought people from overseas to Korea many times, but actually building a team here in Korea is a completely different thing.

The first question was what coming to Korea meant to him. Jin said it was a personally meaningful moment. He had always wanted to tap into Korea's talent hub, but he expected it to be a much longer-term dream.

He was born in Korea, and when he was three months old his whole family immigrated to the US, to Los Angeles specifically. Growing up in the US, he knew he didn't look like the average American, so he always had a curiosity about Korean culture in the back of his mind. And LA has probably the largest Korean diaspora in the world. There's a joke that you can survive in LA without speaking English, and in fact his mother still doesn't speak English. He always heard Korean at home and grew up watching Korean shows.

So he has a lot of pride in being Korean. When he walks through a department store and sees Jennie from BLACKPINK as a Chanel ambassador and Jisoo as a Dior ambassador, he couldn't be prouder. That's also why he believes Koreans succeeding means a lot. Until recently Korea was a less powerful, less influential country, but now it's building global momentum. Yet Korea is a country of about 30 million people, and even counting Koreans abroad it isn't many, fewer than California, just one of the 50 US states. "So I've always felt it's an underrated market."

Then he said what completed the story of the night. He got better opportunities because his parents made the hard decision to move to the US, and now, through the platform that is Fieldguide, he can help a lot of Koreans dream of becoming Silicon Valley builders. A Korean-founded unicorn in Silicon Valley is fairly unusual, so he hopes Fieldguide can become a platform that connects Korea's talent hub with the Silicon Valley ecosystem.

Why he went back to accounting

Before Fieldguide, he worked at a legal AI startup. I wanted to know why he went back to accounting, a profession that had been painful for him. Meeting Chris, now his co-founder, there was a big part of it. They were already best friends, rock climbing together every day and then heading back to the office to work. After leaving that company, they talked every day about what company to build next. They knew they loved building AI for professionals, or machine learning as it was called back then, because they loved solving hard problems.

I believe the easy problems have all been solved. The era of IPOing with a simple product like Asana isn't coming back.

He wanted to avoid lawyers, though. They're extremely argumentative, and he didn't want to build for them. So outside of law, they looked for the next professional services field that was in serious pain and that both of them had experienced. He had been an auditor, and Chris had been on the client side going through audits like SOC 2. Everyone involved in audits was suffering, so solving problems there would be valuable. That was the hypothesis.

Another insight was timing. They started the company during the COVID lockdowns, when everything seemed to have stopped. When he said he was starting a company, his mother asked, can't you just get a normal job? What he told her stuck with me: the best companies have always been founded in times of chaos. Look at today's best companies, and almost all of them started within a year or two of a very hard period. If you do what most people want to do, you don't have a secret insight of your own.

And the last piece was a bit of an accident. Justin Kan, the founder of his previous company, who had watched the two of them for a long time, wrote the first check, and that's how the first investment came together.

Honestly, it just sort of happened. Only after we raised did I think, oh, this is really happening.

SaaS is dying, but

I also asked about the SaaS apocalypse talk that has spread like a fad in Korea. Korea moves fast on trends in everything. When SaaS is hot, everyone does SaaS, and now that everyone is going AI native, they're worried about whether it's still OK to build SaaS.

His answer was firm.

The SaaS apocalypse is real. Traditional SaaS will die. It's only a matter of time.

It's already priced into the public markets, and the same goes for the private markets. Silicon Valley investors don't invest in traditional SaaS anymore, he said.

But Fieldguide is a subscription model too, so I asked what's different. His answer was the most practical part of the night. The problem isn't the SaaS business model itself. It's the hypothesis behind the business. In a world where AI is possible, outcome-based and consumption-based pricing will grow, but that doesn't mean you can't do subscriptions. Fieldguide is a subscription model aligned to outcomes instead of seat count. Anthropic and OpenAI are both seat-based too, and they're doing very well. In the end, whether you do subscriptions or consumption-based pricing, the companies that align their pricing to customer outcomes will win.

Asked in a follow-up how they measure outcomes, he went back to an earlier slide. Think about the incentives in a market where demand for services is growing and the number of people is shrinking. If you charge per seat, you only make money when your customer hires more people. That gives you an incentive to slow customers down so they hire more. But in this industry there's nobody to hire, so that model doesn't work anymore.

"So here's our pitch to accounting firms: we'll help you grow without hiring more people. Hire our AI agents instead of people." Pricing is aligned to the customer's revenue, not their seats. When the customer's revenue grows, Fieldguide grows with it, and as they need more capacity, they hire more agents. They don't need to use every agent from the start either. That's how their pricing is aligned to customer success.

Goldman Sachs convinced us

I was curious about the Goldman Sachs investment too, since you don't often hear about companies getting funding from Goldman.

He said Goldman isn't a traditional VC, but it has been doing venture investing for decades. It used to focus on late-stage rounds right before an IPO, but for the last five years or so it has been investing a lot in somewhat earlier growth stages, because demand for pre-IPO assets has grown. Goldman has many private wealth clients who want access to pre-IPO companies, not just public markets.

He took their money for two reasons. First, the accounting industry is old school. It's a slow, old industry whose software looks like Windows 95, and he figured Goldman has the kind of brand these old accounting firms value. He's seeing that effect in practice now.

The second was more interesting. He has built relationships with almost all the top VC funds in Silicon Valley, and they're all very smart and ambitious and think crazy thoughts, because they have to. But they're not as analytical or research-driven. Goldman showed him all the research it had done on this industry over the past several years.

We didn't convince them. They convinced us.

Of course, he added, it's a big privilege.

Looking back, there was a reason for it. In the Series B, Goldman lost out to Bessemer, because at the time he wanted a more traditional Silicon Valley fund. After missing that round, Goldman built up its research, showed it had the best thesis on the industry, and offered access to every leader at their customer firms. Within a few months it had actually introduced several customers, and it keeps doing so. Even when it doesn't make an introduction, he added, the market already knows who Goldman Sachs is.

If you want the US, there's one answer

I asked the question Korean founders most want answered: if you start a company in Korea and want to expand to the US, what should you do first?

His answer was surprisingly simple.

The only way is to put your headquarters in San Francisco, in Silicon Valley. That's the only way.

The US is the biggest market in the world in dollar terms, so if you treat it as an afterthought, you won't get good results. He gave the same advice to a friend who started a company in India and wanted to break into the US market: if you're serious about the US market, you need a team in San Francisco. Founders need to be where other founders and investors are.

Go to a coffee shop and there's an investor. Cross the street and there's another one. You can't escape them.

Korea has an active investment scene too, but the San Francisco Bay Area is on another level, he said. So if a founder chooses not to be in Silicon Valley, they're choosing to limit themselves and give up the US market, and that's their call.

I asked about New York, Texas, Miami, Seattle, Singapore and London. They're all big, first-class markets, and Fieldguide plans to open offices there too. But the statistics aren't even close. The funding in San Francisco and the funding in New York, Austin or LA are in different leagues. If you're a fintech startup, though, he added, New York can make sense.

B2B sales, he said, depends on the role. Fieldguide's headquarters is in San Francisco, but the sales team is out in the field. Even the sales leader, the CRO, is officially based in San Francisco but is never in the San Francisco office, because enterprise sales means meeting customers face to face, and beyond meeting them, having lunch or dinner with them. Engineers, product managers and designers, on the other hand, come into the San Francisco office almost every day. He himself spends about 90 percent of his time traveling. He summed it up: founders should be among other founders, and sales should be among customers.

You can't build culture later

The last question was about culture. I wanted to know what someone who'd been through several companies made sure of first when founding his own.

For me, culture is probably the single most important thing.

Even with a vision and a strategy, if the culture isn't right, everything falls apart. What he stressed was the order. The moment you say, now that the company is up and running, let's define our values, it's already too late.

If you treat culture as something for later, you can never bring it in or design it. By then the culture is already set, and posting "our five values" or "our six values" on a board won't help.

So when he founded Fieldguide, before doing anything else, he wrote down the values. If they had to change later, fine, but they'd at least be on paper. And he says this again at every company offsite:

The people in this room create the culture for the people who come next.

When you're 10 people, those 10 create the culture for the next 10. When you're 50, those 50 create it for the next 50. That also means being unsentimental about people who don't fit the culture. People who don't fit should go to a company where they'll fit better.

He added something I liked. Startups aren't only hard when things aren't working. They're hard when things are working too. Landing your biggest customer ever looks like the company is doing well, but the moment a customer that size comes in, the product starts breaking. Every process that worked fine for the last six months breaks. Getting through all that pain together takes a strong culture.

And that talk of culture brought him back to Korea. A big reason he's betting on Korea, he said, is that he deeply trusts Junsoo. "I remember his first email because it was so long. I thought, anyone crazy enough to write an essay of an email like this won't fail. If I bet on this person, he'll work himself to death to prove it was the right call." It was a great decision, he said, and he believes building an APAC talent hub in Korea will be too.

Roundtable: which domain would you go after?

The highlight of every Bloom event is always the roundtable. That night there was one question: if it were you, which domain would you apply AI to?

One group came up with content, construction and insurance claims. Fieldguide succeeded by using LLMs to structure what was a tangled mess, they said, so they found common ground in the idea that these hard, complex domains could be cracked the same way. Their takeaway: with good people and a good environment, anything works.

A group with people from Kakao Mobility, Coupang, a chemistry-focused agent company and the Korea Atomic Energy Research Institute brought up areas that haven't been conquered yet: video editing, which takes a long time; chemistry, which has a lot of room; and market research for stocks and finance. They also saw data copyright becoming a bigger issue, and one idea was that people would happily open up their data if a service recorded all their work and analyzed which parts could be automated.

An AI product engineer on an independent development team laughed while presenting that they'd spent the whole time talking about their trip to the US two weeks earlier. They'd built a system for two-way simultaneous interpretation over landlines, still widely used in government offices and hospitals, with no app or extra hardware, and the paper was accepted at ACL, where they had just presented it. At the conference, companies like Apple and NVIDIA showed interest.

A product manager in the gaming vertical framed the choice of domain as money versus romance. They'd started out of romance, because making games is fun, while someone else in the group was on the side of business value: if AI could do CAD design accurately, that would be worth a lot. They had two takeaways. Going into a bigger market means more opportunities. And they asked themselves whether the reason they'd felt boxed in trying to do business in Korea was that they'd been following existing methods. After hearing about 10x thinking that night, they'd started wondering whether they should try something completely new.

The presentation from someone leading company-wide AX (AI transformation) at a medical device company was striking. They're getting rid of every SaaS tool at the company: CRM, ERP, the company-wide HR system, all of it. Others in the group said construction machinery looked likely to see a lot of replacement, and that older algorithms like Naver's shopping recommendations (Naver is Korea's biggest search portal) would change a lot as AI replaced them. At the end, they passed along that one group member wanted to use Fieldguide but had only seen videos, and asked to be let in. That got a laugh.

A group with someone who runs a reverse engineering community talked about the opposite: domains where AI is hard to apply. Ticketing looks simple, but it's closely tied to what happens on the ground and the data is unstructured, so it's hard. Games are hard too, because QA after development has unstructured workflows and complicated conditions for reproducing bugs. Their shared takeaway came from that night's talks: the most important thing might be to go in person to places where people, investors included, are packed close together, like Silicon Valley.

Someone doing a master's at UIUC brought up smart glasses. It's a very hot topic in the US but doesn't seem well known in Korea, they said, and Silicon Valley YC companies are launching lots of applications that use them. One company they'd applied to puts smart glasses on day laborers at industrial sites like HVAC jobs and sells the video data as training datasets for physical AI.

The last group brought up smart homes, agent safety and smart farms. Most were problems they'd felt themselves or that involved people they knew. The night's answer to which domain to choose was all right there. It was exactly what Kibeom said about the founder being the customer, and what Jin said about needing your own insight that you see before others do.

Wrapping up

After the event, I stayed for more than two hours listening to members. I'd tried to draw out as much as I could of what Koreans really want to know, like how to raise money in Silicon Valley and how to expand to the US, but the conversations afterward were even more valuable.

The people you met in your groups were chance encounters, but if you come to the next Bloom session, you might meet again. I hope the people who've already met stay in touch.

And one more time, for the record: the first office Fieldguide is opening outside the US is in Seoul, and it's an AI product team, not a support team. This time, the Silicon Valley our community is always curious about walked over to us. I'm grateful we could make this opportunity happen together and bring new energy and insight to the community.

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