Starting an AI Company in New York as a Foreigner

A conversation with a younger friend from high school who went from a PM in Silicon Valley to an AI founder in New York

A friend from high school, five years younger than me, reached out. She'd spent five years as a product marketer at a publicly traded financial software company in Silicon Valley, then moved to New York and started an AI company. As a foreigner in the US, she has founded a company, raised money, launched a product, and is now networking, selling and landing customers.

When a Korean company tries to expand overseas, you can expect a lot of hurdles. The legal stuff like visas, though, you just deal with as it comes. What gets me is selling to locals in another country without feeling outmatched, when I get nervous selling even in Korea. But I think a global mindset isn't anything special. It's realizing that people in that country are human beings just like me. The rule is the same everywhere: focus on solving the problem of the customer in front of you.

Anyway, after hearing how she started and runs a startup in the US as a foreigner, it came down to two things.

First, she builds on the experience and knowledge from the field she worked in to get ahead in the AI era. The opportunity is even bigger if that field is a legacy industry, and it gets more powerful when it connects with pain points that only Americans locally feel. For example, Americans buy on credit far more than Koreans do, so there are many more people with low credit scores. That's why there are startups offering loans to people with low credit, and some of them have quickly become unicorns. Businesses grow out of places we'd never think of, ones only locals know about.

Second, overseas becomes local. Of course, even traveling abroad makes you tense. But what if the place stops being "abroad" and becomes home? Once you clear the visa or immigration hurdle, people adapt quickly and just get on with life. The harder part, harder than visas or sales, might actually be blending into the country's systems in your daily life: opening a bank account, signing an apartment lease, getting a phone plan. You might think that has nothing to do with business, but this is how you become part of the country.

Today I talked with her casually and heard what startup idea she came up with that works in America, why she left a good job at a public company, and why she's starting out competing with her former employer. It's a business you could only come up with by actually living in the US, which I found pretty fascinating.

1. What five years gave her

She spent five years at the financial management platform most widely used by small and medium-sized businesses in the US. The platform's main customers were small service businesses: cleaning services, landscaping services, HVAC (heating and cooling) services. She said that after five years of working directly with owners in these industries, she knew their workflows inside and out.

What matters is that what she built up during this time was domain knowledge rather than technical skill. Which kind of business owner works in what order, where the bottlenecks are, what they handle by hand. She never wrote a single line of code, but those five years became the foundation of her company.

2. Every business repeats the same manual work

She found a pattern. Small service businesses all follow the same workflow. A new customer inquiry comes in, and they respond, negotiate, put together a quote, get the contract signed, send the invoice, and collect payment. Whether it was cleaning, landscaping or HVAC, the flow was identical.

And new customers asked almost the same questions. The business owners were answering those similar questions in similar ways. All by hand. Some owners hired virtual assistants (VAs) in the Philippines to handle this work. Using offshore staff meant time zone differences, communicating over WhatsApp, and frequent mistakes. Even so, this was accepted as "the most efficient option there is."

She saw an opportunity here. What if an AI agent could take over this repetitive, manual workflow? No time zone gap, no WhatsApp, just AI working alongside the business owner. Instead of hiring an offshore VA, AI fills that seat.

3. Narrowing end to end down to one thing

Her first plan was to automate the entire workflow with AI, from answering inquiries to collecting payment. But once she actually started with small businesses, she said their expectations were too high. "If you're replacing my virtual assistant, you need to do everything that assistant did." That's what customers expected. Covering the whole workflow meant too many use cases, and the product's scope grew wider than she could handle.

Around then, she met a customer through a private equity (PE) firm: a cleaning services company with 200 franchises across the US and Canada. The company was struggling with customers paying late after invoices went out. So it had hired a separate four-person team in the Philippines to send emails by hand and chase down unpaid invoices.

This is where she changed direction. Instead of doing the whole workflow, she'd do one thing at the very end of it, and do it well: AR (accounts receivable) collection, chasing unpaid invoices. She decided to focus on making an AI agent good at this one job. She went from "I'll do everything" to "I'll do one thing well."

4. The work of four people in the Philippines

The situation at that 200-franchise cleaning company, which became her first paying customer, was very specific. They send invoices and customers don't pay on time. So a four-person team in the Philippines was sending emails one by one, by hand, chasing unpaid invoices. Because of the time difference they couldn't respond in real time, and errors crept into the emails. But the core of the job was simple: send a set message at a set time, and follow up based on the reply. It was exactly the kind of work an AI agent does best.

The PE firm introduced her to this customer. The firm invested in cleaning service franchises and wanted to make its portfolio companies run more efficiently. Her solution was a perfect fit.

5. From small businesses to mid-sized companies

Her target changed too. She originally went after small businesses, but the ones that actually became paying customers were mid-sized companies with more than $30 million in revenue. Companies this size are too big to use her former employer's platform, but they aren't using modern SaaS either. One customer she met was running its business on a 1998 version of a Microsoft product. That's how legacy the environment was.

I found this part interesting. Legacy industries, legacy software, legacy processes. All manual, all old. But that very "oldness" is an opportunity for AI agents. Modern tech companies are already automating internally or can build it themselves. Companies running 1998 software don't have the capacity to adopt AI on their own, let alone do vibe coding. That gap is the market.

6. Angel money, with no product

We talked about fundraising too. She said she raised an angel round before she even had a product. $300K for 9%, a $3.3 million valuation. She said that's a standard angel valuation by Silicon Valley standards. The team is just two people. She's the CEO, and her cofounder is the CTO. A computer science engineer paired with a domain expert.

What struck me was that she raised it "with no product." Her domain expertise, her understanding of the market and how she defined the problem were convincing enough on their own.

I asked about hiring plans too. She said once paying customers grow from 1 to 10, she plans to raise a proper pre-seed or seed round and hire then. And even then, she wants people who are good at using AI.

7. Finding customers is still manual

She's building an AI agent company, but finding customers is completely manual. The lead list she showed me was impressive. First customer: an introduction from a PE firm. Second customer: also an introduction from a PE partner. Third: the Columbia MBA network. (She isn't an alum. She just showed up at a school event uninvited.) Fourth: an investor's former business contact. Nearly all of them were warm leads.

Selling through PE firms was her core channel. PE firms invest in franchise businesses like cleaning and landscaping services and use a roll-up strategy, combining several of them into a platform. From the PE firm's point of view, an AI solution that can make its portfolio companies run more efficiently is attractive. She was going straight at exactly that point of contact.

She was using conferences too. Before going, she figures out who'll be there and researches which industries each PE firm invests in. Cold outreach, on the other hand, gets almost no response, she said. You send emails and nobody replies. In B2B sales, especially to mid-sized and larger companies, relationships are still at the core.

8. New York and San Francisco

She's in New York right now, but that's because of an accelerator program run by a venture capital firm. She said she'll probably go back to San Francisco once the program ends.

I asked about the difference between New York and San Francisco. Her summary was clean. San Francisco is where people shooting for the moon gather. Crazy ideas, big visions. New York is more diverse, and the problems people are solving are more practical. Lots of people there are actually solving problems or building something grounded in reality, she said.

She didn't go as far as saying New York is "way better." She called it "working." Networking happens, you can meet customers, there are builders, things are moving. The nuance was that this is enough. Her attitude was that what matters is which city fits your business, more than which city is objectively better.

9. Customers before news

I asked where she gets the latest tech insights. Whether you have to be in San Francisco to get them, or whether there's another way. Her answer was clear: "I don't read that much tech news. I listen to podcasts and follow newsletters, but I spend most of my time talking to customers and meeting customers."

Her point was that understanding customers' problems matters more than chasing trends. News moves on, but customers' problems stay. Especially in B2B, how much you learn from dealing with customers directly is on a completely different level from reading ten newsletters. She brought up Paul Graham's "Do things that don't scale" during our conversation, and that's exactly where she is right now. Meeting customers in ways that don't scale, managing leads by hand, building relationships one person at a time. I found the irony funny: a company that builds automation is growing without any automation.

10. Becoming a local

As we wrapped up, we came back to the question we started with. How is it possible to start a company in the US as a foreigner?

What I noticed watching her is that she doesn't even have a "starting a company in America" frame. She's just solving a problem she saw, in a field she knows, in the place where she lives. She has domain knowledge from five years of working with small businesses in the US, those customers are still working by hand, and AI can replace that manual work. That connection is all there is to it.

If you sit in Korea wondering "how do we break into the US market," everything from market research to sales looks like an endlessly high wall. But for someone already inside that market, it isn't a wall. You meet local PE firms, go to local conferences, find local networks. For her it isn't overseas expansion. It's just business. Half of how a foreigner starts a company in the US was "being there." The other half was domain.

Originally written in Korean on Brunch, April 17, 2026.