Building an AI Startup in New York as a Foreigner
He raised 300 thousand dollars for 9 percent with no product, then narrowed from the whole workflow down to collections for mid-sized legacy firms.
What five years bought
He spent five years at the financial management platform most used by small businesses in America. The main customers there were the service businesses of the country, the ones doing cleaning, landscaping, HVAC. Five years of sitting in direct contact with them taught him their workflows inside out. The important part is that what those years built was not technical skill but domain knowledge. Which trades work in which order, where the bottlenecks form, and what still gets handled by hand. He never wrote a line of code himself, and those five years became the foundation of the company.
Everyone repeats the same manual work
He found a pattern, which is that small service businesses all follow the same flow. An inquiry comes in, they respond, negotiate, produce a quote, get the contract signed, send the invoice, and collect the payment. Cleaning, landscaping, HVAC, the flow is identical. Customers ask nearly the same questions, and owners answer them in nearly the same ways. All of it is done manually.
Some of them hire virtual assistants overseas to manage this, with the time zones, the messenger-only communication, and the frequent mistakes that come with it, and even that passes for the efficient option. There he saw the opening, which was what an AI agent could do with this repetitive manual workflow, with no time zone and no separate messenger, the AI simply working alongside the owner.
From everything to one thing
The first concept was automating the whole workflow, from inquiry all the way to collection. But starting with small businesses, the expectations ran too high.
If you are replacing my virtual assistant, you have to do everything my assistant did.
Covering everything meant too many use cases and a product surface he could not control. Then, through a private equity firm, he met a customer, a cleaning services company with 200 franchises across the US and Canada, which was struggling with customers who paid late after invoicing and was paying a separate overseas team of four to chase the receivables manually over email.
That is where the direction turned, so that instead of the whole workflow he would do one thing at the very end, collections, and do it properly. From doing everything to doing one thing well. The essence of that job is simple enough, which is to send the right message at the right time and then follow up based on the response. Exactly the kind of work an AI agent does best.
The market of 1998 software
The target changed along with it. He had aimed at small businesses, but the actual paying customers turned out to be mid-sized companies above 30 million dollars in revenue, too big for his former platform and not modern SaaS users either. One customer was still running the business on a 1998 version of its software.
This is the crux of it, a legacy industry running legacy software through a legacy process. All of it manual, all of it old. And that oldness is exactly the opportunity for AI agents. Modern tech companies already automate internally or can build the thing themselves. A company running on 1998 software has no internal capacity to adopt AI at all. That gap is exactly the market.
An angel round with no product
He raised angel money before the product existed, 300 thousand dollars for 9 percent at a 3.3 million valuation, which he said is standard for Silicon Valley angels. The team is two people, himself as CEO and a computer science engineer co-founder as CTO, the pairing of an engineer with a domain expert. The striking part is raising anything at all without a product, and what carried it was the domain expertise, the market understanding, and the problem definition. Hiring will wait until there are ten paying customers and a proper round, and even then he will only take people who use AI well.
Customer discovery is still manual
He is building an AI agent company, and finding customers for it is thoroughly manual. The first customer came through a private equity introduction, the second through a PE partner, the third through the MBA network, and the fourth through the previous client of an investor. Nearly every lead so far has arrived by introduction. Private equity became the core channel, because funds invested in franchises want their portfolio companies to run more efficiently, and the interests align exactly there.
When abroad becomes local
We asked about founding in the US as a foreigner. The answer came in two parts. One: use the experience and knowledge of the domain you were in as your edge in the AI era, because the more legacy the field is, the bigger the opportunity, and it gets stronger still when it touches pain points only locals know.
Two: the moment abroad becomes local. Once the visa and immigration hurdles are behind you, people adapt quickly. What is harder may be the living part, the opening of accounts, the signing of a lease, the setting up of a phone, the dissolving into the system of daily life. It looks separate from the business, and it is exactly the process that makes you local.
Keep reading
Join Bloom
Bloom builds offline rooms where people and technology meet. We run them in Seoul, and now beyond it.
- 💬 Discord community, where events are announced first
- ▶️ YouTube, full talks from past events
- 📸 Instagram, photos from our events
- 🎟️ Bloom app, upcoming events and Mixers
This post is based on a conversation with a founder who worked as a product marketer in Silicon Valley before moving to New York to start an AI company.
