He fired almost everyone, kept 2 engineers—then grew 12x and raised a $30M Series A. | Rafael Broshi, Co-Founder of Notch

Notch cut to two engineers, then won a $1m-a-year contract

Notch sells AI agents that run enterprise customer support, and it exists because its first product failed. The digital-asset insurance policy it launched in 2021 won praise and no customers, so when ChatGPT arrived the founders cut to two engineers and turned an internal tool for servicing policyholders into the company. A late entry into a carrier's proof of concept became a four-year contract worth about $1m a year, then a $30m Series A.

Takeaways

• Product-market fit arrived as a won POC. Notch entered a large insurance carrier's proof of concept late and through the back door, then shipped three versions of its agent in days, with the team sleeping in the office twice. Winning produced a four-year contract worth about $1m a year, and the client is now in production.
• Generic requirements were the real signal of fit. The carrier's asks were generic, so Notch made small tweaks and two integrations instead of rebuilding. Broshi treats winning a POC by customizing everything as a yellow light, because that turns a startup into one client's personal dev shop.
• The pivot cut the team to two engineers. Notch had raised $7m by 2023 and had $2m left. It fired almost everyone, leaving one or two engineers, and built for eighteen months without hiring, burning $10,000 to $20,000 a month. A first six-figure contract closed at the end of 2024.
• The agent scales by adding components, not prompts. As complexity grows, Notch's architecture adds LLM components rather than lengthening the prompt, so it scales by adding compute. The carrier ran thousands of tests to confirm response quality held as more guidelines and processes were covered.
• Don't educate a market, ride a breaking wave. The digital-asset insurance policy launched in 47 states and was called genius, but it solved no real pain. When ChatGPT arrived in November 2022, companies that had spent millions on earlier support technology were suddenly on an even technological playing field.
• Competition rarely kills startups at this stage. Rivals in Notch's space raised $500m against its $7m, and Broshi argues competition mostly pushes founders to copy others. A crowded market demands a crystallized story, and a seed investor only needs to believe someone will lead the next round.

Enterprise selling playbook

• Work bottom-up and top-down at once, and ask for introductions from anyone, including family and investors. Broshi says people rarely refuse a request for help.
• Respect working-level people and never reveal that you came in through the CEO. Give people what they want rather than what you have.