We Raised $8M and Built AI That Knows When Customers Will Leave
Sean and Dan met at IBM out of college, spent 15 years working together, and kept seeing the same problem: customers churning with zero warning. In 2023 they quit to build QuadSci, an AI that reads product usage data and predicts which customers will leave or expand 12 to 18 months before it happens. Three years later they’ve raised an $8M Series A, grow 5X year over year, run everything with 37 people, and no customer has ever turned the AI off. In this episode, they break down how they won enterprise customers with no product, why they run as co-CEOs, and why they don’t charge until the AI proves itself.What we cover:- 15 years of working together before starting a company- Why ChatGPT’s launch made the business possible- “Telemetry is truth”: why usage data predicts churn better than what customers say- Finding 90% of churn and 92% of expansion 12–18 months early- How they prove it works: rewinding the model a year to show what it would have predicted- Landing $100M+ ARR customers as an unknown startup- Early hires: people who’ve won in a bank, a tech company, and a startup- Why co-CEOs works, and why staying flat makes them faster- 37 people, 2 in support, everyone else building or selling- Investors who doubled as advisors, board members, and customer intros- Free until it works: where performance-based pricing is going- The churn signal most founders have but can’t act on early enough- Why you should still finish your CS degree in the age of AIKey takeaway: Solve a real problem, prove it works before asking for money, and let trust compound. It’s the only moat that survives everyone innovating at once.QuadSci: https://quadsci.ai/
