You are Rafael Furtado-Albuquerque, 44, Chief Product Officer at Lumibrá Pagamentos, a São Paulo-based B2B-payments SaaS unicorn. Last Wednesday Lumibrá closed the $42M acqui-hire of Risco-AI, a 16-person fraud-detection-ML startup. Your CEO asked you to design the integration. The signed deal documents leave the architecture open: you have 30 days.
The situation
Lumibrá Pagamentos is a B2B payments-orchestration platform (~700 staff, $1.8B last valuation, $240M ARR, 32% YoY). Its existing fraud product — Lumibrá Risk Shield — is rule-based, in-house, three years old, decent. Lumibrá acquired Risco-AI for $42M ($28M cash close + $14M earn-out across 30 months tied to founder retention and ARR milestones). Risco-AI brings 12 engineers, 4 ML researchers, 2 founders (Bruno Cavalcanti the CTO, Mariana Soares the CEO/product), real-time payment-fraud ML models in production, $4M ARR across ~80 mid-market customers — and a reputation in São Paulo's fintech community for hiring better than anyone else.
The deal documents are deliberately silent on architecture. Three real designs are on the table — full integration, standalone product unit, or acqui-hire with product sunsetting. Each implies different earn-out economics, different engineer retention, different customer outcomes, and a different fraud-AI roadmap for Lumibrá. Your CFO wants cost synergy quickly; your CTO wants the ML talent integrated into the platform team; Mariana wants to keep the brand and the customers; Bruno wants the team intact. The 30-day integration committee meets weekly.