AI Summary
→ WHAT IT COVERS Santiago Suarez, founder and CEO of Colombian fintech Adi, details how his company scaled to 3 million consumers and 50,000 merchants by combining buy-now-pay-later, payments, logistics, and banking — while deploying over 200 AI agents that handle 100% of customer service queries with 80% full resolution rates. → KEY INSIGHTS - **Monorepo architecture as AI foundation:** Adi built on a single-code monorepo rather than microservices — a contrarian choice when microservices dominated. This decision means all code is readable by AI agents in one place, enabling faster deployment. Paired with event-sourcing via Kafka and Databricks, the system processes 10 million+ daily events queryable in real time for LLM consumption. - **Start AI deployment with hardest use case first:** Rather than beginning with customer service like most companies, Adi launched AI with legal response automation — Colombian law requires constitutional lawsuit responses within 48 hours or the CEO faces jail. Building those data pipelines first meant customer service agents reached 60% full resolution immediately at launch, versus a slower ramp from easier starting points. - **North Star metric over OKR cascades:** Replace sprawling OKR frameworks — which can generate 25+ tracked items — with one to three L1 metrics the entire company can name. Adi sequenced through risk-adjusted margin, gross margin, gross margin minus sales and marketing, then EBITDA. This single-metric focus drove Adi from losses to profitability and kept weekly business reviews actionable. - **Remote-first operations accelerate AI adoption:** Running a remote company forces all context to be written explicitly rather than shared verbally. This creates structured APIs and documented SOPs that AI agents can directly consume. Adi currently operates 150 headcount below budget while exceeding growth targets, attributing the efficiency gap directly to agents operating on this explicit written-context infrastructure. - **Contrarian market selection over consensus geography:** Most Latin American fintech founders target Brazil or Mexico first. Adi chose Colombia — then considered too cash-dependent (65-70% cash transactions, under 20% credit card penetration) — because smartphone adoption surged from elite to mass-market between 2014 and 2016, creating zero-marginal-cost distribution. Focusing deeply on one country rather than spreading thin across markets mirrors Kaspi's Kazakhstan playbook. → NOTABLE MOMENT Suarez described flying to Kazakhstan after four unanswered LinkedIn messages to meet Kaspi's CEO, who advised him to ignore equity investors for up to a decade when building a deeply integrated single-country platform — predicting they would eventually ask why he never called. 💼 SPONSORS None detected 🏷️ Latin America Fintech, AI Agents, Monorepo Architecture, Financial Inclusion, Startup Scaling