Building Enduring Companies Outside Silicon Valley | Santiago Suarez of Addi

Key points

• The wedge thesis: 'be very big in your ambition, but be very contrarian in how you get there' — 'if you're a consensus play, there's just no alpha'; the LATAM consensus was Brazil/Mexico or a thin layer across markets, and Addi went deep in one 'too hard' country instead.
• Ignorance as an edge: deep US-fintech expertise made him reflexively see 'the 10 ways your startup's gonna fail' (the Bank Holding Company Act, Durbin, SWIFT); not knowing the Colombian rulebook let him act — 'a little bit of ignorance is always bliss.'
• He read the market through customer pain, not decks: smartphones went from years-behind to universal (2014-16), giving 'a distribution device in everyone's hands at zero marginal cost'; against that, legacy UX was absurd (a 22-minute t-shirt instalment requiring fingerprints, two photos and calling two friends; banks charging ~20 cents to use online banking).
• Craft acquired on purpose: fired from his first startup (CEO for 35 days), he listed the best operators (Costco, Amazon, Wynn — not finance) and engineered six years inside Jamie Dimon's 10-person strategy group, 'a whole university in how to run companies'; he also insisted on a green card first so he'd never build 'on a visa' (bigger company, lower upside, lower downside).
• Studying Kaspi (Kazakhstan): four cold LinkedIn messages and a flight to Central Asia bought three lessons — obsess over NPS / 'just listen to customer service calls' (weekly business reviews open with two random customer transcripts), sequence the roadmap rather than do everything at once, and 'equity investors will not understand this — just ignore them for 5 to 10 years'; plus 'e-commerce without logistics.'
• The foundation is the moat: a monorepo (chosen 7 years ago over then-dominant microservices because 'I hired the best CTO') and an event-sourcing architecture logging 10M+ events/day via Kafka into Databricks — exposing events as vectors for LLMs, SQL for classical ML, tabular for humans — which made the company genuinely AI-ready when LLMs arrived.
• AI-native operations at scale: 200+ agents in production; in-house agents 'handle 100% of all customer service queries' and fully resolve ~80% with no human in the loop; a merchant-onboarding agent brings on 2,000-3,000 merchants/month at 100% handle and 20%+ better conversion; a web rebuild took '2 engineers, 2 months' versus the old '6-9 months, 5 engineers.'
• Counterintuitive sequencing — start AI with the hardest problem: they began with legal, not support, because Colombian 'tutelas' demand a 48-hour response or the CEO is liable to jail, and Addi never settles (sets precedent); building those RAG/RLHF pipelines was a 6-month investment that then yielded a customer-service agent at 60% resolution within 90 days — 'if you could resolve a lawsuit, you could resolve most customer service interactions.'
• Talent and culture: run the company in English (attracts global talent and, counterintuitively, raises the bar so local talent wants in — 'a premium'), keep a very high editorial bar ('we only need a few hundred'), be remote-first so context is explicit and agents can use it; to move the org to AI, the founder first built his own stack from an empty EC2 instance ('if this is helpful for me at this scale... the entire company should be able to do this'), then made adoption 'not optional.'
• Operating principles that transfer: 'remember you're a technology company' (where equity value compounds — they obsessed over compile time creeping from 10 to 33 minutes); replace a wall of OKRs with one nameable North Star metric (citing a Sequoia post that Elon 'spends all his time on the most important thing'), they rode profitability — risk-adjusted margin → gross margin → minus S&M → EBITDA; and write everything down, always able to 'articulate the why even if it's the most obvious why' — which made the company's SOPs agent-ready by accident, and 'you don't have to be in San Francisco to push the envelope.'

Notable quotes

Themes

• contrarian wedge over consensus
• building enduring companies outside Silicon Valley
• AI-native operations on a clean data foundation
• learning the operator's craft from the best
• focus (a single North Star) over sprawl