Building the Elemental Fabric: Lessons from Addi’s Scale

In this conversation, Addi CEO Santiago Suárez explains that the most durable competitive advantages often come from ignoring standard industry playbooks. By prioritizing systemic infrastructure, specifically mono-repo architectures and event-sourcing, over the common modular microservices approach, Addi built a platform where AI acts as a core operational layer. The result of these difficult early technical choices is a compounding efficiency that allows the company to outpace competitors with fewer people. This analysis is useful for founders who want to understand why solving harder problems first, such as legal automation, creates a moat that protects the business long after the initial growth phase. The advantage is not just speed; it is the ability to sustain high-velocity innovation by avoiding the technical debt that slows down most scaling startups.

The Hidden Cost of Easy Architecture

Most engineering teams default to microservices because it is the industry standard. Suárez argues this is a mistake for companies aiming for high-velocity AI integration. By choosing a mono-repo architecture, the same approach used by Google, Addi ensured that their AI agents could access the entire codebase as a single, coherent context.

"The monorepo gives you, obviously, for folks who don't know what the monorepo is, basically it's all your code is in a single repository. So for agents to read it, it's a lot easier than if you have multiple code bases on many parts of the org."

-- Santiago Suárez

When you fragment your architecture, you fragment your AI’s ability to understand the system. Addi’s decision to keep everything in one place was not about current convenience; it was a bet that future AI agents would need holistic visibility to perform complex tasks. This is an example of immediate discomfort, managing a massive repo, creating a lasting advantage in the age of LLMs.

Why the Obvious Fix Makes Things Worse

Conventional wisdom in fintech suggests that when building a new product, you should outsource non-core tasks like KYC or loan management to third-party services. Suárez challenges this, noting that in emerging markets, many of these services are either non-existent or commoditized to the point of zero equity value.

Addi’s strategy was to build the entire stack in-house. While this created pressure early on, it prevented the system from becoming a collection of third-party dependencies that break when the market shifts. By owning the full stack, from payment rails to settlement, they gained the ability to move with a speed that competitors relying on external vendors could not match.

The 18-Month Payoff: Why Solving Legal First Actually Scales

When implementing AI, most companies begin with customer service to reduce ticket volume. Addi took a different path: they started with legal automation.

"I remember when we were discussing this, I was like, but let's start with customer service like everyone else... they were like, no, this is the harder problem, but this is the one that's scarce."

-- Santiago Suárez

The system dynamics here are important: legal challenges in Colombia carry a 48-hour response window and potential jail time for leadership. By building an AI agent to solve this hard problem, they were forced to build robust data pipelines that could pull from every corner of the company. Once those pipelines were built, solving customer service was trivial. By tackling the high-stakes, high-friction problem first, they built a foundation that made all subsequent AI deployments easier.

Where Immediate Pain Creates Lasting Moats

Suárez emphasizes that following the consensus is a mistake if you want to stand out. If you are doing what every other startup in Brazil or Mexico is doing, you are not building a moat; you are just competing on price.

Addi’s decision to stay in Colombia and build their own infrastructure, despite the advice to expand horizontally across many markets, allowed them to achieve deep penetration in 1,000 cities. The system rewards this focus with higher margins and lower cost-to-serve. While competitors were busy managing the complexity of multiple regulatory environments, Addi was busy refining the elemental fabric of a single, high-growth market.

Key Action Items

• Audit your architecture for AI-readiness: Over the next quarter, evaluate if your current service boundaries prevent an AI agent from having a holistic view of your product logic. If they do, consider the long-term cost of that fragmentation.
• Prioritize High-Friction automation: Instead of automating the easiest tasks, identify the high-stakes, high-liability processes in your business. Solving these first forces you to build the data infrastructure that makes everything else easier.
• Enforce a Written-Down culture: Immediately begin requiring that all major decisions and the reasoning behind them be written down. This creates the explicit context necessary for AI agents to eventually take over those processes.
• Kill the OKRs as a crutch model: Move toward a North Star metric. If your team cannot name your top three KPIs, you have too many. This pays off in organizational alignment within 3-6 months.
• Build your own stack: If you are building a core product, stop relying on third-party services for the commodity parts of your stack if those parts are where your unique data lives. Own the data pipelines now to avoid dependency traps later.
• Standardize your operating language: If you are hiring globally, consider running the company in a single language to raise the bar for talent and ensure all documentation is accessible to future AI agents.