AI Is Not an Efficiency Play — It's an Amplifier of What You Already Have

This post is inspired by two conversations that reframed how we think about AI: the "Automation at Scale: Unlocking AI's Full Potential" panel at the Kellogg Executive Leadership Forum (February 2026), featuring Tarek Elmasry of McKinsey, Eric Acher of Monashees, and Anas Osman of Google Cloud — and the presentation by Shu Nyatta of Bicycle Capital at Caricaco Summit, where he argued that in a world of AI, the most durable competitive advantage isn't the model you deploy, but the infrastructure, channels, and relationships you already own.

There's a telling irony in the name of the Kellogg panel that sparked much of this thinking: "Automation at Scale." Automation is real, and it matters. But if that's the ceiling of your AI ambition, you're leaving the most valuable part on the table.

The businesses that will define the next decade are not the ones that automate the fastest. They're the ones that recognize what they already have — and ask a different question: what becomes possible now that was simply impossible before?

The Asset You're Sitting On

Most established organizations have spent years — often decades — accumulating three categories of durable value: infrastructure, channels, and relationships.

Infrastructure is the backbone: operational systems, logistics networks, institutional knowledge, data accumulated through every transaction and interaction. Channels are the pathways to market — distribution partners, sales teams, digital touchpoints, the reach that took years to build. And relationships are perhaps the most underappreciated asset of all: the trust, context, and history embedded in every customer, partner, and supplier connection.

A startup can access the same AI models you can. What it cannot do is manufacture 15 years of customer behavior data, a trusted regional brand, or a distribution network spanning hundreds of cities. That asymmetry is the whole game — and AI is the tool that converts it from latent potential into active competitive advantage.

Shu Nyatta put it precisely at Caricaco Summit: Latin America is "under construction." That framing matters. It doesn't mean the region is behind — it means that the incumbents with established trust and reach have the raw material for something extraordinary, if they choose to see AI as fuel rather than threat.

From Efficiency to Market Creation

The clearest way to understand the difference between automation and exponential innovation is this: automation makes existing processes cheaper. Exponential innovation creates markets that didn't exist.

MercadoLibre is the most instructive case in Latin America. For decades, hundreds of millions of people across the region had no access to formal credit — not because they were bad borrowers, but because traditional banks had no data to assess them. MercadoLibre had that data: 25 years of transaction history, buyer and seller behavior, payment patterns, dispute resolutions, across 200 million users. AI turned that existing marketplace infrastructure into an underwriting engine, enabling MercadoLibre to extend credit to tens of millions of people who had never qualified before. The result is a credit portfolio growing at 56% annually, now exceeding $7.8 billion. They didn't optimize a lending process. They created a lending market.

Kavak did the same for used cars in Mexico. The used car market in Latin America was functionally broken — opaque pricing, rampant fraud, and no financing, which meant 90% of transactions happened in cash. Kavak used AI to solve the information asymmetry that made the market dysfunctional: pricing models, quality assessment, risk underwriting. In doing so, they didn't improve an existing market. They built one from scratch, becoming Mexico's most valuable startup by making a trillion-peso industry accessible for the first time. The infrastructure they built — inspection centers, reconditioning facilities, a financing arm — is now the moat that no new entrant can quickly replicate. AI created the market; existing infrastructure defends it.

Relationships as the Engine, Not Just the Asset

The most powerful global example of AI amplifying existing relationships into entirely new businesses is Ping An, China's largest insurer. Ping An had 240 million insurance customers — an extraordinary base of trust and longitudinal health data. Rather than using AI to make claims processing faster, they used that existing relationship base to build Good Doctor, a healthcare delivery platform that now serves over 500 million users. AI made it possible to diagnose, consult, and triage at a scale no physical hospital network could ever match. In 2024, AI service representatives handled 1.84 billion customer interactions — 80% of Ping An's total service volume. But the headline number isn't the efficiency gain. It's that they built a healthcare company from an insurance company, using AI to make the leap.

That is the pattern: existing relationships as the on-ramp to a new business, not just a better version of the old one.

This is precisely what Shu Nyatta's thesis articulates for Latin America. The region's incumbent businesses — banks, retailers, telecoms, distributors — have something that no AI-native startup can replicate overnight: deep, earned trust across massive populations. AI doesn't threaten that advantage. It multiplies it. The question is whether those incumbents see their existing customer base as a constraint to be defended, or as the raw material for building things that were never before possible.

The Compounding Moat

There is a compounding dynamic here that is easy to underestimate. When an organization deploys AI on top of existing data and relationships, the system improves with every interaction. The moat widens over time, not shrinks. A new entrant deploying the same foundation models faces a cold-start problem: no history, no context, no trust. The incumbent deploying AI on a decade of behavioral data builds an insight layer that deepens daily.

This is the inverse of the usual narrative — that AI is a great equalizer that lets startups challenge incumbents. For organizations that choose to act, AI is a great amplifier of the advantages they already have.

The Kellogg panel asked how to unlock AI's full potential at scale. The answer isn't a faster automation roadmap. It's a strategic reframe: stop asking what AI can automate, and start asking what it makes possible for the first time, using what you've already built.

The Strategic Imperative

The window for this reframe is open. Companies like MercadoLibre, Kavak, and Ping An moved early — not because they had the most sophisticated AI, but because they connected AI most deliberately to their existing assets. The pattern is replicable. The opportunity is not.

For any organization with established infrastructure, channels, and relationships — especially in markets that are, as Shu Nyatta would say, still under construction — the question is not whether to adopt AI. It's whether you choose to use it to trim costs, or to build something that simply wasn't possible before.

The engine is here. The assets are yours. The only question is what you choose to build.

iitos works with organizations to design AI strategies that compound on what they've already built — not replace it.