Trust Is the Currency, Knowledge Is the Engine

_What Mastercard's AI lead understands about enterprise transformation that most organizations are still missing._

Federico Cohen Freue receives around a thousand AI requests annually. This is the incoming volume to Mastercard’s central AI and data team—proposals, ideas, and requests from a global enterprise trying to determine where and how to deploy the technology.

A few years ago, more than half of these requests were for chatbots. Today, more than half are for agents.

Fed views this as a positive sign, not because agents are better, but because it indicates that people’s mental models of what AI can do have matured. They are no longer simply imagining a question-and-answer interface. Instead, they envision AI taking action, integrated into workflows, performing tasks on their behalf.

The issue, as he carefully notes, is that wanting agents and being prepared for agents are two distinct matters.

The Ball Bearing Problem

Robb introduced an analogy in this discussion that is worth pondering. Ball bearings: two that appear identical. One that has been precisely machined, using the right material and tolerance. The other has not. You cannot tell the difference by looking at them. Insert the defective one into an aircraft engine, and the engine will fail.

Agent demonstrations work similarly. Even a sophisticated observer cannot distinguish a demonstration that represents a viable solution from one that is a beautifully polished failure waiting to occur. The visual experience is identical, but the underlying engineering differs.

This is why Fed’s team invests heavily in training and fluency before deployment. Simply letting demand drive the agenda is insufficient. If people understand the conditions under which AI functions effectively—what it means to be "machined correctly"—they will make better requests, build better things, and identify failures before they become significant issues.

A Framework Simple Enough to Be True

One of the more practically interesting aspects Federico described is how Mastercard approaches prioritization at scale. With a thousand incoming ideas and a technology that continues to expand its capabilities, how do you decide what to focus on?

The answer is a framework that fits in a single sentence: use AI to enhance security, intelligence, personalization, and strengthen Mastercard.

That’s it. The simplicity is the point. When teams across the organization share a common language for where AI should be deployed (a unified framework that remains relevant as models improve and use cases increase), prioritization becomes a conversation rather than a negotiation. While it doesn’t answer every question, it addresses the most fundamental one: does this fit?

What is noteworthy is what the framework does not do. It does not establish guidelines or create a compliance checklist. Instead, it serves as a strategic lens that helps people think before they ask.

Trust at the Moment of Transaction

As the conversation shifts to agentic payments, the stakes become more tangible. Executing financial transactions on behalf of users via AI agents is no longer a speculative scenario. Consumer demand exists, and product discovery patterns are already shifting towards LLM-driven searches. The loop is about to close.

Mastercard’s response to this situation is telling. The top priority is not developing exciting downstream applications such as multi-vendor trip booking, smart replenishment, or algorithmic negotiations between buyer and seller agents. The primary focus is ensuring that the foundational case works. This includes verifying agent identities, establishing delegated authority frameworks, setting acceptance standards for merchants, and creating a rules infrastructure that guarantees when an agent executes a transaction, all parties in the ecosystem can trust what occurred.

Fed stated plainly: trust is the currency of innovation. As transactions become more complex (involving more parties, dynamic pricing, and autonomous decision-making along the chain), the role of a trusted network does not diminish; it grows.

The intermediary who was supposed to become obsolete becomes the most critical node in the system.

Knowledge Before Action

In the latter part of the episode, we explore a demo of something we have been working on: an AI-centric approach to knowledge management and learning.

The premise critiques how enterprise AI typically operates. You build a knowledge base, then wait for someone to query it. You develop an AI that knows things, then ask it questions. The chatbot model places the responsibility of knowing what to ask on the person who needs to learn the most.

An alternative is a proactive system that does not wait to be queried but determines what you need to know and delivers it.

The architecture begins with a knowledge model—a structured, canonical source of truth for a specific domain. Unlike a document repository where the same idea exists in seventeen versions across seventeen files, this is a map where each idea resides only once, connects to what it relates to, and carries a history of its evolution over time.

From this map, the system creates a learning twin—a representation of what a particular person knows and does not know. It then solves what Robb refers to as a traveling salesman problem: given your current location and destination, what is the most efficient route? Rather than a static curriculum designed for an average learner, it offers a dynamic path recalculated at each step based on what you have recently learned and any changes in the domain since your last inquiry.

GPS for expertise. Here is your current position. Here is your destination. We will guide you step-by-step, and if the route changes, we will reroute.

Fed's response was immediate: this isn't just a technology problem. It's a cultural one. Asking people to engage with knowledge differently (to treat learning as a dynamic, ongoing process rather than a thing you did once during onboarding) requires a shift in how organizations think about readiness and what they reward. The technology can be ready before the culture is.

What Comes Before Doing

The thread running through this conversation is sequence. At Mastercard, there’s a deliberate ordering: understand first, then act. Build fluency before you deploy. Verify identity before you authorize a transaction. Know what you know before you build a curriculum.

This sounds obvious. It’s not how most AI initiatives actually work. Most AI initiatives start with the doing. An agent to automate this. A model to replace that. A chatbot to answer questions no one knows they have. Then they fail, and people call it a technology problem.

It’s usually a knowledge problem. The system didn’t know enough or the people managing it didn’t do the thing reliably. You don’t fix that by improving the model. You fix it by treating knowledge as infrastructure.

That’s the reframe this conversation is pushing toward: before you ask what AI can do, ask what your organization actually knows. Because the agents are only as good as the knowledge they’re built on.

Listen to the full conversation with Federico Cohen Freue on Invisible Machines.