The conversation about AI in commerce has moved from speculative to operational. Agents — autonomous software that acts on a user's behalf — are already comparing prices, evaluating shipping options, and routing purchases. They're doing the work that a human used to do by opening tabs, scrolling through options, and clicking "add to cart."

When an AI agent handles that workflow, it can only engage with what it can see. And what it can see is defined by what's machine-readable — APIs, structured data, queryable endpoints. Everything else is invisible.

Most loyalty programs are everything else.

The Visibility Problem

Consider how a typical loyalty program works from the consumer's perspective: open an app, navigate to the rewards section, check your balance, browse a redemption catalog, select an offer, apply it at checkout. Every step requires a human interacting with a visual interface.

Now consider what an AI agent can do with that same program: nothing. The agent can't tap buttons. It can't navigate app screens. It can't parse a visual rewards catalog that renders as images and custom UI components. The entire value proposition of the loyalty program — the balance, the offers, the redemption options — exists behind a UI that only humans can operate.

In that scenario, three out of four loyalty programs are functionally nonexistent. The agent can't factor in rewards it can't query. The consumer might have 50,000 points with Merchant A — enough for a significant discount — but the agent will never know, because the program doesn't expose that data in a machine-readable format.

The result: the program with the API gets the sale. Not because it had the best product or the best price, but because it was the only one the agent could see.

The Scale of Invisibility

This isn't a hypothetical edge case. The Australian Loyalty Association recently published research on the readiness of loyalty infrastructure for agentic commerce. Their assessment aligns with what any developer who's tried to integrate with loyalty programs already knows: the vast majority of programs are built for human interaction, not machine interaction.

Visa's VP of Loyalty Solutions, Avery Miller, recently noted that in an agentic world, issuers will be competing for every single transaction through the agent layer. That's correct — but it understates the prerequisite. Before you can compete for transactions through agents, your program needs to be visible to agents. Most aren't.

The programs that survive the transition to agentic commerce won't be the ones with the best app design, the most engaging gamification, or the cleverest push notification strategy. They'll be the ones with an API.

MCP and the Machine-Readable Standard

The Model Context Protocol (MCP) is emerging as a standard for how AI agents interact with external services. It provides a structured way for agents to discover capabilities, query data, and execute actions across different platforms.

For loyalty infrastructure, MCP changes the game. A program with an MCP-compatible server isn't just queryable — it's discoverable. The agent doesn't need to know in advance that a loyalty program exists. It can discover the program, understand its capabilities, check balances, evaluate offers, and execute redemptions — all programmatically.

That entire flow — discovery, balance check, offer evaluation, redemption — happens without a human opening an app, navigating a menu, or scanning a barcode. The agent handles it because the infrastructure is machine-readable by design, not as an afterthought.

The Survival Filter

Agentic commerce creates a natural selection pressure on loyalty infrastructure. Programs that are machine-readable survive. Programs that aren't become invisible — and invisible programs don't influence purchasing decisions.

The math is striking. Over 90% of companies have a loyalty program. A vanishingly small percentage expose those programs as machine-readable infrastructure. As AI agents increasingly mediate commerce — comparing options, applying rewards, routing transactions — the gap between having a loyalty program and having a visible loyalty program becomes the gap between relevance and obsolescence.

The Australian Loyalty Association's research frames this as a "readiness gap." That's diplomatic. It's closer to an extinction event for loyalty architectures that were built around the assumption that a human would always be the one opening the app.

What This Means for Brands

If your loyalty program requires a customer to open an app, navigate to a rewards section, and manually apply a discount — your program is one generation away from invisible. Not because consumers will stop caring about rewards. Because the agent that makes decisions on their behalf can't see what you're offering.

The transition isn't coming in some distant future. Agents are already comparing options across merchants. Visa is already talking about issuers competing at the transaction level through the agent layer. The programs that don't show up in that layer don't compete.

The infrastructure requirement is straightforward: documented APIs, structured data, machine-readable earn and redeem rules, programmatic authentication. The programs that have it become part of the agent's decision landscape. The programs that don't become noise — indistinguishable from not having a loyalty program at all.

The signal carries — but only through infrastructure that's built to transmit it.