Amazon Isn’t Afraid of AI Shopping, It’s Afraid of Losing the Customer

Last week, Amazon blocked third-party consumer agents from its site.

Y Combinator cofounder Paul Graham reacted quickly with a post on X, arguing that Amazon's restrictions represent the first real opportunity he's seen for a competitor to challenge its dominance since the company was founded in 1994.

I think he's onto something.

But here's what makes this interesting for founders: The next major Amazon competitor may not need warehouses, a delivery network or even a marketplace. It may only need to control the relationship between consumers and the companies selling them products. 

Most interestingly, Amazon's restrictions could provide a roadmap to an entirely new category of commerce startups.

Amazon has spent more than 30 years building an almost impossible competitive position: warehouses, supply chains, hundreds of millions of products and a delivery network few companies could replicate even with unlimited funds.

But its greatest advantage is that hundreds of millions of consumers start their shopping journeys on Amazon. This gives the company unmatched influence over what people discover, which products they consider and what they buy.

It's also the foundation of a staggeringly large advertising business. In 2025, Amazon made $68.63 billion from advertising, up 22% from 2024. Brands pay Amazon because that's where purchasing decisions happen.

Now imagine telling your personal AI assistant:

"Find me a reliable espresso machine under $800. I want something easy to maintain, with good reviews, that arrives before Friday. Compare the best options and buy the one you'd recommend."

Your assistant might buy from Amazon, directly from the manufacturer or from another retailer you've never heard of.

You don't care. On Friday morning, you're sipping fresh espresso.

That is the threat. The agent becomes the interface, and Amazon becomes one potential supplier among many. The customer relationship has moved elsewhere.

Amazon isn't opposed to autonomous shopping. It's invested heavily in its own AI shopping assistant, Alexa for Shopping. What it doesn't want is an independent agent sitting between Amazon and its customers.

We had a preview of this over the past few months as Amazon began restricting Meta's Muse and fighting with Perplexity.

Amazon does want AI to make shopping easier — just so long as it can keep consumers inside its ecosystem, which is essential for its advertising revenue. Independent agents want to make shopping easier without requiring consumers to use any particular retailer.

This is where Graham's argument gets especially interesting.

What if the next major competitor doesn't need to replicate Amazon's infrastructure? Which, realistically, is nearly impossible. What if it only needs to own the customer relationship?

I see three opportunities.

Independent shopping assistants. An agent that works across retailers, remembers your preferences, tracks prices, evaluates products and buys wherever the best option is available. Its loyalty is to the customer, not the retailer.

Agent-first commerce platforms. Instead of attracting shoppers to a traditional storefront, these businesses connect merchants directly with AI agents, making products easy to discover, compare and purchase.

Infrastructure for agent-driven transactions. Agents need reliable product data, inventory information, payments, identity verification, returns and ways to resolve problems when transactions go wrong.

Somebody has to make all of that work.

None of these businesses needs to commit to two-hour delivery if they are effective at competing for the decision that happens before the package ships.

We're already seeing early signs of this behavior. Public co-CEO Leif Abraham recently described using Meta's Muse to buy products directly from brands rather than defaulting to Amazon.

When the assistant becomes the shopper, the traditional storefront becomes less important.

There's a much broader lesson here for founders.

If customers want to delegate a task to an agent, and an incumbent makes that difficult because it threatens its economics, there's an opening. The incumbent has to protect its existing business, but a startup doesn't.

Look for industries where customers spend time navigating complicated interfaces, comparing options, filling out forms or managing transactions. Booking travel, buying insurance, financial services and subscription management all come to mind.  

Then ask: What happens if the customer no longer needs to interact with the incumbent's interface at all?

I've written before about why every major AI company wants to become your assistant. This is the commercial consequence of that race. An assistant that knows your preferences, understands your needs and can take action on your behalf becomes the interface between you and the companies you do business with.

Amazon isn't going away. Its logistics network, selection, pricing and customer trust remain enormous advantages. Left to their own devices, independent agents would send Amazon substantial business.

But there's a major difference between being the company a customer chooses and being a supplier an agent selects. One owns the relationship; the other competes for the transaction.

For founders, that's the opportunity Graham points to. The next generation of commerce startups don’t need to out-Amazon Amazon.