The Amazon Trap

I love Amazon – both the product and the company. It’s the reason I find Amazon’s reaction to Instinct and Muse so jarring:

https://x.com/i/web/status/2104988337660420186 (opens in new tab)

I’m not mad at Amazon – I get it.

Over the past 20 years, Amazon did the impossible – they wrested a search vertical out of Google’s hands.

In the early days of the Internet, it was unclear if there would be a single search engine to rule them all or a set of vertical search engines. Consumers’ revealed preference was that they wanted a single place to ask all questions. Vertical search experiences (local, travel) slowly collapsed into the core Google search product.

There was one exception – shopping (and Amazon).

Amazon built an incredible shopping experience – the Everything Store – and then trained users that it was faster to start their shopping experience on Amazon, not Google.

This was not an accident – it was relentless execution. Amazon has an incredibly sophisticated SEO strategy. They don’t let Google index their whole product catalog. Many SEO links lead to a category page at Amazon designed to capture your search intent and train you to start your search on Amazon. Amazon has spent 15 years moving your search behavior from Google to their product. The $80B advertising business is the reward (and that $80B comes at Google’s direct loss).

This is why I empathize with Amazon’s current predicament. On the one hand – they have spent 20 years fighting to make Amazon the starting point for online commerce. They have done that – in part – by building a beloved, customer-delighting product rooted in ever-rising customer expectations.

Horizontal consumer assistants are going to work

I have been using Instinct for a while (disclosure, we are the initial investors and on the board). I have also been using Muse since launch and know many of the folks on the team (we were also seed investors in the Dreamer team, which now works on Muse!). They are both great products.

They’re both going to work.

Personal agents are just such a better experience. I had to book a hotel in NYC yesterday. With Instinct, I can just say “Can you book me a room at [the hotel] on the 21st?” It has my credit card and flight info. It knows what rooms I’ve booked before. It has my loyalty number. It can just book the room and add it to my calendar in just the way I like it. It’s not going to make a mistake with the dates. It is so much better than a dozen, slow-loading, form-heavy hotel web pages.

Given both my respect for Amazon and my belief that Instinct and Muse will work, I have been spending the past couple of days trying to think through the game theory of how this market will play out. What would I do if I were Andy Jassy?

Companies often anchor to their priors.

In this case, I suspect Amazon will default to the Amazon / Google playbook of the past two decades, i.e., make Amazon the “personal assistant for shopping.”

Amazon seems to be trying this. Over the past month, an “Alexa for Shopping” tab has appeared (and tends to randomly expand) on the Amazon home page:

On paper, this is a plausible strategy.

The problem is – I don’t think it’s going to work.

The magic of horizontal personal agents is the totality of their context – your email, calendar, files, messages, etc. In the same way that Google was the home page for the Internet, products like Instinct will be the starting point for anything you want to do. I don’t want to go to one agent to order food, another to book a car, another to book a trip, and another to buy baseball socks for my kids. I just want to be able to text a single agent that can figure it out and gets smarter with each interaction.

This puts Amazon in a Catch 22 – give consumers what they want and give up a core pillar of their strategic positioning (not to mention $80B/year of ad revenue), or stall and antagonize their consumers in the hope that horizontal agents don’t work.

The problem is – they’re going to work. And the optimal strategy for Amazon’s competitors is to go all-in on supporting these agents. If the market is shifting and Amazon refuses to participate, this creates a huge opportunity for Walmart, Shopify and others to steal market share as consumer demand evolves.

What should Amazon do?

Amazon should launch a new subscription tier above Amazon Prime. Call it Prime+.

Amazon Prime is $139/year in the US today and I think >65%(!) of US households are subscribers.

Set the Prime+ pricing at something like $299/year or $30/month. Make agentic purchasing exclusive to this tier. This solves multiple problems at once:

• It gives Amazon a way of saying yes to its most loyal customers

• It provides replacement revenue for any potential lost advertising revenue

• It lets Amazon dip its toes into agentic use cases without betting the farm

Amazon doesn’t even need to market this aggressively to customers – leave that job to tools like Muse and Instinct: “To order from Amazon, you need to upgrade to Amazon Prime+ for $40/month or $399/year – do you want me to upgrade your plan?”

I think Amazon would be surprised by the uptake.

There is already some precedent for this. What is YouTube Premium? It is a way for consumers to pay for the convenience of an ad-free YouTube experience. Amazon Prime+ is the same thing – a way for consumers to pay for the convenience of an ad-free, agentic Amazon experience.

YouTube Premium customers are some of YouTube’s most avid, loyal customers. The same will be true of Amazon Prime+. Every consumer product developer knows decreasing friction increases conversion – I suspect that enabling agentic use will meaningfully increase Prime+ ARPU, too.

Uber, DoorDash and Instacart each have their own loyalty programs with similar dynamics (Uber One – $96/year, DashPass – $96/year, Instacart+ – $99/year). I don’t know if they share Amazon’s trepidation about enabling agentic use. But if they do, this offers a similar path forward.

Is this the market equilibrium?

I don’t know.

If I were a competitor, I would optimize for agentic use to try to take market share. Markets don’t reconfigure often.

But here’s the thing – I don’t want to leave Amazon. Would I prefer to just have them enable agentic shopping for free? Of course. But if given the choice, I would rather just pay more to have the Amazon experience within experiences like Instinct and Muse than explore other shopping tools.

Amazon Prime led to the most engaged customers buying more, more often. I think there’s a good chance that Prime+ will do the same. Maybe the marginal user will give Walmart a try, but I suspect the incrementality of Prime+ might outweigh the loss from marginal switchers. It seems worth a try.

Divine Discontent

This is the crux of the innovator’s dilemma – you’re damned if you do, damned if you don’t. I understand why marketplaces will want to stick their head in the sand. Apple has been fighting tooth-and-nail to keep the App Store tax for almost twenty years.

But I think there is another way for now. Introduce a new, premium tier. Enable agentic access there first. Keep your most active, tech-forward customers happy while you let the broader market shake out. I am confident that both (a) horizontal, personal agents are the way forward and (b) beloved consumer services like Amazon will find a way.

Consumer expectations are not static – they only ever go up. Those that meet these ever-increasing expectations are the ones that will ultimately win.