Zero Clicks #15: Amazon’s AI Conundrum

Just a quick note before we dive into this week’s topic. We’ve built a nice little community here and are planning an in person get together on December 10th in NYC - I’ll be there along with the Martech Record crew - we’ll have cocktails and cookies for everyone - along with some brief remarks from a few industry leaders.

I have been writing this missive for the better part of the year and have had a lot of great interaction via email. But it would be great to meeting in person. Martech Record has gifted me a few free passes to hand out. If you guess the State of the Chart (bottom of the email) correctly, I’ll send you one. If you can’t guess, then you can register the old fashioned way, here. Would love to meet you in person there.

- Mike Mallazzo 

LLMs vs. ads, AppLovin keeps ripping, and consultants are here for your retail media dollars.

The age of AI product search has officially arrived. Search GPT can now answer highly complex product queries frighteningly well, featuring the kind of nuance that was once the exclusive domain of commerce publishers (and possibly reliant on their scraped data).

Perplexity is cooking up something called Pro Shop, a dedicated commerce feature with the humble goal of taking on Amazon. And oh yeah, the retailer responsible for 40% of all online commerce in the US is fundamentally rebuilding its search experience around AI.

The old adage about Amazon was that it’s a search engine with a warehouse attached. Today, the business is realistically more of an advertising exchange masquerading as a marketplace. This creates a major challenge for Amazon as it embraces LLM-led shopping experiences.

AI is the most obvious place where the everything store is up against the full weight of the innovator’s dilemma. Amazon’s business today is essentially geared towards showing as many product recommendations and ads as humanly possible, in a cluttered interface that cuts against the grain of where genAI wants to pull retail. 

Amazon has every conceivable training data advantage known to man and their first deployments of Rufus were almost comically bad. Still, that isn’t stopping Amazon, a company that generally feels like it is besieged by a day 2 mentality everywhere else across its business, from thinking first principles about product discovery.

Driven largely by increases in its advertising business, Amazon had a monster Q3 allowing the retailer to increase capital expenditures to $75B, which can mostly be attributed to AI development. Ads are funding AI development that makes the future of Amazon’s advertising business considerably murkier.

Rufus, of course, can and will sell ads- in the form of sponsored answers to shopper queries. Moving discovery to LLMs doesn’t eliminate advertising altogether, but it’s a much more minimalist form factor.

Rufus is improving, even if its bravado is running ahead of its recommendation quality. The tool is showing noticeable conviction in its product recommendations and beginning to support the types of semantic, natural language queries that were previously Google’s exclusive domain. Answering these types of searches requires a Herculean investment in high-quality content to properly capture the qualitative attributes of hundreds of millions of products.

To that end, Amazon recently hired Shareen Pathak, a longtime journalist and brilliant editorial product mind, to lead “editorial development for genAI.” Imagine hearing that job title even two years ago.

Amazon, of course, once had a pipeline of human-generated, “expert” product review data in the form of the Onsite Associates program, a five-year experiment to embed expert review content from media publishers in their product search experience that was shuttered last year. 

In and of itself, this was not particularly big news outside of the hardcore Amazon ecosystem. The Onsite Associates program was highly opaque, quietly powered by a network of third parties, and lived in a weird sort of strategic limbo at Amazon. While an estimated 25% of product searches were served editorial recommendations, Amazon never seemed truly sure what to do with the program.   

In sunsetting Onsite Associates, Amazon effectively removed a cost center to their business and replaced it with more juicy ad real estate. They also potentially killed a valuable data pipeline for precisely the AI-driven search experience they are now rolling out.

From the perspective of brands who sell on Amazon, the retailer’s AI embrace comes at an interesting inflection point where the cost of sponsored product ads has become largely untenable for many. But will brands that sell on Amazon benefit from the retailer shifting to an LLM-driven experience? In the short term, it’s hard to imagine that a potentially black box process with less available inventory than conventional product search will be beneficial to brands.

Zooming out, there is an emerging paradox for Amazon and other large marketplaces that are increasingly reliant on advertising revenue. Even if LLMs prove they can objectively offer a better experience for shoppers, they might result in less overall yield for retailers, who now have less surface area to show additional advertisements and product recommendations. How long is Amazon willing to invest in an experience that is less profitable to play the long game of where commerce might be headed?

With the stock near all-time highs, Temu mania dying down, and Lina Khan likely on the way out of Washington, Amazon looks less vulnerable than ever. But historically speaking, these sorts of existential business conundrums for incumbents are the kindling in which challengers catch fire. 

Head of Business Development (China), AppLovin

I’m running out of superlatives to describe the tear AppLovin is on. The stock nearly doubled last week, is up 600% on the year and its $95B+ market cap now makes it worth more than The Trade Desk. What. The. Hell.

Amazingly, the hottest company in advertising only has 23 (!) open roles posted across its entire business, with many of those based in China to build out operations selling its new eCommerce ad offering directly to Chinese-based brands.  In VP of global BD Rafael Vivas’s words, the app-to-web advertising product that launched less than a year ago is “already dominating with DTC brands in the U.S” and is now looking to replicate that success with deep-pocketed Chinese sellers.

Courting Chinese cash directly is going to be the name of the game for challenger ad tech companies in 2025 as Meta, Amazon, and TikTok have been on this grind for years.  Meta will likely clear $10B in ad revenue this year from Chinese brands and Amazon could easily double that. 

Mark my words, fortunes will be made in this arena.

GTM Lead, Commerce Media, Accenture Song

Retail media is growing up and as it hits critical mass, the consulting-industrial complex is arriving in force to get a piece of the action.  

As with any consulting job, it’s hard to tell exactly what the remit is here but given the number of openings that Accenture has in Bentonville, it seems likely that there’s a connection to selling Walmart Connect into large brands. From a job seeker’s point of view, this is a great gig. Positioning yourself as a “thought leader” in retail media and getting wide exposure to several enterprise commerce media deployments in 2025 is rock-solid career positioning.

Remember, in any gold rush- AI, retail media, or otherwise, the consultants always win.

In case this hasn’t been clear to date, all of the memes you see in Zero Clicks are originals. We can’t write a newsletter bemoaning the amount of derivative content on the internet and then copy memes :)

In 2018, this company was ranked the “best stock to buy and hold for 10 years” in a CB Insights poll. It’s down 50% over the last 5 years and almost 60% from when that piece was published. Name the mystery company.

Reply directly to this email with your guess. First one to guess correctly (without help from AI) gets a shoutout next week (plus unlimited bragging rights in their company Slack).