(From my Marketing Week column).

“Your package was shipped: 1 Nutrition & Wellness, 1 Wireless Accessories.” Some Amazon customers are wondering why their confirmation emails suddenly read weird. Didn’t I order one shampoo and one USB charger? It’s a sign of things to come. Asked about the change, Amazon told a US reporter: “This also reduces customer information shared outside the Amazon app and website to further improve customer privacy.”

When big tech suddenly discovers “customer privacy”, it’s worth digging in.

Amazon has good reason to “protect” its customer data. AI is beginning to squeeze itself between brands and customers. The more data AI systems gather, the easier it becomes for them to advise customers on what to buy—and eventually make the purchase decision.

AI decision making sounds futuristic. It isn’t. Gemini Spark, Google’s personal AI agent, can scrape information from Gmail and other Google services to, for example, compile itineraries and build trips. In May, Google unveiled a “Universal Cart” that works across Google apps. Nike, Sephora, Target, Walmart, Wayfair and others have signed up. Amazon, notably, hasn’t.

For customers, the attraction of an AI interface is obvious. Let the system deal with your purchase of “four packs of kitchen towels on promotion.” Fine. Let it find your favorite holiday package cheaper. Even better. Customers like things to be easy. DoorDash and Deliveroo have already made shopping easier. If AI removes another step, consumers will use it.

For brands, an AI intermediary could range from a business accelerator to a complete disaster.

Brands have always had intermediaries. Most consumer brands never enjoyed much of a direct customer relationship in the first place. People buy juice and diapers through retailers, physical or online. Brands create consumer pull through advertising and retailer push through promotions. It’s „pay for salience“. Service brands have traditionally been closer to customers. Take travel. Traditionally, people booked holidays through an agent. Digital allowed the best travel firms build even stronger customer relationships and get people to book before they shop around. Others ended up paying Google and TikTok ever larger sums to send customers to their websites.

Now a new intermediary is arriving. If AI starts buying, recommending and assembling products for us, brands and retailers enter a different marketing game. Strong brands should have an advantage. So should those with excellent direct customer relationships. Everyone else risks becoming a row in a comparison table: “sort by price.”

B2B isn’t immune. Historically, long negotiation cycles and complicated contracts made great salespeople enormously valuable. Digital already made suppliers easier to find and compare. Now imagine AI running much of the procurement process. If AI systems are good at anything, it’s digesting complicated problems, comparing options and checking complicated contracts.

Which brings us to the uncomfortable question. Will AI turn the customer interface into something brands eventually have to pay for? There’s a very large reason to think it will. Money.

Current AI revenues don’t remotely justify the capital being invested. Infrastructure spending by the biggest US tech firms is projected to rise from roughly $450 billion in 2025 to around $1.4 trillion in 2027. The Economist has calculated that covering AI capital expenditure would require annual AI revenues on the order of $2.5 trillion. That’s more than the entire tech industry generates today.

Meanwhile, CEOs aren’t yet seeing AI productivity gains at scale. Some companies are even cutting back on AI spend. Microsoft is now cracking down on token overuse among its own staff. And most consumers still use AI for free. Wonderful technology. Terrible economics.

Someone has to pay.

For AI firms, finding much larger sources of revenue is becoming existential. Google became the world’s largest advertising business after discovering how valuable control of the consumer interface could be. AI may be about to discover exactly the same thing. Once AI systems begin influencing which products consumers see, compare and ultimately buy, that access becomes extremely valuable.

Monetization is already happening. 200 days after introducing ads, Open AI just reported 1bn US$ Chat GPT advertising revenue. A press release hints at more: „Our next phase of growth will bring ChatGPT Ads to more markets and introduce additional formats, objectives, buying options, and measurement capabilities.” Gemini is experimenting with AI shopping. Anthropic has told the world during the Super Bowl that Claude will remain ad-free. But ad-free doesn’t mean monetization-free. There are other ways to charge brands for access, transactions, services, or placement.

Yet many executives are looking in the wrong direction. Around the world, they are flocking to AI tool courses just as they flocked to digital tool courses twenty years ago. Everyone worries about AI replacing basic activities such as Seach Machine Optimization (SEO) or social media campaigning (it probably will). And everyone wants to become an expert in the new tools.

There’s much bigger question: when machines squeeze between us and our customers, how do we maintain those relationships? When the machine recommends, how do we become the brand people still want?

Executives need to step back from the tools and start doing their actual job. Help the organization work out how to thrive in an AI-mediated world. Decide where AI intermediaries can help the business, where the direct customer relationship must be protected, and where the company risks becoming interchangeable.

Knowing AI tools is tactics.
Knowing how to win as AI changes the game is strategy.