Stratechery Author Ben Thompson: Consumer AI Has a Marketing Problem Because Ordinary Consumers Don't Really Want to Be More Efficient
Ben Thompson, author of Stratechery, stated in a program that consumer-grade artificial intelligence has a marketing problem because ordinary consumers do not truly want to become more efficient.
He used booking flights as an example: tech companies always demonstrate "let us book your flight for you," but he has employed a personal assistant for ten years and has never let them handle bookings. "That's nonsense." He knows which flights will be delayed and which seat to choose, and he finds these demonstrations laughable.
The second judgment is that people enjoy shopping and do not want to be pushed to improve efficiency; they want to be entertained instead. Shopping itself is one of the greatest forms of entertainment. Product pages are filled with information, and companies invest a lot of effort into this. Clothing items specify the size the model is wearing and showcase different styles, details that consumers want to see for themselves.
He describes the assumption of agent-assisted shopping as a flawed interaction hypothesis: the agent retrieves some information, and the user immediately asks another question; tech executives say, "I have an assistant who can handle that." Thompson's response is that he not only wants to do it himself but also wants to complain while doing it, enjoying the act of complaining and wanting to live his own life.
This argument targets the standard demonstrations of current consumer agent products—booking, price comparison, and placing orders. Thompson considers preferences, seat selection, delay judgments, and browsing product pages as non-outsourcable aspects of consumption, rather than friction to be eliminated. Having an assistant manage backend tasks for ten years while retaining booking rights indicates that he distinguishes between "handling my chores" and "making choices for me."
In market mechanisms, this is a demand-defining event, not a buy-sell transaction. Buyers still need to tell the story of agent e-commerce model companies and platforms; seller pressure falls on retailers who build conversion rates on product pages, filters, and shopping experiences. If consumers view shopping as entertainment, traffic and advertising budgets will remain on interfaces that are browseable, comparable, and complainable, rather than handed over to one-time purchasing agents. Benefiting are brands that manage product details, sizes, and visual content as entertainment, while consumer AI products that treat "placing orders for you" as a default value proposition are under pressure.
On a supplementary level, he has long used personal assistants for emails, taxes, and task systems, rather than travel decisions; this aligns with the path of making AI into coding, customer service, procurement, and other B-end agents, where efficiency is the demand, while the consumer side may first require richer interfaces rather than shorter checkouts.
Source: Public Information
ABAB AI Insight
Thompson's analytical framework has consistently placed aggregators and consumer surplus at the center: who controls the demand entry, who extracts the profits. This time he rewrites the entry question from "who searches" to "do consumers really want to hand over choices?" Booking demonstrations eliminate friction in the lab, but in his life, they eliminate judgment and complaint rights. The fact that he has only used an assistant for backend tasks for ten years shows that high-end users have already separated out tasks that can be outsourced from preferences that cannot, while consumer AI has also written the latter as a list to be automated.
Capital continues to flow into agency commerce: model companies, payment companies, and e-commerce platforms are all vying for the "last click order." The logic of money is to compress the funnel and take retail advertising. The resistance Thompson points out is that the funnel itself is the product—the size, style, comparison, and complaints on clothing pages constitute the duration of stay. Retailers like Walmart have already seen public discussions about conversion declines after agency intervention, indicating that adding an extra layer of "helping you buy" can dismantle pages optimized by merchants for ten years without necessarily improving transactions.
Similar splits appear in voice assistants promising to handle life but stopping at timers and music; they also occur in travel super apps wanting to replace price comparisons, while users still open multiple tabs. The industry is at a stage where corporate and coding agents are expanding, while consumer agents repeatedly demonstrate but struggle to change habits. Consumer companies like Airbnb have publicly stated that chat boxes are not the interface for travel and e-commerce; they require richer controls and comparisons.
The essence is that pricing power remains on the entertainment side, not the efficiency side. The mechanism is that the utility of consumer decision-making includes selection, comparison, and expressing dissatisfaction; removing these turns the product from entertainment into a task, reducing the number of people willing to pay. Tech companies sell based on productivity, while consumers pay based on experience; this mismatch can lead marketing budgets to be wasted on unwanted "living for you" services, while what can truly scale is helping people enjoy the information they want to see more.