Walmart swears one price for all and McDonald's lets AI personalize the menu: a warning about shopping with agents

By Zendoric · October 6, 2026.

Tech investor Howard Tullman writes a column in Inc.com (October 5) that pits two pricing philosophies against each other in the era of AI-agent shopping. Walmart, as the author notes, holds that a product's price is the same for every customer. McDonald's, by contrast, uses AI to recommend personalized menu prices that the chain itself admits are based on transaction data and on each area's willingness to pay. Tullman calls that practice "predatory"; it should be read as his editorial opinion, not as a regulatory or legal conclusion about McDonald's.

The backdrop to that comparison is the flood of AI assistants that tech companies and startups have begun to offer: Tullman cites Meta's Muse, OpenAI's dots and a startup he does not clearly identify ("Instinct AI", he says, founded by a 23-year-old with multimillion-dollar funding), all of them requesting broad data permissions and terms of use that almost nobody reads. To this he adds a striking figure: non-human actors —bots, scripts, AI agents— already generate around 55% of global web traffic, according to the number he uses in his column. Jeff Bezos himself, the author recalls, has argued that consumers do not want to negotiate the price every time they buy; the implicit reading is that we have to trust that whichever algorithm is in play will get us the best deal.

That is the real problem, and it is more structural than anecdotal: when your shopping agent sits down to negotiate, it is not doing so against a neutral seller, but against a pricing system optimized with your own behavioral data. This is not a negotiation between equals. It is your assistant, trained to save you time, against a pricing engine trained to maximize what you, specifically, are willing to pay. Whoever has more data and computing infrastructure —almost always the seller, not the buyer— starts with a structural advantage, however much the marketing talks about "your personal agent".

In the short term, this is a risk that should not be downplayed: price personalization can turn into price discrimination dressed up as convenience, and the flood of opaque consents these assistants demand erodes a privacy we have already been giving away in exchange for convenience for years. As sector context, dynamic or personalized AI pricing has been under regulatory scrutiny in various markets for some time, precisely because nobody audits which variables the algorithm uses or whether it discriminates by ability to pay, geography or consumption profile. Walmart and McDonald's represent two different brand bets here —transparency versus margin optimization— in a field where customer trust is beginning to be, in itself, a competitive advantage.

Our reading is that agentic commerce is not bad by design; it is blind by design unless the asymmetry is corrected. If both sides of the transaction —buyer and seller— operate with agents of comparable capability and under verifiable transparency rules, the friction of comparing prices, reading the fine print and wasting time on repetitive decisions can genuinely be reduced, freeing up that time for whatever each person values. That is the underlying promise of the world of abundance toward which, we maintain, AI is taking us in the long run. But that promise does not fulfill itself: it is being decided now, in how we design these agents, what data companies can use to set prices and what guarantees the user has that their assistant is working for them and not, unknowingly, for the seller across the table.