Notes from the Digital Floor 7. Why Online Retailers Are Going to Prison Now.

Notes from the Digital Floor 7

Why Online Retailers Are Going to Prison Now.

"Don't go to prison voluntarily!" Alexander Graf advises retailers in his recent brand eins interview¹, referring to Google's Universal Commerce Protocol. The image is memorable, the warning clear. And that is precisely the problem, because what initially reads like a precise diagnosis of the market unfolds, on closer inspection, into a dynamic where distinct levels are collapsed into one another in ways that generate urgency where differentiation would be required first.

The interview addresses virtually every relevant trend in the industry simultaneously, from agentic commerce through Chinese platforms to the transformation of the working world in online retail. The individual observations are not unfounded, but the manner in which they are connected produces a picture that appears considerably more coherent than reality warrants. What follows is an attempt to examine the central claims individually and to sharpen the focus where generalisation obscures the actual state of affairs.

UCP as prison, but who isn't already inside?

The core thesis of the interview holds that UCP strips retailers of all agency and leads them into total dependence on Google. As a warning, that is understandable. As analysis, it overlooks a central fact, for the overwhelming majority of retailers already find themselves in precisely this dependence. Anyone selling on Amazon today, reliant on Google Ads or purchasing visibility through Meta, is already operating within structures they do not control. UCP shifts this asymmetry. It does not create it.

The more interesting question is what UCP actually changes. The protocol is designed so that the agent (in this case Google's Gemini) handles the entire purchase process, from product search through comparison to payment. The retailer formally remains the seller (Seller of Record) but loses all contact with the customer, and with it the ability to build relationships, collect data or influence repurchase. That is a meaningful loss of control. But it primarily affects those retailers who have never built a direct customer relationship in the first place, and there are a great many of those.

The real question, therefore, is not whether UCP is dangerous, but for whom it is worth resisting, and for whom the ship has already sailed.

The urgency that isn't

Graf himself concedes that agentic commerce will likely become relevant "in five to ten years rather than two to three."¹ At the same time, he advises waiting "until the very last possible day."¹ These statements sit in a remarkable tension with one another, for if the time horizon is genuinely that distant and the customer need still unclear, where precisely does the acute urgency lie?

Small and medium-sized retailers without clear differentiating features or customer advantages face diminishing viability in digital commerce with or without UCP. This is not a new phenomenon and not a symptom of a protocol unveiled in early 2026. It is the result of a concentration that has been intensifying for years, in which platforms control the customer interface and retailers without their own brand, exclusive assortments or demonstrable customer value systematically become interchangeable.

That customers increasingly buy directly from brands and manufacturers via marketplaces is likewise not a UCP-specific development. This trend has been measurable on Amazon for years and is being actively driven by direct-to-consumer strategies from numerous manufacturers. UCP may accelerate this tendency. It does not cause it.

Equally undifferentiated is Graf's claim that customers only want "cheap, cheap, cheap" and that a single euro of markup is grounds for abandonment.¹ This is a generalisation that, in its sweeping form, is not supported by the evidence. A study by the Qualtrics XM Institute found that 72 per cent of consumers are willing to pay more for a better experience.² The Capgemini Consumer Trends Report 2026 concludes that price alone no longer defines perceived value, with quality, trust and emotional connection becoming increasingly decisive in purchasing decisions.³ What Graf describes applies to generic, interchangeable assortments where price is indeed the only differentiator. But retailers who can offer genuine added value, whether through advisory expertise, curated ranges, exclusive products or a superior shopping experience, need not fight over every euro. Equating price sensitivity with the entirety of consumer behaviour mistakes a segment for the market. And it ignores that differentiation is precisely the only sustainable response to platform dominance, regardless of whether UCP materialises or not.

Amazon's answer already exists

One of the interview's most striking claims is the assessment that UCP poses a "massive threat" to Amazon and that the company is "increasingly running out of answers."¹ This reading considerably underestimates the strategic positioning.

Amazon has already implemented its own AI assistant, Rufus, which since 2024 has been progressively expanded from a conversational product adviser to a fully agentic system.⁴ The "Auto Buy" function now enables Rufus to purchase products automatically based on price targets, and the "Buy for Me" function allows the system to shop at third-party retailers outside the Amazon ecosystem.⁴ Amazon is not watching passively but pursuing its own agentic commerce strategy, built on an advantage no other player possesses, namely a functioning logistics network, an established payment infrastructure and the purchase histories of hundreds of millions of customers.

That Amazon made an unconvincing attempt with Haul to compete with Temu in the ultra-low-price segment is undisputed. To conclude from this that the company is generally running out of answers, however, confuses a tactical experiment with a strategic position. In 2026, Amazon has further shortened delivery times in the US alone, expanded Prime membership to new services and deepened AI integration into the core purchasing process. These are not the hallmarks of a company without a plan.

What agentic commerce actually means

At another point in the interview, a more fundamental problem surfaces, namely the definition of the very thing being discussed. Asked about competition from feed-based providers such as TikTok Shop, Temu and Shein, Graf distinguishes between agentic commerce as a "conscious selection process" and the feed model, in which users "don't need to input anything at all."¹ The contrast sounds intuitively plausible but contradicts the concept of agentic commerce at its core.

Agentic commerce does not mean that customers select consciously and actively. It means that an AI agent assumes that selection autonomously. Amazon Rufus purchases automatically on the basis of price targets.⁴ Google's UCP envisions Gemini handling the entire purchase process, from search through comparison to payment. The user delegates the decision to the agent, and it is precisely this delegation that constitutes the disruption, not a new interface for deliberate shopping.

Structurally, feed-based shopping and agentic commerce therefore have more in common than Graf's distinction suggests. Both reduce active user decision-making, merely through different mechanisms. The algorithmic feed curates by steering attention; the agent curates by acting autonomously. In both cases, control over product selection shifts away from the customer.

That this conceptual imprecision appears in one of the industry's most widely noted interviews is rather telling, because it illustrates how swiftly a category becomes a screen for projection before its actual mechanics are properly understood.

China as opportunity, but for whom?

At scarcely any point in the interview does the gap between observation and recommendation become as visible as in the passage on "opening towards China."¹ Graf recommends partnerships with Chinese retailers who require a "cultural translation" to sell in Europe. This sounds initially plausible but underestimates both the speed at which Chinese players are delivering that translation themselves and the structural barriers facing European retailers.

The data paints a sober picture. More than 57 per cent of active sellers on Amazon.de are now based in China.⁵ In 2025, JD.com acquired Ceconomy AG (MediaMarkt, Saturn), the first major takeover of a European retailer by a Chinese company, and is building its own European last-mile delivery network through JoyExpress.⁶ DHL has formalised a trade corridor with JD.com between China and the EU.⁷ Platforms such as Temu and Shein have built European market shares within a few years that no local competitor has achieved in comparable time.

To recommend in this context that European retailers should "seek partnerships" presupposes that these retailers have something to offer which Chinese providers cannot develop or acquire themselves. For some specialised firms with deep market knowledge, regulatory expertise or exclusive market access, that may hold true. For the average online retailer with a generic assortment and limited organisational depth, it is an abstraction bearing little resemblance to operational reality.

The working world in two years?

The most far-reaching claim, and one supported only in passing, concerns the working world. "In two years, most of those jobs won't exist any more," Graf says of computer-based roles in retail.¹ This is a statement whose absoluteness is supported neither by current data nor by the observed pace of AI adoption in organisations.

What can be empirically observed is this. A CFO survey from March 2026 indicates that AI-related redundancies this year are expected to be nine times higher than the previous year, yet still represent a fraction of "doomsday predictions."⁸ Only around 44 per cent of surveyed chief financial officers are planning AI-related headcount reductions at all.⁸ Industry analyses project that by 2030, approximately 85 per cent of customer interactions in retail will be AI-supported, albeit with considerable variation by segment and region.⁹

The line "I reckon you could pull half the people out of any business with computer workstations right now without anyone noticing"¹ is a provocation that works rhetorically but lacks empirical foundation. In most organisations, full automation fails not for want of tools but because of processes, data quality, regulatory requirements and the sheer complexity of structures that have grown organically over years. The transformation is coming, but it arrives gradually, unevenly and considerably more slowly than the tenor of the interview suggests.

The illusion of optionality

What unites all these claims is a pattern remarkably prevalent in E-Commerce discourse. Observable market movements are used to derive a kind of universal optionality, even though the reality of most companies is shaped by considerably tighter operational, organisational and economic boundaries.

An option is not an abstract market movement but a concrete possibility to act within one's own structure, and that structure is rarely as flexible as public debate implies. The result is a form of strategic overload that arises not from the complexity of the market itself but from the way it is spoken about, since the simultaneity of urgency and imprecision places organisations in a state where decisions are either taken in haste or not taken at all.

Breaking through this dynamic requires starting from a different point entirely. Not at the next supposed opportunity, but at the question of which options genuinely exist. And that question seldom leads to spectacular answers. It leads to a rather sober examination of one's own position in the market, one's own capabilities and the limits of what can realistically be changed, which often means that the strategic space does not expand but initially contracts.

Yet precisely this contraction is not a weakness but the precondition for robust decisions, because it makes visible the difference between what appears possible and what is actually achievable.

In an environment shaped by ever-new narratives about possibilities, clarity does not emerge from additional options but from their reduction.

And therein lies the real counterpoint to a discourse that instantly derives an opportunity from every development.

Not every movement in the market is an invitation to move.

And not every theoretical option deserves to become a strategic consideration.

Alexander Graf's powers of observation and his willingness to raise uncomfortable subjects publicly are, on balance, a service to E-Commerce discourse. The industry needs voices that generate movement and challenge certainties. But it needs those voices in a form that genuinely helps the retailers affected, rather than plunging them into a reactive panic whose operational foundations remain unexamined. Less theatrics and more contextualising substance would do greater justice to the gravity of these topics, for the questions Graf raises are too important to be subordinated to rhetorical effect.

Sources

¹ Alexander Graf, interview with brand eins: "Alexander Graf geht nicht freiwillig ins Gefängnis", brand eins Wirtschaftsmagazin, "Radikal denken" issue, 2026. brandeins.de

² Qualtrics XM Institute: "ROI of Customer Experience", 2024. The study finds that 72% of consumers are willing to pay more for a better experience.

³ Capgemini Research Institute: "Consumer Trends Report 2026". The report concludes that price alone no longer defines perceived value.

⁴ Amazon: Rufus AI Shopping Assistant, progressively rolled out since February 2024 with "Auto Buy" and "Buy for Me" functions introduced during 2025.

⁵ Marketplace Pulse: "Amazon Marketplace Seller Data", 2025. More than 57% of active sellers on Amazon.de are based in China.

⁶ JD.com: Acquisition of Ceconomy AG (MediaMarkt, Saturn) and establishment of the European delivery network JoyExpress, 2025.

⁷ DHL Group: Formalised China-EU trade corridor in partnership with JD.com, 2025.

⁸ BDO CFO Outlook Survey, March 2026: AI-related redundancies projected at nine times the previous year's level, yet only 44% of CFOs planning corresponding headcount reductions.

⁹ Juniper Research / Gartner: Forecast that approximately 85% of retail customer interactions will be AI-supported by 2030.