When the buyer is a bot, unreadable product data becomes an out-of-stock

Brett Davis, US Chief Innovation Officer for Deloitte, argued in a Forbes Tech Council column that consumer AI has crossed from assisting decisions to making them on people's behalf, and that being absent from an agent's consideration set is becoming the new version of being out of stock ^[1][2][3]. If that framing holds, product data stops being an input to quarterly reporting and starts behaving like shelf space.

Davis names three determinants of who wins: whether a large language model can read, understand and trust a brand's data, whether the brand's value can be quantified in terms an agent can compare, and how fast the organisation can act on both ^[4]. Two of the three are questions about machine legibility. That is closer to a supplier onboarding checklist than a marketing brief, and it is why he says the shift changes channel strategy and how tightly supply chain planning has to integrate with agent engagement ^[5].

The budget consequence is the sharper part of his case. Most organisations, he writes, still spend on rear-view analytics to explain the past and forecast the future, while agentic commerce demands forward-looking signal data that most brands do not possess ^[6]. He calls the change potentially bigger than the move to e-commerce 25 years ago ^[7], which is the sort of scale claim a consultancy selling the transition will make. The underlying mechanism is duller: feeds, attributes, provenance.

The demand-side evidence is mixed, and Davis cites it as such. Prophet's 2026 AI Powered Consumer Research found 54 percent of consumers say autonomous agents acting on their behalf would be helpful ^[8]. YouGov's study of American trust in AI for retail found 65 percent trust it to compare prices but only 14 percent currently trust AI to place an order ^[9][10], a gap of 51 points ^[11]. So delegated checkout is not the near-term exposure. The shortlist is, because agents already shape what consumers see, compare and consider ^[12].

On trust, Davis argues that as AI-generated reviews and synthetic engagement proliferate, provenance becomes essential: participants need to know where a data point originated and whether it can be traced and audited ^[13]. He adds a durability test, noting that many reviews and sentiment signals sit unrefreshed for months or years ^[14], and that data which is not human-verified, always on, auditable and predictive of purchase intent becomes a liability agents will learn to discount ^[15].

The value argument is where the reader should apply the most discount, because the instrument is the author's own. Deloitte's research on the value-seeking consumer holds that between 10 and 40 percent of what drives perceived value has nothing to do with price, and is instead quality, trust and attitude ^[16]. That underpins Deloitte's More-Value-for-the-Price framework, under which only about one in three brands studied qualifies ^[17][18], and Deloitte's Converge Data Signals platform shows consumers moving from lower-value brands to those brands ^[19]. The same framework sits behind the Forbes Best Brands for Value list, built with HundredX from more than 160,000 consumers, 4.7 million ratings and more than 5,500 brands ^[20] -- roughly 850 ratings per brand ^[21]. A proprietary framework, a co-published list, and a share-shift finding measured on the vendor's own platform.

Worth tracking: whether the 14 percent order-placement figure moves in the next wave of the same surveys; whether agent platforms publish attribute and provenance requirements that brands can be graded against; and whether anyone outside Deloitte reproduces the claim that value-defined leaders are taking share.