Skip to main content
Welcome to Deloitte

If we have selected the wrong experience for you, please change it above.

Abstract digital artwork features a luminous circular ring surrounded by reflective spheres. The scene is set against a dark background with vibrant gradients of purple, blue, and orange light. Smaller glass-like orbs rest on the surface.

The invisible shelf

How agentic commerce rewrites the rules for CPG brands

Authors:

  • Gaetano Rinaldi | Partner – Strategy & Business Design
  • Amandine Gillet | Partner – Tech & Transformation – Customer
  • Giulia Pescatore | Director – Operations Excellence
  • Laura Mathieu | Senior Manager – Customer Strategy & Design
  • AI agents are starting to decide what gets seen and what gets skipped. For consumer packaged goods (CPG) brands, that changes the rules of discovery, evaluation, and purchase.
  • New pressure points are emerging across the purchase journey. Product data, visibility, comparison, and customer access can shape whether a brand is chosen or quietly replaced.
  • Some products face far greater exposure than others. The article explores how to identify where the value at stake is highest and where brands are best positioned to compete.
  • The challenge goes well beyond technology. It calls for closer coordination across content, data, commerce, media, and technology teams.
  • The market is moving before a clear model has emerged. Retailer-led, AI-led, and interoperable approaches are all taking shape, creating both risk and opportunity for brands that act early.

Introduction

For decades, CPG companies have competed by building brands, securing shelf space, optimizing packaging for the human eye, and using trade promotions to influence purchase decisions. Whether on a physical shelf or a digital product page, the formula has remained largely unchanged: capture consumer attention, trigger emotional affinity, and convert that affinity into selection.

AI agents are disrupting that formula. As they increasingly mediate product discovery, evaluation, and purchase, an invisible shelf is emerging, where algorithms determine which products consumers consider.

The shift is already visible. More than half of European consumers already use AI for shopping, with 80% reporting a better experience,1 while AI agents influenced 20% of global online Cyber Week orders in 2025.2

The United States is leading early experimentation, with large retailers and platforms deploying proprietary shopping agents at scale. Europe is close behind, driven by rapid consumer adoption and growing retailer investment, as leading players begin to embed agents into their digital experiences.

Much of the discussion about agentic commerce has rightly focused on the implications for retailers and platforms, those who own the transaction and the customer relationship. But for CPG companies, the stakes are arguably more acute. Retailers control assortments and fulfilment infrastructure. Large language model providers control algorithms. CPG brands, by contrast, have historically relied on the strength of their consumer relationships and the persuasive power of their marketing. In an agent-mediated world, both of those advantages must be re-earned in an entirely new language, one spoken by machines.

This article explores where brand equity is most exposed, how CPG companies can identify the products and categories that matter most, and what organizational changes are required to compete on the invisible shelf.

The four vulnerabilities CPG brands cannot ignore

The path from consumer intent to product purchase has always involved four broad gates: awareness, consideration, evaluation, and selection. CPG companies have built sophisticated capabilities to influence each one. Agentic commerce does not remove these gates, but it changes who navigates them and what determines success.

This creates four structural vulnerabilities for CPG brands:

Before an AI agent can recommend a product, it must be able to interpret what it is, what it contains, who it is suitable for, and whether its claims are verifiable. Poor or inconsistent product data is not a new challenge, but AI agents amplify its consequences. What once resulted in a weaker website or e-commerce listing can now exclude a product from consideration entirely, creating an invisible out-of-stock: the product exists in the supply chain, but not to the agent.

Being machine-readable does not guarantee being surfaced. Each AI platform applies its own logic to determine which products appear in response to a given request, with brands having limited visibility into when, how often or why their products are recommended. As a result, influence shifts away from optimizing individual keywords towards strengthening the underlying signals agents use to rank products, which are explored in the following sections.

Once a product enters the consideration set, agents can compare alternatives against criteria such as price, ratings, ingredients, certifications, availability, and delivery speed. Packaging, displays, and other traditional point-of-purchase cues carry less influence at this stage. The agent compares on data, and the product that scores highest on the consumer's stated or inferred criteria wins.

As AI platforms and retailers increasingly mediate discovery, recommendation, and transaction, the CPG brand's direct relationship with the consumer weakens. Pricing power may erode as agents optimize for value, consumer data increasingly sits with platforms, and brands have less influence over cross-sell, up-sell, and basket composition.

Consider a routine replenishment mission: a parent asks an AI agent to “reorder the same hypoallergenic laundry detergent, but in a larger size.” The agent must determine which products meet the requirements (eligibility), identify those it can find and interpret across retailer and platform catalogues (discovery), and compare them on price, certifications, reviews, and delivery speed to decide whether to retain the incumbent brand or substitute a higher-value alternative (selection and control). At each step, the brand’s underlying product data and its integration across retailer and platform ecosystems influence whether a brand is surfaced, selected, or silently replaced.

Not all products are equal: A framework for prioritization

CPG organizations do not need to transform every product, category, and market simultaneously. Exposure to agentic commerce varies across the portfolio, and resources are finite. The priority is to identify where action will create the greatest impact.

Two dimensions matter most:

This assesses how exposed a product or category is to agentic commerce, considering purchase frequency and replenishment potential, suitability for agent-mediated shopping (mission-based purchasing, comparison-heavy decisions, subscription models), current AI adoption among category shoppers, and strategic importance to the broader portfolio. Early evidence suggests this impact is already significant: nearly a quarter of consumers plan to make AI‑mediated shopping their default behavior within the next year, shifting a growing share of demand into agent‑controlled pathways.3

This assesses how ready a product is to compete in an agent-mediated environment, based on product data completeness and consistency, the quality of retailer feeds and catalogue syndication, current visibility across major AI platforms, and relative competitive position on the attributes agents evaluate. In practice, CPG teams can translate these factors into a readiness scorecard rating each product on data quality, feed coverage, algorithmic visibility, and attribute competitiveness to identify where they are currently strong and where remediation is required.

Products that score high on both dimensions should be the priority. These are the hero stock keeping units (SKUs) and core categories where investment in agentic readiness will deliver near-term returns.

For products with high value at stake but low ability to win, the focus should be to build the foundation: improving product data and content by standardizing attributes, enriching ingredient and usage information, closing gaps in retailer feeds, and ensuring claims are consistently machine‑readable across platforms.

Products with lower strategic value can be monitored and addressed as the capability matures.

Expand table

This prioritization should remain dynamic. As AI adoption accelerates, category behaviors change, and competitors strengthen their capabilities, organizations should regularly reassess both the value at stake and their ability to win.

The operating model shift: From channel silos to scenario orchestration

Perhaps one of the most challenging implications of agentic commerce for CPG organizations is not technological but organizational. Today, the functions that collectively determine how a product appears in the digital marketplace (search engine optimization (SEO), e-commerce, content, retail media, product data management, and technology) typically operate in separate teams with separate goals. While that structure made sense when channels could be managed independently, it is fundamentally misaligned with a world where a single consumer query can trigger an agent to evaluate product data, compare pricing across retailers, assess review sentiment, verify ingredient claims, and complete a purchase, all within seconds and across multiple platforms simultaneously.

The shift is from managing individual channels to orchestrating the consumer scenarios through which products are discovered, evaluated, and purchased. Connecting channels is not new to CPG companies and has been central to digital and omnichannel strategies for over a decade, but agentic commerce raises the level of coordination required. Gaps in data, content or platform integration can now directly affect whether a product is recommended or passed over.

Several established practices therefore need to evolve:

  • From SEO and keyword rankings to AI recommendation share, supported by structured product data, verified claims, review quality, and accurate retailer feeds.
  • From campaign-led content to continuous product data publishing.
  • From retailer-specific feed management to multi-platform data governance.
  • From channel-level key performance indicators (KPIs) to scenario-based performance measures, tracking visibility, recommendation, and conversion across the agent-mediated journey.

This is not a peripheral initiative that can be delegated to a single team. Instead, it requires a dedicated cross-functional capability, an agentic commerce hub, which sets shared priorities across content, data, technology, and analytics.

Building for an uncertain architecture

No single architecture for agentic commerce has yet emerged as dominant. Instead, three models are developing in parallel:

  1. Retailer-led: Retailers use their own agents to guide discovery and purchase within their ecosystems.
  2. General AI-led: Horizontal agents, such as ChatGPT or Gemini, orchestrate discovery before directing consumers to retailers for fulfilment.
  3. Interoperable: Agents connect to live merchant catalogues through common protocols, enabling cross-merchant orchestration.

Which model scales fastest is likely to vary by market, with regulation shaping the pace and direction of adoption. The United States and China are currently moving ahead with relatively permissive approaches to AI‑mediated commerce, enabling rapid experimentation by retailers and platforms across closed and general‑purpose agent models.

In Europe, regulators are more actively shaping how AI systems handle consumer data, transparency of recommendations, and liability for autonomous decisions. As forthcoming rules are defined and implemented, they are likely to influence which models become predominant and at what pace.

For CPG companies, the answer is not to bet on a single end-state. Instead, they should build a common agentic commerce foundation of structured data, verified claims, real-time feeds, and schema-ready content, and then adapt activation by market, retailer, and emerging platform. This provides the flexibility to compete across all three models and leaves the organization well placed to respond as the landscape evolves.

The imperative to act

AI-referred traffic to retail sites grew 393% year-over-year in Q1 2026, while AI-referred visits converted at rates 42% higher than non-AI traffic.4 The trajectory is clear, and the window for establishing algorithmic visibility is finite. Positions earned early in an emerging channel are disproportionately difficult to displace once habits and algorithms consolidate.

For CPG brands, the imperative is equally clear. The invisible shelf is growing, and competing on it requires translating brand equity into machine-readable signals. The companies that treat product data as a strategic asset, build cross-functional orchestration capabilities, and invest in readiness across multiple commerce architectures will not merely survive the agentic shift. They will define it.

Brand equity remains essential, but it must now be translated into machine-readable proof.

Discover our Future of Advice Blog Homepage

1 Deloitte Survey on Agentic Commerce, Europe, 2026.

2 Salesforce, 2025.

3 Deloitte Emerging retail and consumer trends, 2026.

4 Adobe, 2026.

Did you find this useful?

Thanks for your feedback