Brief

Rewiring Demand Generation in the Age of AI Agents

When consumers delegate shopping to AI, brands must win the algorithm or disappear from consideration.

Di Shweta Bhardwaj, Lesley Butler, e Sinead Fox

First published on maggio 08, 2026
Tempo di lettura min.


In evidenza As consumers shift to mission-based prompts (“plan my son’s birthday”) and algorithms handle discovery, comparison, and purchase, AI agents are collapsing the marketing funnel. Agents favor distinctive, concrete, review-backed attributes over broad brand equity, filtering out vague positioning. Product innovation will likely move toward step-change features and more personalized variations, tightly aligned to use cases, because novelty must justify relearning by algorithms. Commercial models will rewire around agents: Retailers become technology partners, KPIs shift to agent performance, and cross-functional teams replace siloed activities. The most important customer in the consumer packaged goods (CPG) industry may soon be an algorithm. Shopping has emerged as one of the leading use cases for generative AI, with agents influencing CPG research, comparison, and, increasingly, transactions—the end state of agentic commerce. Bain & Company surveys have found that 30% to 45% of US consumers use AI for shopping support, and 64% said they have used or are open to using AI to complete a purchase. In addition, 44% of online buyers surveyed mostly start their journey in a large language model (LLM) or split their search between AI tools and traditional search engines. By 2030, we estimate, fully agentic commerce could reach $300 billion to $500 billion in revenues in the US, representing up to one-quarter of total e-commerce. The implications extend far beyond e-commerce to brand building, innovation, commercial strategy, and how companies operate. The pace of change leaves little room for incremental adaptation. This transformation will involve mobilizing cross-functional teams that share a view of consumer demand across marketing and sales groups—a prerequisite for building agent-ready brands. From browsing to prompting CPG firms now need to sell to both consumers and bots. Agent-supported shopping—and eventually, automated agent transactions—differs from traditional online shopping. Occasion-driven prompts, such as “reorder groceries for my family next week” or “set up a clean skincare routine for dry skin and restock as needed,” replace traditional queries. The steps of browsing, comparing, and purchasing often collapse into a single interaction. Machine-interpreted signals increasingly shape consumers’ decisions through AI-generated product comparisons and embedded retailer copilots. What’s emerging is a system of instant checkout and agent-controlled pricing, inventory, and logistics. Over the long run, this might remove the consumer’s hands-on involvement for many purchases. The impact will vary by category. Complex, research-heavy segments such as beauty and personal care are already seeing strong agent influence. Commoditized household essentials will follow quickly as consumers prioritize convenience and price. Categories shaped by taste and habit, such as packaged food, may shift more gradually, but even here, multi-retailer orchestration and mission-based planning will gain traction. Across categories, one dynamic holds: If a product is not selected by an agent, it effectively does not exist. Agentic commerce thus upends the traditional demand generation model and forces executives to confront four enterprise-level questions. Does brand building still matter? Agents value sharp distinctiveness over generic salience. They reward reviews, tangible attributes, and brands known for specific traits such as “designed to be soluble in cold water” (see Figure 1). However, emotional resonance still matters, as LLMs pick up narratives expressed in reviews and commentary. Figure 1 Agents recommend brands based on strength along specific criteria Source: Bain Generative AI US Consumer Survey, September 2025 (n=1,500) To stay relevant, brands must become both humanly compelling and machine legible. This entails amplifying signals of relevance in paid advertising; enriching content on the brand’s own website; using agents to educate other agents on topics such as the loyalty program, product availability, and shipping speed; and optimizing retailers’ product detail pages and reviews. What will it take to launch winning innovations? Occasion- and mission-driven demand—say, searches for “quick family dinners” rather than “macaroni and cheese”—changes the alignment of innovation themes to consumers’ needs. Incremental updates struggle in this environment. Agents favor products that clearly outperform alternatives for a distinctive use case. That is raising the bar toward step-change features, personalized variations, and occasion-based offerings. New sources of insight will emerge. Consumer prompts provide a direct window into intent, and these can be validated through retail partners. A company’s own website or app can also serve to inform innovation. Beauty care giant L’Oréal’s AI-powered Beauty Genius personalizes skincare routines based on facial scans and customer input, helping users discover their ideal products based on skin tone, hair type, and individual concerns. Customers’ prompts provide L’Oréal with data for future offerings and inform how the company talks about them. This is what it means to generate demand signals in an agent-led world. What is the most effective commercial approach? In the near term, a hybrid marketplace will likely emerge, with broadly adopted agents such as Claude and Gemini leading discovery and retailers retaining checkout and fulfillment. As retailers become technology partners, not just sales channels, CPG firms will need to train retailers’ own agents. They will also need to determine how to partner with retailers on data, experiments, and a view of transactions. That’s why Unilever, maker of many beauty and household products, is using content systems that “snap into” Amazon, Walmart, Alibaba, and emerging AI shopping interfaces. Unilever teams explicitly manage for a retailer’s search, algorithms, and generative answers, not just paid media. Given the hybrid nature of the new marketplace, CPG firms will need to partner with both LLMs and retailers in order to make the right decisions. For example, should a brand sell its products directly through ChatGPT or through a retailer? Considerations include the economics, such as the cut ChatGPT would take on an order and how this might grow over time, along with fulfillment implications and retailer blowback. Even if they don’t sell directly to consumers, LLMs are evolving into marketers with sponsored recommendations. LLMs also may start to monetize their data sets that feature a library of consumer prompts and their consumer targeting capabilities—another possible reason to partner. Is the operating model built for speed and integration? Most CPG firms are not structured for agentic commerce, with responsibilities scattered across the organization. That fragmentation hinders meaningful progress. Instead, companies should create a small, global team to orchestrate agentic commerce. The team bears responsibility for understanding how the ecosystem is evolving, establishing partnerships, standing up the data center for prompt analytics, monitoring progress, and cascading best practices to local market teams. Complementing that global team, cross-functional execution squads should be introduced in the most advanced markets, such as the US, India, and China. They will translate ongoing prompt analytics and broader understanding of bot behavior into priorities for the brand, marketing, sales, and technology teams, who can rapidly test, learn, and adapt to sudden market shifts. Governance should be clear on who owns AI-related functions, to enable faster trade-off decisions. And key performance indicators should move beyond clicks to track agent performance and outcomes. No-regrets moves to make now While the ecosystem evolves, several actions are critical for CPG firms. Build an agent-ready brand. If you’re not steering the consumer conversation, you’re invisible. Start by understanding the brand’s status relative to competitors across customers’ highest priorities. Then act on it—for example, by replacing vague claims with concrete, occasion-based descriptors. Research on prompts—how consumers search in specific categories, what their missions are—will inform how to tailor marketing innovation. The brand website should become a structured, machine-readable source of truth. Marketers should amplify validation that comes in the form of reviews, expert commentary, and forums. And the brand should appear in a consistent manner in places where marketers have that control. Prepare for agentic commerce. To facilitate actual transactions through agents, nitty-gritty details around pricing, inventory, and logistics matter a lot. Companies must define “when to choose us” logic as executable rules, clarifying returns, shipping, and guarantees in machine-readable ways. They may want to experiment with sponsored agent recommendations and mission bundles, while also shaping a direct-to-consumer channel. This checklist will of course vary by sector, with some items having a low priority. (Most food companies, for instance, will not build out fulfillment capabilities.) Mobilize a cross-functional approach. Generative AI shopping collapses the funnel. It requires bringing together marketing and sales groups into an executive-sponsored agile squad that integrates brand, e-commerce, revenue growth management, technology, and data teams. Agentic commerce is still taking shape, but the general direction is clear. Even small, niche players can compete through distinctive offerings, so CPG companies of any size cannot afford to ignore that more consumers are gravitating to AI tools. Those that do risk losing visibility in mission-based prompts and failing to have their innovations selected by algorithms. Companies that move quickly to highlight distinctive brand attributes stand to build an early advantage and become the choice the machines make.

In evidenza

As consumers shift to mission-based prompts (“plan my son’s birthday”) and algorithms handle discovery, comparison, and purchase, AI agents are collapsing the marketing funnel.
Agents favor distinctive, concrete, review-backed attributes over broad brand equity, filtering out vague positioning.
Product innovation will likely move toward step-change features and more personalized variations, tightly aligned to use cases, because novelty must justify relearning by algorithms.
Commercial models will rewire around agents: Retailers become technology partners, KPIs shift to agent performance, and cross-functional teams replace siloed activities.

The most important customer in the consumer packaged goods (CPG) industry may soon be an algorithm. Shopping has emerged as one of the leading use cases for generative AI, with agents influencing CPG research, comparison, and, increasingly, transactions—the end state of agentic commerce. Bain & Company surveys have found that 30% to 45% of US consumers use AI for shopping support, and 64% said they have used or are open to using AI to complete a purchase. In addition, 44% of online buyers surveyed mostly start their journey in a large language model (LLM) or split their search between AI tools and traditional search engines.

By 2030, we estimate, fully agentic commerce could reach $300 billion to $500 billion in revenues in the US, representing up to one-quarter of total e-commerce. The implications extend far beyond e-commerce to brand building, innovation, commercial strategy, and how companies operate. The pace of change leaves little room for incremental adaptation. This transformation will involve mobilizing cross-functional teams that share a view of consumer demand across marketing and sales groups—a prerequisite for building agent-ready brands.

From browsing to prompting

CPG firms now need to sell to both consumers and bots. Agent-supported shopping—and eventually, automated agent transactions—differs from traditional online shopping. Occasion-driven prompts, such as “reorder groceries for my family next week” or “set up a clean skincare routine for dry skin and restock as needed,” replace traditional queries. The steps of browsing, comparing, and purchasing often collapse into a single interaction.

Machine-interpreted signals increasingly shape consumers’ decisions through AI-generated product comparisons and embedded retailer copilots. What’s emerging is a system of instant checkout and agent-controlled pricing, inventory, and logistics. Over the long run, this might remove the consumer’s hands-on involvement for many purchases.

The impact will vary by category. Complex, research-heavy segments such as beauty and personal care are already seeing strong agent influence. Commoditized household essentials will follow quickly as consumers prioritize convenience and price. Categories shaped by taste and habit, such as packaged food, may shift more gradually, but even here, multi-retailer orchestration and mission-based planning will gain traction. Across categories, one dynamic holds: If a product is not selected by an agent, it effectively does not exist.

Agentic commerce thus upends the traditional demand generation model and forces executives to confront four enterprise-level questions.

Does brand building still matter? Agents value sharp distinctiveness over generic salience. They reward reviews, tangible attributes, and brands known for specific traits such as “designed to be soluble in cold water” (see Figure 1). However, emotional resonance still matters, as LLMs pick up narratives expressed in reviews and commentary.

visualization — Agents recommend brands based on strength along specific criteria

To stay relevant, brands must become both humanly compelling and machine legible. This entails amplifying signals of relevance in paid advertising; enriching content on the brand’s own website; using agents to educate other agents on topics such as the loyalty program, product availability, and shipping speed; and optimizing retailers’ product detail pages and reviews.

What will it take to launch winning innovations? Occasion- and mission-driven demand—say, searches for “quick family dinners” rather than “macaroni and cheese”—changes the alignment of innovation themes to consumers’ needs. Incremental updates struggle in this environment. Agents favor products that clearly outperform alternatives for a distinctive use case. That is raising the bar toward step-change features, personalized variations, and occasion-based offerings.

New sources of insight will emerge. Consumer prompts provide a direct window into intent, and these can be validated through retail partners. A company’s own website or app can also serve to inform innovation. Beauty care giant L’Oréal’s AI-powered Beauty Genius personalizes skincare routines based on facial scans and customer input, helping users discover their ideal products based on skin tone, hair type, and individual concerns. Customers’ prompts provide L’Oréal with data for future offerings and inform how the company talks about them. This is what it means to generate demand signals in an agent-led world.

What is the most effective commercial approach? In the near term, a hybrid marketplace will likely emerge, with broadly adopted agents such as Claude and Gemini leading discovery and retailers retaining checkout and fulfillment. As retailers become technology partners, not just sales channels, CPG firms will need to train retailers’ own agents. They will also need to determine how to partner with retailers on data, experiments, and a view of transactions. That’s why Unilever, maker of many beauty and household products, is using content systems that “snap into” Amazon, Walmart, Alibaba, and emerging AI shopping interfaces. Unilever teams explicitly manage for a retailer’s search, algorithms, and generative answers, not just paid media.

Given the hybrid nature of the new marketplace, CPG firms will need to partner with both LLMs and retailers in order to make the right decisions. For example, should a brand sell its products directly through ChatGPT or through a retailer? Considerations include the economics, such as the cut ChatGPT would take on an order and how this might grow over time, along with fulfillment implications and retailer blowback. Even if they don’t sell directly to consumers, LLMs are evolving into marketers with sponsored recommendations. LLMs also may start to monetize their data sets that feature a library of consumer prompts and their consumer targeting capabilities—another possible reason to partner.

Is the operating model built for speed and integration? Most CPG firms are not structured for agentic commerce, with responsibilities scattered across the organization. That fragmentation hinders meaningful progress. Instead, companies should create a small, global team to orchestrate agentic commerce. The team bears responsibility for understanding how the ecosystem is evolving, establishing partnerships, standing up the data center for prompt analytics, monitoring progress, and cascading best practices to local market teams.

Complementing that global team, cross-functional execution squads should be introduced in the most advanced markets, such as the US, India, and China. They will translate ongoing prompt analytics and broader understanding of bot behavior into priorities for the brand, marketing, sales, and technology teams, who can rapidly test, learn, and adapt to sudden market shifts. Governance should be clear on who owns AI-related functions, to enable faster trade-off decisions. And key performance indicators should move beyond clicks to track agent performance and outcomes.

No-regrets moves to make now

While the ecosystem evolves, several actions are critical for CPG firms.

Build an agent-ready brand. If you’re not steering the consumer conversation, you’re invisible. Start by understanding the brand’s status relative to competitors across customers’ highest priorities. Then act on it—for example, by replacing vague claims with concrete, occasion-based descriptors.

Research on prompts—how consumers search in specific categories, what their missions are—will inform how to tailor marketing innovation. The brand website should become a structured, machine-readable source of truth. Marketers should amplify validation that comes in the form of reviews, expert commentary, and forums. And the brand should appear in a consistent manner in places where marketers have that control.

Prepare for agentic commerce. To facilitate actual transactions through agents, nitty-gritty details around pricing, inventory, and logistics matter a lot. Companies must define “when to choose us” logic as executable rules, clarifying returns, shipping, and guarantees in machine-readable ways. They may want to experiment with sponsored agent recommendations and mission bundles, while also shaping a direct-to-consumer channel. This checklist will of course vary by sector, with some items having a low priority. (Most food companies, for instance, will not build out fulfillment capabilities.)

Mobilize a cross-functional approach. Generative AI shopping collapses the funnel. It requires bringing together marketing and sales groups into an executive-sponsored agile squad that integrates brand, e-commerce, revenue growth management, technology, and data teams.

Agentic commerce is still taking shape, but the general direction is clear. Even small, niche players can compete through distinctive offerings, so CPG companies of any size cannot afford to ignore that more consumers are gravitating to AI tools. Those that do risk losing visibility in mission-based prompts and failing to have their innovations selected by algorithms. Companies that move quickly to highlight distinctive brand attributes stand to build an early advantage and become the choice the machines make.