Agentic commerce for CPG brands: what AI shopping agents change, and what to do now

For decades, consumer packaged goods (CPG) brands competed for two things: space on the physical shelf and attention on the digital one. A third shelf is now forming, and no person ever walks past it. A shopper types "find me a fragrance-free laundry detergent that’s safe for a toddler and under $20," and an AI assistant reads hundreds of product listings, weighs reviews, and returns two or three options. Sometimes it completes the purchase too.

This shift is called agentic commerce: AI agents, meaning software that can take actions toward a goal rather than just answer questions, doing parts of the shopping journey for a person. This article explains what is actually happening, what it means for a CPG commerce team, the real benefits and limits, and a practical 90-day starting plan.

What is actually happening in 2026

The movement is real, though still early. Adobe, which analyzes more than a trillion visits to US retail sites, reported that traffic arriving from generative AI tools rose 393% in the first quarter of 2026 compared with a year earlier ¹. More telling than the volume is the quality. In March 2026, AI-referred visitors converted 42% better than other visitors and generated 37% more revenue per visit. A year earlier, the same AI traffic converted 38% worse ¹. For the 2026 holiday season, Adobe forecasts AI traffic to US retail sites will grow another 130%, with grocery among the fastest-growing online categories at a projected 10.3% ².

The infrastructure is being built at the same time. In late 2025, OpenAI launched in-chat checkout in ChatGPT on an open standard it co-developed with Stripe, the Agentic Commerce Protocol ¹². In January 2026, Google announced the Universal Commerce Protocol with Shopify, Target, Walmart and Etsy, designed to let agents handle discovery, payment and post-sale support ¹⁰. By March, Walmart’s own shopping agent, Sparky, was running inside ChatGPT, and retailers including Target, Sephora, Lowe’s and The Home Depot had joined OpenAI’s protocol ¹¹. Shopify, headquartered in Ottawa, now lets merchants place products inside ChatGPT, Perplexity and Microsoft Copilot conversations, with attribution data flowing back to the merchant ¹³.

CPG sits close to the center of this. Morgan Stanley estimates agentic shoppers could account for 10% to 20% of US e-commerce spending by 2030, up to $385 billion, and expects grocery and CPG to lead because they are frequent, repeat purchases ³. Consumers agree: in a Checkout.com survey, groceries (41%) and household supplies (31%) topped the list of purchases people would hand to an agent ⁶.

The business problem: your next customer may read, not look

A human shopper responds to packaging, placement and familiarity. An agent responds to data. It reads attributes, ingredient lists, certifications, reviews, prices and availability, then matches them against the shopper’s stated need. If a product’s "plastic-free packaging" or "nut-free facility" claim exists only on the physical box, the agent cannot see it.

This creates three problems for CPG commerce teams.

Brand loyalty becomes negotiable. In Accenture’s 2026 survey of 25,590 consumers in 16 countries, 37% of behaviorally loyal shoppers said they would let an agent switch them away from their preferred brand for a better fit ⁴. Checkout.com found 57% of consumers would let an agent switch brands for better value ⁶. When an agent compares on claims, price and fit, weak differentiation becomes visible quickly.

Many products are not readable. Adobe found that roughly a third of retail product pages cannot be properly accessed by AI systems ¹. Content built for human eyes, such as claims embedded in images or vague marketing copy, is often invisible to a machine.

The rules of discovery are shifting. NIQ and Kearney argue that agents favor relevance to a shopper’s specific need over brand size or ad spend, and that a "trust layer" of video, community forums and creator content shapes nearly half of agentic search results ⁸. That can help challenger brands, and it can hurt incumbents that relied on shelf presence.

How AI changes the work of a CPG commerce team

Agentic commerce changes the job in four practical ways.

1. Product data becomes packaging. Google Cloud’s CPG team frames this well: structured product data is now the package an agent "sees" ¹⁶. That means complete, consistent attributes across every retailer feed: sizes, ingredients, allergens, certifications, use cases and verified claims, written as text and structured fields rather than images. Accenture’s research reaches the same conclusion: agents reward verified claims, clean product data, transparent pricing and reliable availability ⁴.

2. Content shifts from keywords to need states. Search engine optimization targets the words people type. Generative engine optimization (GEO) targets the questions people ask an assistant: "gentle on sensitive skin," "high-protein snack for kids," "best value for a family of five." NIQ recommends moving from product-centric to consumer-goal-led messaging ⁸. In practice, product descriptions should answer the real questions shoppers ask, in plain language.

3. The digital shelf gets a new aisle to monitor. Commerce teams already track ratings, content compliance and out-of-stocks across retailers. Now they also need to track how often AI assistants recommend their products for priority questions, and what those assistants say. This is a natural job for internal AI agents: running a fixed set of shopper prompts weekly across assistants, logging which brands appear, flagging wrong or outdated claims, and drafting content fixes for a person to approve.

4. Retailer relationships get more strategic. Retailers are building their own agents, like Walmart’s Sparky and Amazon’s Rufus, and they will decide which outside agents can access their sites. In March 2026, a US federal judge temporarily blocked Perplexity’s shopping agent from Amazon, a decision the Ninth Circuit overturned in August while the underlying lawsuit continues ¹⁴¹⁵. For brands, the takeaway is that the rules of agent access are being set right now by platforms and courts, and they will vary by retailer. Retail media plans and joint business plans should start including agent visibility, not just sponsored placements.

The benefits, and the limits

The upside is clear in direction, if not yet in size. AI-referred shoppers arrive better informed and convert better ¹. Adobe’s consumer survey found 69% of AI shoppers said they were less likely to return items ². Brands with genuinely differentiated products can win on merit, and smaller brands can compete with larger budgets ⁸.

The limits matter just as much.

• Most of today’s numbers are about assistance, not autonomy. Adobe measures visits referred by AI tools, not purchases agents made alone. Merchants in Checkout.com’s survey said only about 3% of their transactions currently involve AI agents ⁶. Just 9% of consumers in Accenture’s study are open to fully autonomous purchases ⁴.
• Trust is uneven, and Canadians are more cautious. In Accenture’s data, 21% of Canadians would let an agent make a final purchase decision within set limits, versus 32% globally ⁵. A Léger survey for National Bank found 39% of Canadians used generative AI to help with a purchase decision in the past year, but 28% of those users said AI had recommended something they later regretted ⁷.
• The value is hard to prove. In a 2026 Consumer Goods Forum and BCG survey of 39 senior CPG and retail executives, about three quarters of CPG respondents were still piloting AI, and 53% of all respondents did not formally measure ROI on consumer AI investments ⁹. The sample is small and directional, but the gap is familiar.
• Forecasts vary widely. Estimates of the agentic commerce market range from hundreds of billions to trillions of dollars, depending on definitions ³¹⁰. Treat any single number as a scenario, not a plan.

Where human oversight stays essential

Agents can draft, compare and monitor. People should still own the decisions that carry risk.

• Product claims. Health, nutrition, allergen and environmental claims are regulated in both Canada and the US. An AI system can find gaps and propose wording, but a qualified person should approve every claim before it reaches a feed.
• Pricing and promotions. Some vendors pitch agents that negotiate promotions or adjust prices in real time with retailer agents ¹⁶. Until the guardrails and audit trails are mature, keep a person approving price moves.
• What the brand stands for. Optimizing only for what agents rank highly can flatten a brand into a list of attributes. Positioning remains a leadership decision.
• Monitoring the monitors. Internal agents that track the AI shelf should log every prompt, result and suggested change, so a manager can review what the agent saw and why it recommended a fix.

A 90-day starting plan

1. Pick 10 to 20 priority products and write down the 5 shopper questions each should win.
2. Audit what agents see. Run those questions in the major assistants and on your top retailers’ AI tools. Record which products appear and what is said.
3. Fix the data before the copy. Close attribute gaps, move claims out of images into text, and make the same facts consistent across every retailer feed.
4. Rewrite descriptions around need states, in plain language, with verified claims only.
5. Set up a weekly AI-shelf report, ideally run by an internal agent with a person reviewing and approving changes.
6. Add agent visibility to retailer conversations and retail media planning.
7. Define one metric to track, such as share of AI answers for your priority questions, so you can tell whether the work is paying off.

Sources

1. Adobe Analytics, reported by TechCrunch (April 16, 2026). AI traffic to US retailers rose 393% in Q1, and it’s boosting their revenue too.
2. Adobe 2026 holiday forecast, reported by Digital Commerce 360 (September 29, 2026). Adobe projection: 2026 online holiday sales to reach $275.1 billion.
3. Morgan Stanley research, reported by Digital Commerce 360 (December 9, 2025). Morgan Stanley: AI agentic shoppers could drive $385 billion in online sales.
4. Accenture (June 3, 2026). Consumer Pulse Research 2026: Talk to my AI agent. Survey of 25,590 consumers in 16 countries.
5. Retail Insider (June 30, 2026). Accenture study: Canadians more cautious than global consumers about AI-powered shopping.
6. Checkout.com (June 9, 2026). Consumer demand for AI shopping is forming fast but trust for agentic commerce is still catching up. Vendor survey.
7. National Bank of Canada and Léger (June 18, 2026). AI is becoming a factor in purchasing decisions. Online survey of 1,518 Canadians, June 5–7, 2026.
8. NIQ and Kearney (March 2, 2026). The New Growth Frontier: Agentic Commerce and AI in CPG.
9. The Consumer Goods Forum and Boston Consulting Group (2026). State of AI for CPG. Survey of 39 senior executives, April 2026.
10. SiliconANGLE (January 11, 2026). Google debuts Universal Commerce Protocol to streamline agentic shopping.
11. Retail Dive (March 25, 2026). Walmart brings Sparky into ChatGPT.
12. Digital Commerce 360 (February 16, 2026). OpenAI expands agentic commerce push.
13. The Fintech Times (December 10, 2025). Shopify unveils agentic commerce era with Winter ’26 Edition.
14. GeekWire (March 10, 2026). Judge blocks Perplexity’s AI bot from shopping on Amazon in early test of agentic commerce.
15. eWeek (August 2026). Perplexity keeps its AI shopping agent on Amazon: what happens next?
16. Google Cloud, Sonia Fife (January 9, 2026). The invisible shelf: how retail and CPG can prepare for agentic commerce. Vendor perspective.