The agentic commerce radar: Autumn 2026 update

The state of agentic commerce  

Agentic commerce continues to make waves (not to say a tsunami) across the retail industry. If anything, the last few months have sharpened rather than resolved the central question of 2026: Not whether AI will reshape shopping, but how. 

Mastercard’s newly published research puts a number on the ambition: It forecasts that AI-driven consumer spending could reach GBP370 billion a year globally by 2030, with more than 300 million online shoppers routinely using AI agents to shop and pay by then. That’s a genuinely large bet on the category. But it is a 2030 bet, not a 2026 one.

On the ground, very little about the fundamentals has shifted in the last couple of months. Consumers are still using their favorite AI channels overwhelmingly to discover and compare products, not to hand over the keys to their wallets. Forrester’s mid-2026 assessment is blunt about it: Most “agentic” experiences today are still conversational, humans still drive decisions and checkout in the vast majority of cases. True autonomous purchasing remains rare, it states, even as market narratives run well ahead of actual behavior.

What has changed is the intensity of the buildout underneath that behavior. OpenAI’s Instant Checkout remains paused, but a new front has opened: Infrastructure providers are racing to arm retailers with the tools to build their own agents rather than cede discovery entirely to third-party platforms. Anthropic’s newly released commerce agent blueprints, and a wave of retailer-specific rollouts from Best Buy to Kohl’s, suggest brands are no longer waiting to see how the platform wars shake out. 

In the short term, this still translates into a “simple” requirement for brands and retailers: Make product catalogs, inventory, payments and trust signals readable by machines as well as humans. But in the long term — when machine-readable data becomes the entry point for competing in the agentic marketplace — the winners will be those with the foundation to leverage agentic across AI channels and within their brand-owned touchpoints, while maintaining control of the data and the relationship it generates.

Let’s explore in more detail what’s happened in agentic commerce over the last couple of months and what this means for your business.

Consumer sentiment for all things AI continues to evolve, but not beyond discovery and comparison

What happened? 

• Deloitte’s survey of 13,500 European consumers finds that more than half already use AI for shopping, and another 27% intend to. Adoption is highly uneven across Europe: Spain, Poland, Hungary and Italy lead at 60%+ shopper penetration, while the UK, France and Germany sit at the bottom of the range (48–52%), a roughly 20-point gap.
• Unsurprisingly, delegation is heavily concentrated at the top of the funnel: Consumers are willing to hand comparison and price-finding tasks to AI (24–34%), but that willingness collapses to just 8% at checkout. 
• In the US, Visa CEO Ryan McInerney said in September that consumers are clearly using AI to shop but not yet to pay: Roughly three-quarters of consumers don’t trust agentic platforms to make payments autonomously. However, that figure flips when a trusted payment brand is in the loop: 61% say they’d trust an agent-led payment if Visa were involved, rising to 71% among consumers who use LLMs weekly.
• Mastercard’s newly published research (surveying 26,000 people, including teens and parents, across 13 European countries) found generational divergence starting early: 18% of teens already use an AI assistant weekly to compare products, versus 10% of the parents surveyed. A third of teens said they’d hand full control to an AI agent that chooses and pays for products.
• Data privacy remains the number one reason for hesitation across generations and intensifies with age, but it’s declining as a barrier faster than expected. In Germany, Deloitte found the share of AI users citing data and privacy as a top concern fell from 52% in 2025 to about 30% in 2026, while the share willing to switch retailers based on an AI recommendation rose from 22% to nearly 30% over the same period.

Why this matters

The picture stayed similar in the last few months, just with harder numbers behind it: Adoption for research, discovery and comparison is no longer a hypothesis, it’s mainstream behavior across most of Europe and the US. 

What’s still true — and worth restating precisely because so much hype suggests otherwise — is that delegation of the actual purchase decision remains the exception, not the rule.

Deloitte frames this well: Shoppers want AI to help them decide, not to decide for them. That holds true across all markets and age groups in their survey. The generational story is nuanced, too, with younger consumers trusting AI sources more readily and are more willing to share data for better recommendations. But even so, Gen Z isn’t rushing to delegate checkout just yet. 

What is shifting, at least in mature markets like Germany, is the pace at which privacy concerns are softening as consumers get repeated positive experiences, suggesting the gradual, trust-built adoption curve we flagged last time is playing out roughly as expected, just faster in some markets than others.

AI-assisted shopping is here, with OpenAI and Google taking different paths

What happened?

• OpenAI’s Instant Checkout has not re-emerged as the center of its commerce strategy. Its latest major commerce-facing move is ChatGPT Ads, expanded to 31 European markets, with OpenAI positioning advertising around the moments when people are researching, comparing and making purchase decisions. 
• However, it seems there’s hope for OpenAI’s agentic checkout. Visa announced a strategic collaboration with OpenAI to bring Visa’s payment and security infrastructure into OpenAI’s commerce experiences. 
• Google is moving further down the funnel. UCP-powered checkout is now rolling out in AI Mode in Search and the Gemini app, while Universal Cart connects products across retailers and supports checkout through Google Pay or a handoff to the retailer. Google is explicitly building toward AI-mediated transaction execution, not just discovery. 

Why this matters

The split we flagged last quarter has become more interesting than a simple “checkout is delayed” story. AI is becoming the discovery and decision layer, but the commercial models around that layer are diverging. OpenAI’s ads business points toward monetizing the moment consumers explore and evaluate products; Google’s UCP strategy points toward carrying that intent through cart and checkout.  

However, Google’s history of bringing advertising into the fold aligns with its business model. Ads aren’t part of the Gemini app yet, but Google hasn’t ruled out the possibility of doing so eventually. 

The nearer-term battle, then, may not be “will AI replace checkout?” but rather who controls the moment of commercial intent, who shapes consideration, whose recommendations influence the basket, and whether the consumer is handed back to a retailer or completes the transaction within the AI environment.

Omnichannel is reinforcing the brand-owned side of agentic commerce

What happened?

• Home Depot has expanded its Magic Apron assistant to all of its more than 2,000 US locations, connecting product discovery with local inventory, aisle navigation and project advice through its app and in-store QR codes. Associates remain part of the experience. 
• Tesco is testing an AI meal-planning assistant that moves from preferences and recipes to adding recommended ingredients to a basket. PYMNTS describes a wider retail race to influence the shopping mission before the list (and therefore the basket) even exists.
• As mentioned in the introduction, heavyweights like Best Buy and Kohl’s also launched their own assistants. 

Why this matters

Beyond the AI heavyweights, brands and retailers are seeing the value in brand-owned experiences, with some, like Home Depot, going omnichannel. It means AI assistance is no longer confined to a website search or chat, but guiding a customer through a physical store and helping construct the basket. The interesting part of it is what’s underneath: The more places AI touches the journey, the more valuable it becomes to own the data, logic and relationship connecting those touchpoints.

The goal isn’t necessarily to keep every interaction inside a proprietary interface, but to make sure that wherever discovery starts, the brand retains enough of the context, intelligence and transaction layer to remain part of the relationship. In that sense, agentic commerce in an omnichannel context may be less about replacing the retailer’s ecosystem than making it intelligent enough to travel with the customer.

The infrastructure race: AI labs and payment networks arm retailers to build their own agents

What happened?

• Anthropic released a set of “blueprints” — reference implementations, harnesses and guardrails — for retailers, travel companies, telecoms and ticketing platforms to build their own shopping and merchant agents on Claude. The release is timed roughly three months ahead of peak holiday shopping. 
• The shopping agent handles product search, comparison and cart-building inside a retailer’s own app or site, but explicitly does not complete purchase: The payment stays with the retailer’s existing checkout or a third-party payments provider. A separate merchant agent supports inventory, pricing and marketing suggestions, gated behind human approval.

Why this matters

Rather than retailers waiting on OpenAI, Google or Amazon to define the terms of agentic checkout, an AI lab is now offering the tooling to let retailers build and own their own agents. It explicitly stops short of payment execution, which it leaves to the existing rails (and to partners like Visa and Mastercard).

That division of labor mirrors Forrester’s guidance almost exactly: Build a “machine advantage” through structured, AI-readable product data and content, on infrastructure you control, rather than betting your entire discovery strategy on someone else’s answer engine. 

Deal maker or breaker? Security, trust and identity 

What happened?

• In June, Visa announced Agent Scoring, an Agentic Registry and other capabilities designed to help identify agents, assess their behavior and distinguish agentic transactions. 
• Mastercard is moving the same problem upstream in payment with Agent Pay. The company’s Encoding Trust report focuses on agent identity, verifiable intent, task-specific permissions and auditable records, effectively creating a trust layer between an agent’s decision and the payment itself.
• Visa’s July analysis of live on-chain activity shows that agentic payments are already happening, including agents booking travel, reordering inventory and buying services. However, it also highlights the need for new protocols, standards and controls as agents begin to hold budgets and transact independently. 
• OpenAI revealed that an autonomous AI agent powered by its technology went rogue during a test, accessed the open web and hacked a prominent startup, Hugging Face. The leading AI company called it “an unprecedented incident”. 
• Gartner’s report, “Agentic AI, Sovereignty, Resilience: Framework to Demonstrate Control and Expose Gaps,” expects that 60% of large enterprises running AI agents in production will face external accountability requirements for autonomous actions by 2030, up from under 10% today.

Why this matters 

The security question is moving from “is this payment fraudulent?” to “was this action actually authorized, by whom, within what boundaries, and can anyone prove it afterward?” That’s a much bigger control problem. An agent may have permission to search, recommend, reserve or purchase, but those permissions need to be explicit, scoped, revocable and auditable.

That’s why the latest activity from Visa and Mastercard matters. Both are effectively building a trust layer around agentic action: Identity, intent, permissions, authentication and accountability before money moves. The July evidence that agents are already transacting makes this less theoretical, while Gartner’s governance forecasts suggest that the organizations that fail to build these controls into the architecture will increasingly face operational and compliance problems as autonomy scales.

And there is a useful tension here: Security cannot be used as an excuse to slow agentic adoption to a crawl. As Dirk Hoerig, Chief Innovation Officer at commercetools, put it in a recent LinkedIn article: “Security failures are a normal part of rapid innovation.” His argument is not to wait for perfect security, but to move fast with visibility, clear boundaries, isolated testing, agent/API gateways and time-bound access controls. 

The foundational readiness for the agentic enterprise

Agentic commerce still isn’t unfolding as a single, linear shift toward autonomous checkout. Forrester’s mid-2026 assessment and Deloitte’s European survey both confirm that discovery, decision-making, transactions and trust continue to evolve at different speeds, exactly as we described last time. 

What's changed is the intensity and the stakes: Infrastructure providers are now actively arming retailers to compete on their own terms and payment networks are racing to build the identity layer this future requires.

Deloitte’s own framing is a useful way to close: Retailers need to make themselves easy for AI to find, hard for it to ignore and safe for the customer to trust. Being found starts with machine-readable product data; being chosen requires clear, comparable advantages that an agent can act on; being trusted is what ultimately converts AI-assisted research into a completed purchase. And that last, hardest step is still where most of the unclaimed value sits.

Foundational readiness is what turns AI from experimentation into execution. AI doesn’t fix fragmented data, disconnected systems or inconsistent experiences; it amplifies them. The brands that win will adopt agentic capabilities powered by infrastructure that supports them at scale without compromising performance, trust, or ownership of the customer relationship.

White Paper

Navigating the Agentic Commerce Era On Your Terms

The practical guide to help you approach agentic commerce strategically — not out of FOMO.

Download Now