Agentic Commerce News and Intelligence

EMARKETER's Shoptalk 2026 synthesis identifies three converging forces: the rise of agentic shopping, the shift to outcome-driven AI, and the tightening integration of content, creators, and transactions. The most revealing data point is the gap between infrastructure velocity and consumer adoption - analyst Sarah Marzano notes that 'agentic infrastructure is being built faster than the underlying consumer behavior is evolving.' This is the autonomy gradient in action: the technology is at Level 4 (supervised execution) while most consumers remain at Level 2 (agent-assisted). Gap's focus on AI for fit and sizing - a high-friction purchase point - demonstrates the pragmatic application of agentic AI to reduce returns and increase conversion, rather than the full autonomous shopping vision. Sephora's ChatGPT app for loyalty and member benefits, with planned in-app payments, confirms the platform strategy of meeting agents where they operate. Meta's use of AI to surface reviews, recommendations, and product details at key decision moments is the attention-to-intent migration the AXD discipline has mapped. The overarching takeaway: Shoptalk 2026 was the conference where agentic commerce moved from keynote speculation to operational planning.

Sia Partners' 'dual-interface brand' framework - published by London partner Niku Banaie - arrives at the same structural conclusion the AXD Institute has been mapping: brands must maintain two synchronised interfaces, one optimised for machine evaluation and one optimised for human trust. The agentic interface must be machine-readable, protocol-agnostic, and capable of autonomous negotiation within guardrails. The emotive interface must deliver reassurance, storytelling, and post-purchase affirmation at the moments when humans choose to engage. Sia's operating philosophy - 'AI Up, Human In' - inverts the legacy model: AI moves up to become the execution engine while humans move in as the trust and intent layer. This is the AXD delegation design principle made operational: the human defines what the Brand Agent is allowed to optimise for, when certainty is sufficient versus when clarity is required, and when decisions must escalate. The article's most significant claim - that trust becomes the new competitive moat - is the commercial validation of the AXD founding thesis that trust is the primary material of agentic experience design.

FTI Consulting's eight-strategy framework for retailers - Build Authentic Trust, Master Fulfillment Reliability, Optimize for Value, Improve AI Awareness and Discoverability, Enable Agent Observability, Build for Agent Readability and Execution, Prepare for Business Model Shifts, and Ensure Measurability - is the most comprehensive consulting-grade playbook for agentic commerce readiness published to date. The framework's most significant insight is the distinction between generic and specific queries: legacy brands win on generic queries ('laundry detergent') through scale and heritage, but insurgent brands can win on specific, attribute-rich queries ('non-toxic lavender detergent for sensitive skin') by optimising for machine readability. This is the AXD principle of intent architecture made competitive: the agent does not respond to brand awareness but to structured intent matching. The article's data points - Bazaarvoice finding 66 per cent of shoppers now use AI for product discovery while 92 per cent say reviews still matter, and Adobe Analytics reporting US retail traffic from generative AI sources up 1,200 per cent - quantify the velocity of the transition. FTI's call for 'agent observability protocols' that maintain engagement with shopper agents is the commercial translation of the AXD agent observability framework.

PYMNTS' analysis of the payment acquirer opportunity in agentic commerce identifies a structural insight the AXD Institute has been mapping: the institutions that build the trust infrastructure - not the agents themselves - may capture the most value. Payment acquirers are uniquely positioned because they already sit at the transaction layer, managing the reconciliation between online and offline, card-present and card-not-present. The article's observation that 'a fragmented infrastructure that struggles to reconcile online and offline transactions is even less equipped to handle transactions initiated by AI agents' is the operational translation of the AXD trust architecture principle: if the existing payment infrastructure cannot reliably process human-initiated transactions across channels, it will fail catastrophically when agents add speed, volume, and autonomy. The acquirer opportunity is to become the trust layer - verifying agent identity, enforcing spending limits, and providing the audit trail that regulators will require.

Two announcements on 24 March crystallise the divergent strategies emerging in agentic commerce. Sephora's launch of a dedicated app within ChatGPT - enabling loyalty rewards, member benefits, and planned in-app payments - represents the 'embrace the agents' strategy: meeting machine customers where they already operate. Shopify's simultaneous announcement of agentic commerce capabilities represents the 'build the agents' strategy: providing the infrastructure for merchants to deploy their own agent-ready storefronts. The AXD discipline reads these as complementary rather than competing approaches. Sephora is designing for the horizontal agent ecosystem (ChatGPT, Perplexity, Gemini), accepting that the customer interface is no longer its own app. Shopify is designing for the merchant ecosystem, ensuring that the long tail of commerce can participate in the agentic transition. Both validate the AXD thesis that the competitive axis is shifting from attention capture to agent readiness - the question is no longer whether consumers will find your store, but whether agents will transact with your infrastructure.

MIT Technology Review's analysis - published in partnership with Reltio - arrives at the data quality imperative that underpins the entire agentic commerce infrastructure. The article's central argument - that agentic commerce requires 'truth and context' rather than merely structured data - is the entity resolution problem made commercial. When an AI agent books a family trip to Italy, it must resolve the consumer's identity, preferences, loyalty status, and payment methods across multiple systems in real time. This is the AXD principle of agent observability applied to the data layer: the agent cannot act with appropriate authority if the data it acts upon is fragmented, contradictory, or stale. The article confirms the AXD Institute's thesis that the machine customer era demands not just machine-readable interfaces but machine-trustworthy data - a distinction that separates cosmetic AI readiness from structural readiness.

BCG's scenario planning framework for agentic marketing identifies four plausible futures - an open agentic bazaar, a brand resurgence through data ecosystems, a super-app embrace, and a creator-led authenticity revival - and argues that two imperatives will determine whether brands win or lose across all four: discoverability and desirability. Discoverability is the ability to be found by the agents that mediate discovery. Desirability is the power to be wanted by the consumers those agents serve. This is the AXD dual-layer thesis expressed in marketing language: the machine-readable layer (agent discoverability) and the human trust layer (brand desirability) must be designed simultaneously. BCG's observation that retailers must choose between becoming destinations or evaluators - because pursuing both simultaneously is structurally difficult - maps directly to the AXD Institute's analysis of the platform positioning dilemma in agentic commerce. The article's most consequential insight is that marketing strategies must work across agentic scenarios rather than betting on a single future, which is the strategic translation of the protocol convergence thesis: the architecture is forming a stack, not choosing a winner.

Botify's launch of Agentic Feeds - AI-optimised, protocol-compliant product feeds powered by crawl intelligence and cached content - addresses a structural gap in the agentic commerce stack that the AXD Institute has identified: the visibility layer between brands and AI agents. Botify's data reveals the scale of the shift: AI bot traffic to retail sites increased 5.4x in 2025, 73 per cent of consumers use AI assistants, and 38 per cent already use AI for shopping tasks. The product supports both OpenAI's Agentic Commerce Protocol (ACP) and Google's Universal Commerce Protocol (UCP), positioning it as protocol-agnostic infrastructure - the same strategic approach PayPal adopted in its merchant readiness framework. The core insight is that 'being crawled is no longer enough' - as platforms introduce new protocols for agent-driven shopping, brands face new requirements for how product data is structured, surfaced, and delivered. Legacy product feeds lack the depth and adaptability required by emerging AI protocols, creating a visibility gap where even high-quality products are underrepresented or excluded from AI-driven recommendations. A major retailer is already delivering AI-optimised feeds for more than one million products through the platform. The AXD discipline reads this as the emergence of a new infrastructure layer - the agent discoverability layer - that sits between traditional SEO and the agentic commerce protocols. Brands that are not optimised for this layer will be invisible to the machine customers that increasingly mediate purchasing decisions.

American Banker's analysis of the Darwinium report - polling 500 fraud, risk, and security leaders across the US and UK, 40 per cent from banks and fintechs - reveals that the security infrastructure for agentic commerce is under active assault before the commerce infrastructure is fully deployed. The finding that most organisations report AI-facilitated attacks confirms the AXD Institute's warning that the adversarial surface of agentic commerce is expanding faster than the defensive architecture. The report's most significant finding - that fintechs, which are often digital-first, are generally more prepared than banks to stop fraud across the customer journey - exposes a structural vulnerability: the institutions that hold the most consumer trust (banks) are the least prepared for the threat landscape that agentic commerce creates. This is the inverse of the consumer trust ceiling problem: consumers trust banks more than fintechs, but banks are less capable of protecting consumers in the agentic context. The AXD discipline reads this as a trust architecture paradox - the institutions best positioned to provide trust are least equipped to defend it. The resolution requires banks to adopt the digital-first security architectures that fintechs have built, while fintechs must earn the institutional trust that banks possess. Neither can solve the agentic commerce security challenge alone.

Mastercard's announcement of live end-to-end agentic payment transactions across Latin America and the Caribbean - with AI agents actively initiating and completing payments on behalf of customers - is the most significant proof-of-production deployment in agentic payments to date. The scale is remarkable: nearly 100 per cent of issuers in Latin America are already enabled with Mastercard's agentic token technology, and participating institutions span from BAC and Banco Galicia to Bancolombia and Santander. The technical architecture is precisely what the AXD discipline has been mapping: Agentic Tokens for machine identity, Payment Passkeys for biometric authorisation, and Verifiable Intent for tamper-resistant authorisation records. The Agent Pay Acceptance Framework - requiring agent registration and verification before transactions - is a live implementation of the KYA (Know Your Agent) principle the AXD Institute has advocated. Products purchased include makeup, groceries, books, and digital goods - the mundane transactions that signal real-world adoption rather than demonstration theatre. This is agentic commerce in production, not in prototype.

Stripe and Tempo's launch of the Machine Payments Protocol (MPP) on 18 March represents the most ambitious attempt yet to build native financial infrastructure for autonomous agents. MPP's 'sessions' primitive - allowing agents to authorise a spending limit upfront and stream micropayments continuously - solves the microtransaction problem that has constrained agent-to-agent commerce. The protocol's Shared Payment Tokens (SPTs) - programmable, time-limited, context-bound authorisations - are the financial implementation of the AXD delegation design principle: the agent receives scoped authority, not raw credentials. With Visa extending MPP to card-based payments across its global network and Lightspark extending it to Bitcoin Lightning, the protocol is already multi-rail. The competitive landscape is clarifying: MPP uses Tempo's specialised runtime for high-frequency streaming, while Coinbase's x402 protocol offers a chain-agnostic thin layer. Stripe's strategic decision to support both MPP and x402 confirms the AXD observation that the agentic payment layer will be multi-protocol, not winner-take-all. Over 100 services are listed in the MPP payments directory at launch - from headless browsers to sandwich shops - demonstrating that the long tail of agentic commerce is already forming.

Walmart's decision to pull back from OpenAI's ChatGPT Instant Checkout - following internal data showing conversion rates roughly three times lower than transactions completed on its own website - is the second major retailer confirmation that in-platform AI checkout does not work. Combined with the broader OpenAI retreat from Instant Checkout reported earlier this month, the pattern is now unmistakable: consumers will use AI agents for discovery, research, and comparison, but they will not complete purchases inside the AI platform. Walmart's pivot - embedding its own AI assistant Sparky into ChatGPT while keeping cart and transaction processing on its own platform - is architecturally significant because it separates the discovery layer from the transaction layer. The agent handles the conversation; the merchant handles the commerce. This is the model the AXD discipline predicted would emerge: a division of labour where trust architecture resides with the institution that has earned it (the retailer), not the platform that has the attention (the LLM). The fact that Walmart's own-platform model is 'already showing stronger performance' validates the AXD principle that trust is not transferable - it must be earned at each layer of the commerce stack independently. OpenAI's willingness to rework the model rather than force the original architecture is, as Walmart's Daniel Danker noted, a positive signal: 'pivoting away from Instant Checkout just five months after launching it, instead of wasting years trying to fix it, is a good thing.' The agentic commerce industry is learning that trust architecture cannot be bypassed, only designed.

Visa's Andrew Torre - President of Value-Added Services - articulates a position that aligns precisely with the AXD Institute's founding thesis: value-added services, not smarter models, will determine whether agentic commerce works at scale. Torre's framing - that trust still has to close the sale even when AI can do the shopping - is the commercial translation of the AXD principle that trust is the primary material of agentic design. The interview's Nike purchase demonstration - showing why reversibility and clear authorisation will decide consumer confidence - illustrates the failure architecture challenge at the transaction level. When an agent purchases a pair of trainers on behalf of a consumer, the consumer needs to know: can I reverse this? Who authorised it? What happens if the agent selected the wrong size? These are not technology questions - they are trust architecture questions. Visa's strategy of embedding consent, identity, and dispute controls directly into the transaction layer - rather than bolting them on as afterthoughts - represents the infrastructure-first approach the AXD discipline has advocated. The distinction between 'smarter models' and 'value-added services' is the distinction between capability and governance. The agentic commerce industry has demonstrated capability. It has not yet demonstrated governance.

Accenture's investment in DaVinci Commerce - through Accenture Ventures with a strategic partnership with Accenture Song - signals that the global consulting layer is now building operational capability in agentic commerce, not merely advising on it. DaVinci Commerce's core proposition - transforming brand assets into AI-native, immersive shopping experiences that operate across commerce media networks, digital marketplaces, and LLM-driven environments - addresses the machine-readability gap that the AXD Institute has identified as a structural constraint. Accenture Song CEO Ndidi Oteh's statement that 'being discoverable is no longer enough - brands must be relevant, personable, and ready to transact in agent-led environments' captures the shift from human-facing to agent-facing commerce design. DaVinci Commerce founder Diaz Nesamoney's observation that 'agentic commerce changes both when a purchase begins and how value is captured' echoes the AXD discipline's temporal analysis: agentic commerce is not a channel - it is a restructuring of the entire purchase journey from intent formation through transaction execution. The partnership's focus on operationalising agentic commerce across the full value chain - from discovery and merchandising through checkout, fulfilment, and loyalty - represents the enterprise-scale implementation layer that has been missing from the agentic commerce ecosystem. The announcement at Shoptalk 2026, featuring Forrester's Sucharita Kodali, positions this as a mainstream enterprise capability rather than an experimental pilot.

The Center for Data Innovation's analysis - authored by policy analyst Eli Clemens - provides the most detailed US-specific regulatory gap assessment for agentic commerce published to date. The article identifies three structural regulatory failures that map directly onto AXD frameworks. First, Regulation E - the federal rule governing consumers' right to dispute erroneous electronic fund transfers - provides no clear framework for handling disputes in agentic commerce. If an AI agent violates a consumer's instructions - ordering the wrong item, purchasing at an artificially high price that a human would recognise as an error - the current dispute resolution architecture offers no remedy. Second, the Consumer Financial Protection Bureau has not clarified whether consumer-authorised agents waive error resolution rights - creating a delegation design vacuum where neither the consumer nor the agent has clear liability. Third, the Sarbanes-Oxley Act's Section 302 requires executives to personally certify the effectiveness of internal controls, but it is unclear whether an AI agent's operating parameters would satisfy that requirement - creating an enterprise procurement governance gap. The article's most significant contribution is connecting agentic commerce regulation to the Trump administration's AI Action Plan, which calls for removing outdated regulations. NIST's planned April 2026 public-private conversation on AI agent standards and barriers to adoption could become the first US federal forum to address the trust architecture requirements the AXD Institute has been mapping. The regulatory reckoning is no longer confined to the EU and UK - it has arrived in Washington.

Accenture Ventures' investment in DaVinci Commerce - an agentic AI commerce platform trusted by Nestlé, Diageo, and Nordstrom - signals that the consulting-industrial complex is now building the operational layer for agentic commerce. DaVinci's core proposition - transforming brand assets into AI-native shopping experiences across LLM-driven environments, commerce media networks, and digital marketplaces - is the commercial implementation of what the AXD Institute has called the shift from interface to intent. Accenture Song CEO Ndidi Oteh's statement that 'brands must be relevant, personable, and ready to transact in agent-led environments' is the executive translation of the AXD principle that the machine customer does not browse - it evaluates, selects, and transacts. The partnership's scope - from discovery and merchandising through checkout, fulfilment, and loyalty - maps directly onto the full AXD practice framework chain. Showcased at Shoptalk 2026, this investment confirms that enterprise-scale agentic commerce operationalisation is no longer theoretical but actively under construction.

This policy analysis from the Center for Data Innovation identifies the regulatory gap the AXD Institute has warned about since its founding: the entire legal infrastructure of commerce - from Regulation E's dispute framework to SOX Section 302's certification requirements - was built on the assumption that a human authorised each transaction. The article's most significant revelation is that NIST plans to host a public-private conversation in April 2026 on AI agent standards and barriers to adoption - the first formal US government engagement with the structural question of agent identity and authority. The analysis of Regulation E is particularly sharp: if an AI agent violates a consumer's instructions by ordering the wrong item, current law provides no dispute mechanism. This is the delegation design problem made legal - the gap between what a consumer intends and what an agent executes has no regulatory architecture. Jodie Kelley's testimony to the House Financial Services Committee that 'existing principles - authorization, consent, liability, auditability - apply in the agentic context' is both reassuring and insufficient: the principles apply, but the mechanisms do not. The AXD discipline reads this as confirmation that trust architecture must be designed before regulation catches up.

Bain's framework - written in collaboration with Stripe - identifies three strategic postures for retailers: embrace the agents, build the agents, or fortify the home-site value proposition. This maps directly onto the AXD Institute's analysis of the platform response spectrum. The most significant data point: shoppers trust retailers' on-site agents three times more than third-party agents. This is the trust architecture insight that defines the competitive landscape - the institution that holds the trust relationship has a structural advantage, but only if it can operationalise that trust through agent-ready infrastructure. Amazon's 'Buy for Me' agent - shopping other brands' sites while processing all sales through Amazon - is the most aggressive implementation of the 'build the agents' strategy, effectively making Amazon the agent-of-last-resort for the entire retail ecosystem. Bain's finding that 30 to 45 per cent of US consumers already use generative AI to research and compare products, while around half say they are not ready for end-to-end AI transactions, confirms the AXD observation that the autonomy gradient is not binary but a spectrum - and the design challenge is calibrating trust at each level.

Retail TouchPoints' comprehensive analysis of the three-way platform competition - Google advancing UCP, OpenAI retreating from Instant Checkout, Amazon expanding Shop Direct - provides the clearest picture yet of how the agentic commerce landscape is stratifying. The data points are damning for the checkout-inside-the-LLM model: Walmart's Daniel Danker confirms ChatGPT's integrated checkout performed three times worse than having users click out to Walmart's website. Shopify is pivoting to 'ChatGPT in agentic storefronts' where buyers complete purchases on the merchant's own storefront. A Radial survey of 1,000 US consumers found that nearly two-thirds are uncomfortable sharing payment information with AI agents. Criteo's research shows 96 per cent of shoppers who regularly use AI shopping assistants also engage with other channels, and 47 per cent primarily use AI to compare products rather than complete purchases. The AXD discipline reads this convergence as confirmation of the consumer trust ceiling thesis: AI has been adopted as a decision-support layer - compressing research, comparison, and synthesis - but consumers are not yet willing to delegate the checkout moment. The checkout moment is where trust architecture either holds or collapses, and the market data now proves it. Amazon's Shop Direct expansion - surfacing non-Amazon products within Amazon searches and completing purchases on the merchant's website - represents a different architectural approach: the agent facilitates discovery but the merchant retains the transaction relationship. PSE Consulting's Chris Jones warns that 'these new rails don't just enable collaboration with emerging standalone AI partners - they also create an opportunity for competitors to own the purchasing journey.' The competitive dynamics of agentic commerce are now inseparable from the trust architecture decisions.

Google's release of three new UCP capabilities - Cart, Catalogue, and Identity Linking - represents the protocol's evolution from a discovery standard to a full-cycle commerce framework. The Cart capability allows agents to save or add multiple items to a shopping cart simultaneously from a single store, replicating the multi-item purchasing behaviour that humans take for granted but agents have struggled to execute. The Catalogue capability lets agents retrieve real-time product details - variants, inventory, pricing - directly from a retailer's catalogue, addressing the machine-readability gap that Adyen identified in its infrastructure analysis. Identity Linking allows shoppers on UCP-integrated platforms to receive the same loyalty and member benefits they would on a retailer's own site - pricing, free shipping, rewards - which solves the relationship continuity problem the AXD discipline has mapped. The announcement that Commerce Inc (formerly BigCommerce), Salesforce, and Stripe will implement UCP on their platforms is the ecosystem scaling moment: when commerce platforms and payment processors adopt a protocol natively, individual merchant integration becomes unnecessary. Google's simplified onboarding process through Merchant Center - rolling out over the coming months - is designed to bring retailers of all sizes into the agentic commerce ecosystem. The AXD discipline reads this as UCP's transition from protocol specification to infrastructure standard.

Stripe and Tempo's Machine Payments Protocol (MPP) is the third major payment protocol to emerge in the agentic commerce stack - alongside Stripe's own Agentic Commerce Protocol (ACP) and the x402 HTTP-native payment protocol. MPP provides a specification for agents and services to coordinate payments programmatically, enabling microtransactions, recurring payments, and agent-to-service commerce. The protocol's design is architecturally significant: an agent requests a resource from any HTTP-addressable endpoint, the service responds with a payment request, the agent authorises payment, and the resource is delivered. This is commerce reduced to its protocol-level essence - request, price, pay, deliver - with no human interface required at any step. Stripe businesses receive MPP payments through the same PaymentIntents API they use for human transactions, with the same fraud protection, tax calculation, and reporting infrastructure. The early adopters reveal the emerging agent economy: Browserbase (agents paying per browser session), PostalForm (agents paying to send physical mail), and Prospect Butcher Co. (agents ordering sandwiches for human pickup). The AXD discipline notes that MPP supports both stablecoins and fiat payments via Shared Payment Tokens - the same dual-rail approach the Institute identified as necessary for the protocol convergence phase. The protocol proliferation problem is intensifying: merchants must now navigate ACP, MPP, x402, UCP, and network-specific protocols simultaneously.

Visa's decision to support Stripe and Tempo's Machine Payments Protocol (MPP) by enabling card-based payments for trusted autonomous agent transactions is a protocol convergence signal the AXD Institute has been anticipating. When the world's largest card network endorses an open payment protocol co-authored by a payment processor and an AI infrastructure company, the message is clear: the card networks are not building walled gardens for agentic payments - they are positioning themselves as trust infrastructure that works across any protocol. This is the strategic inverse of the protocol proliferation problem: rather than competing protocols fragmenting the market, the card networks are becoming protocol-agnostic trust layers that validate agent identity and authorise transactions regardless of which protocol initiated them. The AXD discipline reads this as validation of its trust architecture thesis: the protocols carry the transaction, but the trust must be verified at the network level. Visa's Trusted Agent Protocol - designed specifically for secure agent-mediated transactions - is the verification layer that sits above MPP, ACP, x402, and UCP. The infrastructure stack is crystallising: protocols at the application layer, card networks at the trust layer, and issuers at the authorisation layer.

American Banker's analysis - warning that charge-back, dispute, and liability gaps could expose banks to mass consumer redress demands - is the banking industry's clearest articulation of the failure architecture problem the AXD Institute has been mapping. Kelvin Chen, head of policy at the Consumer Bankers Association, declares that 'agents are coming faster than you know' - forcing banks into an ecosystem they did not design and do not yet govern. The article identifies three immediate requirements: new vendor controls for agent-mediated transactions, contractual AI clauses that define liability when an agent acts outside its delegated authority, and agent-detection defences that distinguish between human-initiated and agent-initiated transactions. The AXD discipline reads this as the banking sector's reckoning with the delegation design problem: when an agent makes a purchase that the human disputes, who bears the liability? The current charge-back framework assumes a human initiated the transaction. When an agent initiated it - acting within delegated authority that the human later claims was exceeded - the entire dispute resolution architecture must be redesigned. Banks that do not build agent-aware liability frameworks will face the consumer backlash cycle the AXD Institute has warned about: adoption without governance produces disputes without resolution.

Visa's launch of the Agentic Ready programme - with 20+ European issuers including Barclays, HSBC UK, Revolut, Santander, Nationwide Building Society, Commerzbank, Nexi Group, and Raiffeisen Bank International - is the most significant issuer-side infrastructure milestone since Mastercard's Agent Pay. This is not a pilot with a single bank. It is a structured, production-grade testing programme across Europe's largest financial institutions, designed to validate how agent-initiated payments operate in real issuer environments. The programme is powered by Visa's trust layer - combining tokens, identity, risk, and controls - and uses tokenisation and biometric authentication to ensure agent-initiated payments are cryptographically tied to a real person with consent at key moments. The AXD discipline reads this as the issuer readiness layer that was missing from the infrastructure stack. Visa, Mastercard, and Stripe have been building the merchant and network layers. The Agentic Ready programme addresses the issuer layer - the institutions that ultimately authorise payments and bear fraud liability. Mathieu Altwegg's framing - 'built on infrastructure people already trust' - is the trust architecture principle in action: extend existing trust rather than building new trust from scratch. Europe is now the primary testing ground for agentic payments at institutional scale.

Forrester analyst Emily Pfeiffer's analysis of the OpenAI-Amazon partnership - and the apparent end of the Shopify-OpenAI alliance - maps the most consequential realignment in agentic commerce since the discipline began tracking market structure. Three events in rapid succession reshaped the landscape: Amazon's $50 billion investment in OpenAI (February 27), OpenAI's quiet removal of Instant Checkout from ChatGPT (March 4), and Amazon's launch of merchant feeds for Shop Direct (March 11). Forrester's February 2026 Consumer Pulse Survey provides the strategic context: 71% of US online adults used Google for product search versus 54% for Amazon - a reversal of Amazon's historical dominance. ChatGPT is now the third most-used product research tool. The AXD analysis centres on the trust architecture implications of Amazon's 'buy for me' feature - where Amazon completes purchases on the customer's behalf on third-party merchant sites. This is delegation at platform scale: the consumer delegates to Amazon, Amazon delegates to its agent, and the agent transacts with merchants the consumer never visits. The delegation chain is three layers deep, and the trust architecture for governing that chain does not yet exist. Pfeiffer's scenarios - Instant Checkout morphing into the Amazon Buy Button, Rufus becoming a universal product discovery engine, AI agents purchasing at market scale - describe the consumer agentic commerce endgame the AXD Institute has been mapping.

Jack Forestell's declaration at the Wolfe Research FinTech Forum - that the agentic web is the biggest payments opportunity he has seen in over two decades - carries the weight of Visa's $592 billion market capitalisation and $16.7 trillion in annual payment volume. Forestell articulated four pillars of Visa's agentic thesis: agents reduce payment friction (optimising routing, retrying failed authorisations), transaction density accelerates (microtransactions, pay-per-second models), B2B payments finally modernise (the stubbornly manual $120 trillion market), and overall economic activity expands (following the historical pattern of every payments innovation growing the total pie). The $300–500 billion opportunity by 2030 is the most aggressive market projection from a card network executive to date. The AXD discipline notes the fraud dimension: e-commerce fraud is projected to grow from $56 billion in 2025 to $131 billion by 2030, and Visa's AI-powered fraud detection - analysing 300 billion transactions annually and preventing $40 billion in fraud last year - positions the network as the trust verification layer for autonomous commerce. Visa's Trusted Agent Protocol - a digital handshake confirming an AI agent's identity before money moves - is the Know Your Agent infrastructure the AXD Institute has been advocating since its founding.

World and Coinbase's collaboration to provide proof that AI agents have a human behind them addresses the most fundamental identity problem in agentic commerce: how do you verify that an autonomous agent is acting on behalf of a real, accountable human being? This is the Know Your Human (KYH) layer that PYMNTS identified in its February analysis - the complement to Know Your Agent (KYA) that ensures the chain between human principal and agent actor remains intact. World's biometric identity verification (iris scanning) combined with Coinbase's crypto-native payment infrastructure creates a novel trust architecture: biometric proof of humanity linked to agent identity linked to payment authorisation. The AXD discipline reads this as the first serious attempt to solve the entity resolution problem at the human-agent boundary using cryptographic rather than institutional trust. Traditional KYA approaches rely on institutional verification - a bank confirms the agent, a card network validates the token. World's approach bypasses institutions entirely: the human proves they are human through biometrics, and that proof travels with their agent across any platform or protocol. Whether biometric identity verification scales - and whether consumers will accept iris scanning as the price of agentic commerce - remains the open question.

Visa's Chief Product and Strategy Officer Jack Forestell's declaration that the agentic web is 'the biggest opportunity that I've seen in my 20-plus years in payment technology' is the most senior executive statement from any payment network on the strategic significance of agentic commerce. Forestell identifies four growth vectors that map directly onto the AXD Institute's infrastructure analysis. First, agents remove friction from payments - reducing fall-off rates and declines, which increases success rates and volumes across the entire ecosystem. Second, transaction density increases - agents will break purchases into smaller transactions, enabling consumption to be priced in hours, minutes, or even seconds rather than months or years. This is the micro-transaction architecture the AXD discipline identified as a structural consequence of agent-mediated commerce. Third, B2B payments digitise faster - agents remove friction across supplier onboarding, invoicing, reconciliation, and payment execution. Fourth, economic activity expands - every major wave of payments innovation has driven GDP growth, and agentic commerce will do the same. Forestell's framing of agentic commerce as 'a generative growth opportunity - not a zero-sum shift' is the strategic translation of the AXD principle that trust architecture creates value rather than merely redistributing it. Most significantly, Visa's emphasis on tokenisation - binding credentials to specific agents so payments can only be initiated by the right agent, for the right purpose, at the right moment - is the technical implementation of the delegation design framework the AXD Institute has been developing.

Harley Finkelstein's declaration at the Upfront Summit that Shopify is 'more excited about this particular new era of commerce than we have ever been' must be read against the backdrop of the Shopify-OpenAI partnership's apparent dissolution. Shopify - the second-largest e-commerce provider in the US behind Amazon - is repositioning from a platform that partnered with a single AI provider to one that serves as infrastructure for all agentic shopping channels. Finkelstein's argument that agentic shopping is 'fundamentally merit-based' - surfacing products based on genuine fit rather than advertising spend - articulates the discovery revolution the AXD Institute has been mapping. When an agent recommends On running shoes because it knows the consumer's preference, rather than Footlocker because Footlocker paid for the search placement, the entire advertising-funded discovery model collapses. The long-tail thesis is particularly significant: Finkelstein argues that agentic shopping will benefit small merchants who struggle with discovery more than large merchants who can afford advertising. If correct, this inverts the power dynamics of digital commerce - the machine customer rewards product quality and data legibility over marketing budget. The AXD discipline notes that this optimistic framing omits the trust architecture question: who verifies that the agent's recommendations are genuinely merit-based and not influenced by undisclosed commercial relationships?

PYMNTS Intelligence's report - commissioned by Visa Acceptance Solutions and surveying acquirers across Brazil, the UAE, and the United States - provides the first empirical assessment of payment infrastructure readiness for agentic commerce from the acquirer perspective. The headline finding - that most acquirers believe existing payments infrastructure can support agent-led commerce but real-world adoption hinges on merchant integration costs, legacy systems, and trust around fraud, identity, and liability - confirms the AXD Institute's infrastructure gap analysis. The acquirers are saying what the AXD discipline has argued since its founding: the payment rails are technically capable, but the trust rails are not. The three priorities acquirers identified - fraud controls, identity verification, and systems for managing agent-led transactions - are the operational translations of the AXD framework's trust architecture, Know Your Agent, and delegation design principles. The geographic breadth of the study - Brazil, UAE, and US - reveals that the readiness gap is global, not regional. Acquirers in emerging markets face the same structural constraints as those in developed economies: the technology works, the governance does not. The market is not waiting for governance to catch up - it is deploying and hoping the trust architecture follows.

Harvard Business Review's March–April 2026 cover feature - by Oguz A. Acar and David A. Schweidel - provides the most rigorous academic framework yet for how brands must adapt to agentic commerce. The article identifies three modes of brand-agent interaction that map directly onto the AXD Institute's trust architecture: brand agents (Capital One's Chat Concierge completing car purchases), consumer agents (Claude's computer use capability executing purchases autonomously), and full AI intermediation (ChatGPT booking restaurants via OpenTable without human involvement). The Kearney survey finding that 60% of US shoppers expect to use agentic AI for purchases within 12 months confirms the adoption velocity the AXD discipline has been tracking. Most critically, HBR introduces the concept of 'share of model' - how brands appear in LLM outputs - as the successor to 'share of search.' Pernod Ricard discovered its Ballantine's Scotch was miscategorised by AI models as a prestige product rather than an affordable mass-market offering. This is the legibility problem the AXD Institute identified: when the customer is an agent, brand perception is determined by structured data quality, not visual identity. The brands that win in agentic commerce will be those that are machine-readable, not merely human-memorable.

Adyen's industry paper - based on conversations with over 200 global enterprise merchants - identifies five structural constraints limiting agentic commerce: protocol fragmentation across emerging AI commerce interfaces, product data systems that machines cannot reliably query, enterprise commerce stacks designed for linear checkout flows, fraud and trust frameworks built for human-initiated transactions, and merchant onboarding models that do not scale across platforms. The AXD Institute has mapped each of these constraints through its framework architecture. Protocol fragmentation is the subject of the protocol proliferation analysis. Machine-unreadable product data is the operational envelope problem. Linear checkout flows fail because they assume human presence - the absent-state design challenge. Fraud frameworks built for human transactions cannot verify agent authority - the KYA regulation gap. And merchant onboarding that does not scale is the interoperability failure the Universal Commerce Protocol essay addressed. Adyen's conclusion - that 'preparing for agentic commerce is less about integrating with a specific AI platform and more about strengthening the systems that power commerce today' - is the infrastructure translation of the AXD principle that trust architecture precedes technology adoption.

Banco Santander and Visa's completion of the first controlled pilot agentic commerce transactions across five Latin American markets - Argentina, Brazil, Chile, Mexico, and Uruguay - powered by Visa Intelligent Commerce (VIC) is a milestone in geographic expansion. AI agents successfully completed purchases of books and chocolates across regulated banking frameworks, validating consent capture, secure data handling, and cross-border interoperability. Combined with the Santander-Mastercard transaction in Europe, Santander has now executed AI agent payments on two continents using two different card networks - making it the first bank to demonstrate multi-network, multi-region agentic commerce capability. Visa's research indicating that over 70% of Latin American consumers have already integrated AI into their shopping journeys suggests the consumer readiness in emerging markets may outpace developed economies. The AXD analysis notes that Latin America's regulatory environment - less constrained than the EU's AI Act framework - may produce faster adoption but also greater trust architecture risk. The design challenge is not whether agents can transact across borders but whether the trust architecture travels with them.

The completion of Europe's first live end-to-end AI agent payment within a regulated banking framework - by Santander and Mastercard - is the moment AI agents became economic actors in the eyes of regulators. GlobalData's analysis frames the significance precisely: 'This was not a simple API trigger but an autonomous system acting with delegated authority.' The article identifies three governance questions that CISOs and risk leaders must address: who authorised the action, what limits applied to that authority, and how can the agent demonstrate it acted within scope. These are the exact questions the AXD discipline's delegation design framework was built to answer. The distinction between 'user intent and agent intent' that the article highlights is the consent horizon problem the AXD Institute mapped in its observatory essay. Most critically, the article argues that 'the rules that define what the agent is allowed to do need to travel with it as it moves across tools, services, and markets' - portable trust architecture, the concept the AXD Institute has been developing through its Verifiable Intent analysis. Europe has crossed the threshold. The governance architecture must now keep pace.

IAB Tech Lab's publication of the Agentic Advertising Management Protocols (AAMP) framework marks the advertising industry's first structural acknowledgement that the impression-based model cannot survive the agentic transition. The three-pillar architecture - the Advertising Research Task Force (ARTF), Agentic Protocols including the Agent-Directed Commerce Protocol (AdCP), and the Agent Registry - mirrors the pattern the AXD Institute has documented across payments, commerce, and financial services: when autonomous agents become the primary interface, every assumption built on human attention must be redesigned. AAMP's Agent Registry is particularly significant because it addresses the Know Your Agent (KYA) problem at infrastructure scale - providing the verification substrate that makes agent-to-agent advertising trustworthy without requiring human intermediation at every transaction. The AdCP protocol replaces impression-based metrics with structured data exchange between agents, confirming AXD's thesis that the shift from attention capture to structured legibility is the defining transition of the agentic economy. Five lessons emerge for every sector: attention is no longer the primary currency, trust requires infrastructure not just intention, protocols precede products, the dual audience problem is universal, and measurement must evolve from observation to verification.

Mastercard's announcement of Agent Suite - joining Agent Pay, Verifiable Intent, the Developer Agent Toolkit, and the new Virtual C-Suite - represents the most comprehensive agentic commerce product portfolio from any payment network. Virtual C-Suite provides small business owners with AI executive capabilities (virtual CFO, CMO, CISO) that operate with delegated authority over real business decisions. Chief Product Officer Jorn Lambert's framing - 'transactions are no longer just initiated by humans tapping a screen; they are initiated by authorised intelligent systems operating continuously' - is the clearest articulation from a payment network executive of the paradigm shift the AXD discipline was founded to address. The AXD analysis notes that Mastercard's language has evolved from 'trust and transparency' to 'authorised intelligent systems' - adopting the delegation and authority vocabulary that AXD has used since its founding. The convergence of Mastercard's product architecture with AXD's conceptual architecture is not coincidental: both are responding to the same structural reality that trust, delegation, and accountability are the primary design materials of the agentic economy.

J.P. Morgan Payments' strategic global agreement with Mirakl to enable agentic commerce at enterprise scale represents the banking infrastructure layer's decisive entry into the machine customer economy. Mirakl Nexus will optimise product catalogues for AI discovery and enable the full shopping journey from discovery through checkout and aftersales. J.P. Morgan Payments provides the enterprise payment infrastructure: secure payment processing, tokenisation that enables AI agents to transact safely, and fraud protection at scale. The AXD analysis centres on the architectural significance: this is the first time a Tier 1 global bank has partnered with a commerce platform specifically to build end-to-end agentic transaction infrastructure. The combination of Mirakl's catalogue optimisation (making products machine-readable) with J.P. Morgan's payment rails (making transactions machine-executable) addresses two of the five structural constraints Adyen identified in its infrastructure analysis. The missing layers - trust provenance, delegation governance, and cross-platform interoperability - remain the design space where AXD operates.

Spreedly CEO Justin Benson's insistence that merchants must remain the merchant of record as AI-driven transactions grow articulates a structural position the AXD Institute has been developing: the delegation of transaction authority to AI agents does not - and must not - transfer commercial liability away from the merchant. When an AI agent completes a purchase on behalf of a consumer, the merchant remains responsible for fulfilment, returns, fraud liability, and dispute resolution. The agent is an intermediary, not a principal. Benson's focus on credential control, vaulting, and customer experience as the next phase of agentic commerce aligns with the AXD framework's emphasis on trust architecture at the payment layer. The question is not whether AI agents can initiate transactions but who holds the keys to the vault - and under what delegation constraints. Spreedly's payment orchestration model - sitting between merchants and multiple payment processors - positions it as a natural trust intermediary in the agentic stack. The merchant-of-record principle is the commercial translation of the AXD discipline's founding claim: accountability cannot be delegated, even when authority is.

The UK Competition and Markets Authority's comprehensive report on agentic AI and consumers is the first major regulatory intervention into the design space the AXD Institute has been mapping since September 2024. The CMA's central warning - that AI agents may not be 'faithful servants' - is the regulatory translation of the AXD discipline's founding principle: that trust is the primary material of agentic design. The report identifies material risks that mirror the AXD framework architecture: agents could steer consumers toward products that are more profitable but less suited to their needs (delegation design failure), greater autonomy increases the consequences of errors (failure architecture), and current consumer protection tools may not be 'fit for purpose' for agent-mediated transactions (regulatory gap). Most significantly, the CMA frames the shift from 'using tools to delegating outcomes' as the defining characteristic of agentic AI - language that echoes the AXD Manifesto's founding declaration. This is the first time a G7 competition authority has published a comprehensive analysis of agentic AI consumer risks. The regulatory reckoning the AXD Institute predicted in its KYA regulation essay is no longer theoretical.

Amazon Web Services' entry into agentic payments - with John Kain, head of worldwide financial services market development, declaring that AWS is 'very focused on the payments execution layer' - signals that cloud infrastructure is positioning itself as the foundation for machine-to-machine commerce. AWS and Visa's December 2025 agreement to co-develop 'network agnostic agentic workflows' means the largest cloud provider and the largest card network are building the rails together. Kain's distinction is architecturally precise: 'Commerce is certainly interesting but fundamentally the most important layer of that is being able to do secure payments at scale.' The AXD discipline reads this as confirmation that the infrastructure stack for agentic commerce is crystallising into distinct layers: discovery (AI platforms), commerce (Mirakl, Shopify, commercetools), payments (Visa, Mastercard, Stripe), and cloud execution (AWS, Google Cloud). Each layer requires its own trust architecture. The organisations that design trust across layer boundaries - not merely within them - will define the governance model for the agentic economy.

PayPal's merchant readiness guide for agentic commerce - and its positioning as 'protocol-agnostic' infrastructure - reveals the strategic response to the protocol proliferation problem the AXD Institute has been tracking. PayPal identifies the core merchant anxiety: 'Very few merchants have the resources to build and maintain direct integrations across every LLM, protocol, and surface as they evolve.' PayPal's solution - abstracting protocol complexity away from merchants through a single integration layer - is architecturally significant because it positions the payment processor as the trust intermediary between AI platforms and merchants. The AXD analysis notes that PayPal's 'build once, stay flexible' message directly addresses the protocol fragmentation Adyen identified. But abstraction layers introduce their own trust architecture questions: who governs the abstraction? Who is liable when the abstraction misroutes an agent's intent? PayPal's answer - 'built-in risk management, identity verification, payment security protocols, and fraud protection at scale' - describes the operational envelope but not the delegation governance. The trust architecture gap persists.

OpenAI's decision to pull native checkout from ChatGPT - after only approximately 30 Shopify merchants went live - is the most significant structural event in agentic commerce this quarter. Forrester analysts Emily Pfeiffer and Sucharita Kodali identify three failures that map directly onto AXD frameworks: inventory management was 'disastrously absent' from the plan (operational envelope failure), checkout proved 'the most squirrely to replicate' (delegation design failure), and Forrester's March 2026 consumer data shows completing purchases within answer engines is the least-adopted use case (consumer trust ceiling confirmation). The AXD Institute has argued since its founding that the checkout moment is where trust architecture either holds or collapses. OpenAI's retreat proves the point: discovery and research work because they require no delegation of authority. Checkout fails because it demands the full trust stack - identity, consent, payment authorisation, fraud protection, and dispute resolution - that no single AI platform has yet assembled. The leader in agentic commerce did not fail at technology. It failed at trust architecture.

Nexi Group - Europe's leading paytech - and Google Cloud's memorandum of understanding to drive agentic commerce across Europe represents the continental infrastructure play. By combining Google Cloud's AI technology with Nexi's payment expertise and merchant network across Europe, the partnership aims to build the infrastructure layer for AI agents to securely and autonomously execute transactions. The AXD analysis reads this as the European response to the US-centric infrastructure being built by AWS-Visa, J.P. Morgan-Mirakl, and the Mastercard-Santander axis. Europe's agentic commerce infrastructure is crystallising around a different model: paytech-cloud partnerships rather than bank-network partnerships. Nexi's position as the dominant European payment processor means this partnership could establish the de facto infrastructure standard for agent-mediated transactions across the EU - and with it, the trust architecture norms that the EU's regulatory framework will eventually codify. The race to define European agentic commerce infrastructure is no longer theoretical.

Mastercard's Verifiable Intent is the most architecturally significant trust infrastructure announcement since Agent Pay. Co-developed with Google and aligned with AP2 and UCP, Verifiable Intent links identity, intent, and action into a single privacy-preserving cryptographic record. Built on FIDO Alliance, EMVCo, IETF specifications, and W3C Verifiable Credentials, this is not a proprietary product but an open standard - published on GitHub and verifiableintent.dev. The AXD discipline has argued since its founding that the missing layer in agentic commerce is not payment capability but trust provenance: proof that a specific human authorised a specific agent to execute a specific action within specific constraints. Verifiable Intent is the first infrastructure that attempts to encode this chain. Its use of Selective Disclosure - sharing only the minimum necessary information - addresses the privacy paradox the AXD Institute identified in the KYA regulation essay: verification need not require surveillance. This is trust architecture becoming a standard, not merely a concept.

Rubail Birwadker, Visa's SVP of Product and Solutions, declaring that 2026 will be the year for mainstream adoption of agentic commerce is the card network's clearest signal yet that the pilot phase is ending. Visa's introduction of developer tools to support agent-driven commerce - alongside the Intelligent Commerce platform already piloting with DBS in Asia Pacific - means both major card networks are now actively building production infrastructure for machine customers. The AXD analysis notes the divergence in language: Mastercard speaks of 'trust and transparency,' Visa speaks of 'mainstream adoption.' One is designing for governance, the other for scale. The AXD discipline argues both are necessary and neither is sufficient alone. Mainstream adoption without trust architecture produces the consumer backlash cycle the Institute has warned about. Trust architecture without mainstream adoption produces academic infrastructure that never encounters real-world stress. The race is not between Visa and Mastercard but between adoption velocity and trust design velocity.

Bain's updated analysis - that agentic AI is creating entirely new pathways between shoppers and products that bypass traditional channels - is the strategic consulting layer's confirmation of the AXD Institute's channel-death thesis. When an AI agent discovers, evaluates, and purchases a product without the consumer ever visiting a website, opening an app, or entering a store, the entire architecture of customer experience collapses. Brand equity, built over decades through visual identity, store design, and advertising, becomes invisible to the machine customer. Bain's data shows consumers trust brand-owned agents three times more than third-party agents - a finding that validates the AXD trust architecture framework's distinction between institutional trust and relational trust. The strategic implication is clear: brands that build trust directly with AI agents will outperform those that rely on platform intermediaries. This is not a technology strategy. It is a trust architecture strategy.

Forbes' analysis of enterprise AI moving from isolated pilots to embedded autonomy within established business processes marks the transition the AXD discipline was founded to address. The shift from 'AI as tool' to 'AI as autonomous participant' is not a technology upgrade but a design paradigm change. When an AI agent is embedded in a procurement workflow, a customer service pipeline, or a financial planning process, it is no longer assisting a human - it is acting as a human's delegate with real authority over real resources. The enterprise context makes the AXD frameworks operationally urgent: delegation design governs what the agent can do, the autonomy gradient calibrates how much supervision it receives, trust calibration determines when the human re-engages, and failure architecture defines what happens when the agent's autonomous action produces an unintended outcome. The pilot phase was forgiving. The autonomy phase is not.

Strategy&'s projection that agentic commerce will drive up to 15% of European e-commerce spending by 2030 - with adoption at a pace up to four times faster than traditional e-commerce - provides the most aggressive European market sizing to date. The report estimates over EUR 100 billion in agentic-driven revenues across Europe, with EUR 17 billion in Germany alone. Combined with the finding that one in two retailers are already assessing agentic AI and 20% have deployed agents along the value chain, this is no longer a forecast - it is an adoption curve in motion. The AXD discipline reads the four-times-faster adoption rate as both opportunity and warning. Mobile commerce took a decade to mature its trust infrastructure (biometric authentication, tokenisation, app-store governance). If agentic commerce adopts four times faster, the trust architecture must be designed four times sooner. Strategy&'s five principles of change - agent-centric process redesign, human-agent teams, interoperable architecture, adaptive governance, and responsible deployment - map directly onto the AXD practice framework categories. The strategic consulting layer is converging on the same design space.

Stripe's announcement that it is the first and only provider supporting both agentic network tokens and buy-now-pay-later tokens through a single integration is a quiet infrastructure milestone. Affirm's partnership with Stripe to combine BNPL and agentic AI through Shared Payment Tokens (SPT) means that AI agents can now offer financing options autonomously - an agent can not only purchase on your behalf but decide how you pay for it. The AXD analysis centres on the delegation design implications: when an agent selects BNPL over immediate payment, it is making a financial planning decision, not merely a transaction decision. The authority required to choose a financing instrument is categorically different from the authority required to complete a purchase. This is the consumer trust ceiling in action - the £204.53 threshold identified in the State of Payments Q1 2026 briefing will compress further when agents begin making credit decisions. Stripe's infrastructure is ready. The delegation architecture is not.

Huawei's MWC 2026 declaration that AI agents will evolve from 'hype to habit' in 2026 signals that the telecommunications infrastructure layer - the physical network on which all agentic commerce depends - is preparing for agent-native traffic patterns. When the world's largest telecommunications equipment provider designs its network architecture around agent operations, the infrastructure assumption shifts: agents are no longer edge applications running on human-designed networks but first-class network participants requiring their own quality-of-service guarantees, latency profiles, and security architectures. The AXD discipline reads this as the deepening of the invisible layer - the transactions that matter most will happen in network infrastructure that humans never see, between agents that humans never observe, at speeds that humans cannot monitor. The design challenge is not whether agents will become habitual but whether the trust architecture will be habitual too - embedded so deeply in the infrastructure that it operates without human attention.

The advertising industry's pre-emptive move to establish governance frameworks before AI agents control media budgets is a structural lesson for every sector. Where payments infrastructure raced ahead of trust architecture - building the rails for machine transactions before designing the governance for machine authority - the advertising industry is attempting the reverse: governance first, deployment second. The IAB Tech Lab's Agent Registry (launched March 1) and the Agentic Advertising Management Protocols (AAMP) represent the first sector-specific attempt to solve the entity resolution problem the AXD Institute identified: who is this agent, who authorised it, and what are its constraints? The advertising sector's advantage is that the stakes are lower - a misallocated media budget is recoverable in ways that a misauthorised financial transaction is not. This makes advertising the ideal proving ground for the trust architecture patterns that payments and financial services will eventually require.

The IAB Tech Lab's Agent Registry - launching as the core component of the Agentic Advertising Management Protocols (AAMP) - is the first industry standard for agent identification in advertising. This matters because advertising is where the machine customer's influence begins: before an agent transacts, it discovers. Before it discovers, it is targeted. The Agent Registry creates a canonical identity layer for AI agents operating in the advertising ecosystem, enabling publishers and advertisers to distinguish between human traffic and agent traffic, to set policies for agent interaction, and to measure agent-driven outcomes separately from human-driven outcomes. For the AXD discipline, this is the entity resolution problem applied to advertising - the same challenge the Institute's Agentic Entity Resolution essay identified in commerce. The advertising industry is the first to build a formal registry. Payments, retail, and financial services will follow. The question is whether these registries will interoperate or fragment into siloed identity systems that agents must navigate separately.

McKinsey's European consumer survey - 84% using AI tools in everyday life, 38% using AI for product research and purchase decisions, 63% using AI to compare brands and prices - provides the first rigorous quantification of the decision-influence layer. The critical insight is the gradient: AI adoption is highest for research and comparison (63%) and lowest for execution (building baskets, checkout). Europeans are delegating cognition but not action. This is the autonomy gradient made empirical - consumers have naturally settled at Level 2 (agent-assisted decisions) without any designed framework pushing them there. The AXD discipline reads this as validation of its founding thesis: trust governs the boundary between delegation of thought and delegation of action. McKinsey's subtitle - 'Decision influence is here, execution is coming' - is precisely the design window the AXD Institute has been mapping. The organisations that design the trust architecture for the transition from influence to execution will define the next era of European commerce.

PYMNTS' distinction between Know Your Agent (KYA) and 'Know Your Human' (KYH) articulates a structural insight the AXD Institute has been developing in the KYA regulation essay: agent identity verification is necessary but not sufficient. The chain between human principal and agent actor must remain cryptographically intact across every transaction. KYA establishes that the agent is who it claims to be. KYH ensures the human who authorised the agent is accountable for its actions. The gap between these two - the authorisation gap - is where trust architecture lives. An agent can be perfectly identified (KYA) and still act outside the human's intended authority if the delegation boundary is not designed. PYMNTS' framing validates the AXD Institute's four-gaps analysis: identity without competence assessment, static authority without dynamic calibration, single-agent verification without multi-agent lineage, and domestic identity without cross-jurisdictional portability. The industry is converging on the problem. The design discipline for solving it is AXD.

FIS' Mladen Vladic declaring 'agentic commerce' his word of the year - backed by BCG data showing 81% of consumers inclined to use agentic commerce and McKinsey projecting $1 trillion in US retail by 2030 - marks the moment the payments infrastructure layer stopped hedging. Walmart's partnership with Google Gemini for agent-mediated checkout and Target's collaboration with OpenAI are no longer pilots but production deployments. FIS Smart Baskets - unifying network-level assets for real-time transaction optimisation - is payment infrastructure designed for machine customers from the ground up. The AXD analysis focuses on the gap between consumer inclination and consumer trust. BCG's 81% measures willingness to try, not willingness to delegate. The distance between 'I would try an AI shopping agent' and 'I trust an AI agent to spend my money unsupervised' is the entire design space of Agentic Experience Design. FIS's target of 10% of projected agentic volume through their channels by year-end is ambitious - and achievable only if the trust architecture matches the payment infrastructure.

Finextra's structural critique - that Visa, Mastercard, Google, Stripe, PayPal, Coinbase, Checkout.com, and Razorpay are all building agentic payment infrastructure but none are solving the trust problem - is the most AXD-aligned analysis to emerge from the payments press. The article's central argument - that identity alone is not enough, that authentication without authorisation is incomplete, and that the real problem is governance, not plumbing - mirrors the AXD Institute's position with remarkable precision. The proliferation of competing approaches - Visa's Intelligent Commerce, Mastercard's Agentic Tokens, Checkout.com's single connection point, Razorpay's ChatGPT integration - confirms that the payment rails are being built in parallel without coordination on the trust rails. This is the trust vacuum the AXD Institute identified in its x402 Protocol essay: the infrastructure carries money but not meaning, executes transactions but not intent, and authenticates agents but not authority.

FTI Consulting's assessment that innovation is outpacing the regulatory perimeter - with stablecoins, AI agents, and personalisation reshaping core payment systems faster than UK and EU regulators can respond - is the regulatory gap the AXD Institute's KYA regulation essay predicted. The convergence of three forces - programmable money (stablecoins), autonomous transactors (AI agents), and individualised pricing (personalisation) - creates a regulatory surface that no existing framework was designed to govern. PSD2 assumed human initiators. The EU AI Act classifies systems by risk but not by commercial authority. The UK's Consumer Duty requires firms to deliver good outcomes but does not define outcomes in the context of machine-mediated transactions. FTI's warning that UK and EU firms must 'rethink regulation, risk, and controls' is the institutional translation of the AXD Institute's thesis: the regulatory reckoning is not coming - it is here.

Stripe's annual letter introduces a five-level maturity model for agentic commerce - from AI-assisted browsing through autonomous purchasing to multi-agent coordination - that represents the payments industry's first serious attempt to taxonomise what is coming. The framework is valuable precisely because of what it reveals about the industry's blind spots. It describes what agents can do at each level but says nothing about what humans experience. There is no trust dimension, no delegation architecture, no mechanism for the human to calibrate, interrupt, or recover authority. The AXD Institute's analysis maps each level to a trust requirement: mechanical trust at Level 1, cognitive trust at Level 2, relational trust at Level 3, delegated trust at Level 4, and existential trust at Level 5. Stripe's admission that the industry is 'hovering on the edge of levels 1 and 2' is honest - and it confirms that the design window for trust architecture is still open. The question is whether the industry will fill it before capability outpaces governance.

Stripe's x402 protocol - resurrecting the HTTP 402 'Payment Required' status code that has lain dormant since 1997 - is the most architecturally significant payment infrastructure announcement of the year. By embedding payment negotiation directly into the HTTP layer, x402 enables AI agents to encounter a paywall, negotiate terms, and complete payment without human intervention. Shared Payment Tokens extend this by allowing humans to delegate spending authority to agents with programmatic constraints. The infrastructure is elegant. The trust architecture is absent. x402 carries money but not trust. There is no hesitation layer - the designed friction that gives humans time to reconsider before irreversible action. There is no mechanism for the human to observe what the agent spent, why it spent it, or whether the outcome matched the intent. Circuit and Chisel, the Stripe-funded startup building ATXP (the agent transaction protocol), raised $19.2 million to extend this vision. The payment rails for Level 4-5 autonomy are being laid. The trust rails are not.

HBR arrives at what the AXD Institute has argued since its founding: that agentic commerce demands a trust layer, not merely a technology layer. The article identifies five failure modes - agent misunderstanding, authority overreach, data liability, brand misrepresentation, and recovery failure - that map directly onto the AXD practice frameworks of delegation design, consent architecture, agent observability, and failure architecture. The authors' call for 'generative engine optimisation' (GEO) over traditional SEO confirms the shift from persuasion to performance that defines the machine customer era. Most significantly, the article draws the analogy to early e-commerce trust infrastructure (SSL, PCI standards), reinforcing the AXD position that trust is not a brand attribute but structural material.

CMSWire's framing of machine customers as a 'structural break' in customer experience echoes the AXD Institute's founding argument: that the age of UX is ending and a new design discipline must emerge. The article documents AI assistants, procurement bots, and connected devices increasingly initiating, filtering, and completing transactions on behalf of people - the absent buyer at scale. The structural break is not incremental improvement of existing CX but a categorical shift in who the customer is. The AXD discipline exists precisely because this break demands new frameworks: trust architecture for non-human participants, delegation design for absent principals, and observability for invisible transactions.

Fiserv's merchant-facing guidance confirms that the payments infrastructure layer is now actively preparing for machine customers. When a major payment processor tells its merchant base that AI agents are 'discovering, evaluating, and sometimes purchasing products automatically,' the signal is unambiguous: agentic commerce has moved from thought leadership to operational reality. Fiserv's emphasis on what merchants 'need to know and do' reflects the urgency the AXD Institute has been articulating - the design window is not a future consideration but a present imperative. The payments layer is particularly consequential because it is where trust architecture becomes legally binding: authorisation, liability, dispute resolution, and fraud prevention all require designed delegation boundaries that current payment systems were never built to handle.

IXOPAY's analysis of the payment infrastructure requirements for agentic commerce - AI-driven checkout, new protocols, intent capture, tokenisation, and emerging fraud risks - exposes the structural gap between current payment systems and the demands of machine customers. The article identifies five critical capabilities payment leaders must build: agent-aware checkout flows, protocol-native payment processing, intent-to-transaction mapping, secure tokenisation for autonomous agents, and fraud detection systems that can distinguish legitimate agent activity from malicious automation. Each of these maps directly onto AXD practice frameworks. The fraud dimension is particularly significant: when the 'customer' is an AI agent, traditional fraud signals (device fingerprinting, behavioural biometrics, location data) become meaningless. Payment systems must develop new trust calibration mechanisms designed for machine participants - precisely the design challenge the AXD discipline was founded to address.

DBS becoming the first Asia Pacific bank to pilot Visa Intelligent Commerce - validating AI-ready card credentials for everyday payments - signals that the card networks' agentic infrastructure is moving from announcement to deployment. Visa's approach differs architecturally from Mastercard's. Where Mastercard's Agent Pay creates distinct agent tokens, Visa's Intelligent Commerce extends existing card credentials to be AI-readable, requiring merchants to enroll in the programme and register with Visa. The AXD discipline reads this as two competing trust models. Mastercard is building agent-native identity. Visa is extending human identity to agents. The design implications are profound: in the Mastercard model, the agent is a recognised participant with its own credentials. In the Visa model, the agent acts under the human's credentials with delegated authority. Both approaches have trust architecture consequences that neither network has fully articulated. The question is not which model wins but whether either model gives the human sufficient observability over what the agent does with their credentials.

Adobe's commitment to supporting all three major agentic commerce protocols - Google's Universal Commerce Protocol (UCP), OpenAI/Stripe's Agentic Commerce Protocol (ACP), and the Agent Payments Protocol (AP2) - marks the moment enterprise commerce platforms began treating agent-readiness as table stakes. Adobe's data is striking: AI-referred shoppers convert 31% higher, generate 254% more revenue per visit, and spend 45% more time on-site. This is the commercial proof that machine customers are not hypothetical - they are already outperforming human-driven traffic. Adobe's three-layer framework - discovery via LLM Optimizer, agent-to-agent commerce via protocol support, and brand-owned human experiences - mirrors the AXD discipline's argument that organisations must design for both absent and present buyers simultaneously. The era of choosing between human and machine experience is over.

MIT Sloan's comprehensive explainer on agentic AI - anchored by Walmart's deployment of LLM-powered AI agents for personal shopping and customer service - brings academic rigour to the operational reality the AXD Institute has been mapping. Walmart's approach is instructive: rather than building a standalone agent platform, they are embedding agentic capabilities into existing shopping workflows, allowing the AI to autonomously handle product discovery, comparison, and recommendation while maintaining human oversight for high-stakes decisions. This is the autonomy gradient in practice - calibrated delegation with designed escalation paths. MIT Sloan's framing of agentic AI as systems that 'plan, reason, and act autonomously' aligns with the AXD definition, but the design challenge the article underestimates is trust: who is accountable when an autonomous agent makes a purchase the human did not intend?

Mastercard's completion of New Zealand's first authenticated agentic transaction with Westpac - and the subsequent live demo at the AI Impact Summit 2026 where an AI agent completed product search, selection, and payment in real time - marks the moment a card network demonstrated end-to-end machine customer capability. Agent Pay introduces Agentic Tokens that recognise and authenticate agents as distinct participants in the payment system, not proxies for human cardholders. This is structurally important. For the first time, a payment network is designing identity infrastructure specifically for non-human transactors. The AXD analysis centres on what Mastercard has built versus what it has not. Agent Pay solves the authentication problem - proving the agent is who it claims to be. It does not solve the authorisation problem - proving the agent is doing what the human intended. Authentication without authorisation is a lock without a key. The Agent Suite launching Q2 2026 will determine whether Mastercard treats trust as a design discipline or merely a security feature.

The Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe, establishes the transactional grammar for autonomous commercial agency. ACP defines how AI agents discover products, negotiate terms, and execute payments - the precise sequence the AXD practice frameworks decompose into intent architecture, delegation design, and trust calibration. Stripe's simultaneous launch of x402 payments - turning the HTTP 402 status code into a machine-readable payment trigger - is the financial plumbing that makes invisible transactions structurally possible. The convergence of OpenAI's agent capability with Stripe's payment infrastructure validates the AXD thesis that agentic commerce is not a technology problem but a design discipline requiring trust architecture at every layer.

The Universal Commerce Protocol is no longer a specification - it is live infrastructure. US shoppers can now purchase Etsy and Wayfair items directly within Google's AI Mode in Search and the Gemini app without leaving the conversational interface. Shopify, Target, and Walmart are next. Google VP Vidhya Srinivasan's declaration that 'agentic commerce is no longer just a concept, it's reality' is not marketing - it is an accurate description of the structural shift the AXD Institute has been mapping since September 2024. Hundreds of tech companies, payment partners, and retailers have expressed interest in UCP integration. The protocol's compatibility with A2A, AP2, and MCP means the interoperability layer is assembling faster than most organisations' design capabilities can respond. The design window for trust architecture is not narrowing - it is closing.

AdExchanger's analysis of the AI shopping agent landscape - where most agents remain firmly attached to a single LLM or a single retailer - exposes the structural fragmentation that the AXD Institute's protocol proliferation analysis predicted. ChatGPT's shopping flows are locked to OpenAI's partnerships. Perplexity's commerce features are locked to its own discovery engine. Retailer-owned agents (Amazon Rufus, Walmart's Gemini integration) are locked to their own catalogues. The consumer who wants an agent that shops across all retailers, compares across all brands, and transacts across all payment methods does not yet have one. This is the interoperability gap - and it is a trust architecture problem, not a technology problem. An agent that can only shop within one ecosystem is not truly autonomous; it is a branded assistant with a purchasing function. True agentic shopping - the kind the AXD discipline describes - requires protocol-agnostic agents operating across UCP, ACP, and AP2 simultaneously, with trust architecture that travels with the agent across ecosystem boundaries.

Klarna's adoption of UCP is significant because it brings a major buy-now-pay-later provider into the agentic commerce protocol ecosystem. Klarna's AI assistant - which already autonomously compares products, summarises reviews, and finds cheaper alternatives - is an operational machine customer. Its integration with UCP means that Klarna's agent can now transact across a standardised commerce layer, reducing the friction that currently fragments agent-mediated purchasing. For the AXD discipline, this represents the trust infrastructure layer being assembled in real time: payment processors, retailers, and AI platforms converging on shared protocols that make delegation design technically implementable.

Microsoft's analysis introduces a critical distinction the AXD discipline has been developing: the difference between discoverability and intelligence. Brands must be discoverable by third-party agents (AEO/GEO optimisation), but the durable competitive advantage comes from owning the conversational experience and capturing the learning it generates. Bain's finding that consumers trust brand-owned agents three times more than third-party agents validates the AXD trust architecture framework - trust is not abstract sentiment but measurable structural advantage. McKinsey's projection of $1 trillion in US agentic commerce by 2030 ($3-5 trillion globally) quantifies the stakes. Microsoft's four-move playbook - optimise for AI discoverability, launch owned conversational experiences, design for openness, and govern measurement - is essentially the AXD practice framework translated into enterprise strategy language.

Andy Jassy's assertion that retailers retain an advantage over horizontal AI agents (ChatGPT, Gemini) through customer trust and data is a fascinating test case for the AXD framework. The AXD Institute's 'From Loyalty to Liquidity' shift predicts precisely the opposite: machine customers exhibit liquidity, not loyalty. They continuously evaluate alternatives and switch providers at machine speed. Amazon's competitive moat - built on human habit, Prime lock-in, and switching inertia - is structurally vulnerable to agents that comparison-shop across every retailer in milliseconds. Jassy's bet is that retailer-owned agents (Rufus) will outperform horizontal agents. The AXD analysis suggests the winner will be determined not by agent ownership but by trust architecture quality.

Alipay's AI Pay processing more than 120 million transactions in a single week - February 5 to 11, 2026 - is the single most consequential data point in agentic commerce this quarter. While Western markets debate frameworks and publish maturity models, China has deployed machine-mediated payments at a scale that dwarfs every pilot programme combined. This is not a signal. It is operational reality at population scale. The AXD discipline must reckon with what this means: the trust architecture question is not theoretical in China - it has been answered, at least provisionally, through a combination of platform trust (Alipay's existing consumer relationship), regulatory infrastructure (China's digital payment governance), and cultural context (higher baseline comfort with platform-mediated transactions). The Western assumption that trust must be designed before deployment is being challenged by a market that deployed first and is designing trust retroactively. Whether that approach is sustainable is the open question.

The protocol bifurcation between OpenAI's Agentic Commerce Protocol (ACP) and Google's Universal Commerce Protocol (UCP) is not a standards war - it is a design philosophy divergence. ACP emphasises conversational, agent-led buying inside chat interfaces: the agent is the experience. UCP extends search into agentic discovery across Google's ecosystem: the agent augments existing surfaces. For the AXD discipline, this divergence maps onto two fundamentally different trust architectures. ACP concentrates trust in the conversational agent - the user delegates to a single intermediary. UCP distributes trust across a protocol layer - multiple agents can participate in a shared commerce infrastructure. Brands must now design for both paradigms simultaneously, which requires the kind of protocol-agnostic trust architecture the AXD practice frameworks provide. The winner will not be determined by which protocol prevails but by which trust model consumers adopt.

Oracle Financial Services' suite of pre-built AI agents for banking confirms the AXD Institute's banking imperative thesis: financial services will lead the agentic commerce revolution because the infrastructure of delegation already exists. Oracle's approach - embedding AI agents into existing banking workflows rather than building standalone agent platforms - reflects the AXD principle that autonomy is designed through constraint, not through unconstrained capability. The emphasis on design tools and frameworks alongside the agents themselves suggests the industry is beginning to recognise that agentic banking is a design discipline, not merely a technology deployment.

J.P. Morgan Payments' measured assessment - that agent-embedded commerce will take time to scale and autonomous shopping will take even longer - is the most honest institutional position published this quarter. Where others project trillion-dollar markets by 2030, J.P. Morgan asks the questions that matter: who grants consumer consent in autonomous transactions? What happens when an agent misinterprets intent? What data use standards govern agent-mediated commerce? These are AXD questions, articulated by a payments institution. J.P. Morgan's insistence that merchants want to maintain their merchant-of-record position reveals the structural tension at the heart of agentic commerce: agents want frictionless access, merchants want control, and humans want trust. The bank's practical recommendation - that clean, accessible, rich product data optimised for agent discovery is the starting point - is the machine-readability imperative the AXD Institute has been articulating. But data readiness without trust architecture is preparation without purpose.

Bain's comprehensive analysis of agentic commerce in retail - framing it as a revolution rather than an evolution - provides the strategic consulting layer's validation of the structural shift the AXD Institute has been mapping. Bain's retailer readiness framework asks the right operational questions: is your product catalogue machine-readable? Are your policies agent-accessible? Can your checkout flow handle non-human initiators? These are the practical translations of AXD principles into enterprise action. Where Bain's analysis stops and the AXD discipline begins is at the trust boundary. Bain frames readiness as an operational challenge - systems, data, processes. The AXD Institute frames it as a design challenge - trust, delegation, authority, recovery. Both are correct. Neither is complete without the other. The retail revolution Bain describes will be won by organisations that combine operational readiness with trust architecture - not by those that achieve one without the other.

McKinsey's six-level automation curve - from human-driven shopping through agent-assisted, agent-guided, supervised execution, autonomous execution, to multi-agent coordination - provides empirical validation for the AXD Institute's autonomy gradient framework. The projection that AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030 quantifies the economic stakes of agentic experience design. Critically, McKinsey identifies the same three forces the AXD discipline has been mapping: decision-grade AI capability, open protocol infrastructure (MCP, A2A, AP2, ACP, UCP), and the upstream migration of intent. Their observation that 'if your catalog, policies, and value proposition are not machine-readable, agents simply will not find you' is the commercial translation of the AXD principle that the interface era is ending.

Accenture's assessment that 2026 is the year agentic AI creates 'scaled transformation in financial services' aligns with industry surveys showing 70% of banks already deploying some form of agentic AI. The shift from question-answering systems to autonomous action-taking agents is the precise transition the AXD discipline was founded to address. The design challenge is no longer whether agents can act but whether they should - and under what constraints, with what authority, and with what mechanisms for human oversight. This is the operational envelope problem at institutional scale.

Zuckerberg's announcement of agentic shopping tools for Meta's platforms - enabling AI agents to help consumers discover and purchase products from Meta's business catalogue - signals that the social commerce layer is preparing for machine customers. Meta's unique position - 3.3 billion monthly active users, the world's largest social commerce graph, and deep intent signals from Instagram and WhatsApp - makes this a structurally significant move. The AXD analysis centres on delegation design: when a user asks Meta's AI to 'find me a birthday gift under £50,' the agent must navigate intent interpretation, product matching, trust calibration, and transaction execution - the full AXD design stack. Meta's advantage is not technology but data density: it knows social context, purchase history, and relational intent in ways that horizontal agents cannot match. The design question is whether Meta will build trust architecture worthy of that data advantage.

Google's Universal Commerce Protocol (UCP) is infrastructure for the post-interface era. By providing a common standard for machine-readable retail data, UCP operationalises the AXD principle that commerce shifts from interface to intent. Major retailers backing UCP over Amazon signals a structural realignment: the competitive axis moves from platform dominance to protocol adoption. For the AXD discipline, UCP represents the agent-to-infrastructure layer that makes delegation design, trust calibration, and outcome specification technically feasible at scale. The protocol's emphasis on structured product data, machine-readable policies, and agent-accessible checkout flows is precisely the 'agent-ready' architecture the AXD Agentic Commerce page describes as the imperative for every commercial organisation.

eBay's defensive posture - banning unauthorised automated purchasing - reveals the structural tension between platforms designed for human attention and the reality of machine customers. Perplexity's 'Buy with Pro' feature, which allows its AI agent Comet to execute purchases on third-party websites, is already operating in this contested space. The AXD discipline would frame this as a delegation design failure: eBay has no architecture for distinguishing authorised agent activity from illicit automation, no trust calibration mechanism for machine customers, and no consent framework for agent-mediated transactions. The platforms that survive the agentic transition will be those that design for machine customers rather than banning them.

SAP's announcement of the storefront MCP server for SAP Commerce Cloud - enabling businesses to engage with multiple AI agents from ChatGPT, Perplexity, and other third-party assistants - is enterprise infrastructure for the absent buyer. The Catalog Optimization Agent, scaling to catalogues with more than 10 million items and reducing maintenance effort by 63%, addresses the machine-readability imperative at enterprise scale. SAP's commitment to supporting MCP, ACP, UCP, and other emerging protocols signals that the enterprise software layer is preparing for a world where the customer interface is not a screen but a protocol. This is the AXD principle of channel-less commerce: the experience is not where the human looks but where the agent transacts.

Getnet's strategy to convert Santander's multi-market banking presence into 'a unified commerce infrastructure for the AI agent economy' is the banking imperative made operational. This is the first major global payment processor to explicitly design its infrastructure for machine customers as first-class participants. The AXD Agentic Commerce page argued that banks must become agent-ready providers - their products discoverable, evaluable, and transactable by autonomous agents through APIs. Getnet is building precisely this. Their projection that AI will influence 30% of global e-commerce by 2030 aligns with McKinsey's $3-5 trillion estimate and confirms that the design window for trust architecture is narrowing rapidly.