What is Agentic Commerce?

This KPI measures the purchase completion rate of sessions initiated through an AI agent. The agent conversion rate can, for example, be compared to the conversion rate of human users through traditional channels. Current market data (including Shopify and Adobe Digital Insights) suggests that traffic generated by agents is converting at an increasingly higher rate than traditional traffic.

The Agent Payments Protocol is an open standard built on A2A and MCP that was introduced by Google in September 2025. It is designed to regulate, regardless of the payment method, how AI agents may initiate payments on behalf of humans—for example, via credit card, real-time bank transfer, or stablecoin. At its core are two cryptographically signed—and thus tamper-proof—mandates: The “Intent Mandate” bindingly records the original purchase intent, while the “Cart Mandate” permanently locks in the items and price once the shopping cart has been assembled, before the payment is initiated. AP2 is now supported by more than 60 organizations, including Mastercard, PayPal, American Express, Adyen, and Shopify (as of September 2026). For cryptocurrencies, the A2A-x402 extension—developed in collaboration with Coinbase, the Ethereum Foundation, and MetaMask—supplements the protocol with corresponding payment channels.

An agent-ready catalog refers to a product catalog whose data is formatted in such a way that AI agents can search it, interpret it, and incorporate it into recommendations. Data quality is crucial for this and can be determined using a score. All data in the catalog must be complete, structured, and machine-readable. This requires, among other things, unique, standardized attributes (rather than free-text product descriptions), semantic tagging, and markup structures (e.g., Schema.org/Product).

This KPI measures the percentage of website or storefront traffic generated by AI assistants and shopping agents. Technically, this requires that agent traffic be reliably identified as such, for example, using tools such as the Adobe LLM Optimizer.

A2A is an open standard that Google introduced in April 2025. The goal is to enable AI agents from different vendors and frameworks to communicate securely with one another and coordinate tasks (e.g., when a shopping agent simultaneously requests quotes from multiple vendor agents). To this end, agents publish their capabilities via standardized, JSON-based “Agent Cards” and exchange tasks with a clearly defined lifecycle as well as work outputs (“Artifacts”). Since August 2026, A2A has been further developed in collaboration with MCP under the umbrella of the Agentic AI Foundation. More than 150 organizations support the standard (as of September 2026).

The Agentic AI Foundation (AAIF) is a neutral, cross-vendor foundation operating under the umbrella of the Linux Foundation. It was founded on December 9, 2025, with the goal of transparently and collaboratively advancing open standards for agentic AI. Among the founding contributions were Anthropic’s Model Context Protocol (MCP), Block’s agent framework “goose,” and OpenAI’s AGENTS.md standard. In August 2026, Google’s Agent2Agent (A2A) protocol was added, and it has since been further developed under the foundation’s auspices. The AAIF is supported by a three-tier membership structure: ranging from major technology companies such as AWS, Google, Microsoft, Anthropic, and OpenAI at the top Platinum tier to smaller organizations such as Hugging Face or Zapier at the Silver tier. Since its launch, membership has grown from 40 to over 250 (as of September 2026).

The Agentic Commerce Alliance (ACA) is a vendor-neutral, independent foundation that was launched in October 2025 with 15 founding members (including valantic and Shopware). The ACA offers retail brands practical frameworks, benchmarks, and guidance to help shape the adoption of Agentic Commerce. The goal is transparent governance and independence from individual platform providers. The foundation does not define technical standards, but it has incorporated the ACP—made open source by OpenAI and Stripe—into its roadmap. Since its founding, additional partners such as PayPal, Klaviyo, Nosto, Nexi, and Trusted Shops have joined the alliance.

The open standard initiated by OpenAI and Stripe enables AI agents to make purchases on behalf of users. AI agents such as ChatGPT can use the interface to suggest and compare products directly within the chat, select them, complete the checkout process, and pay without leaving the chat. The ACP also allows multiple agents to initiate transactions simultaneously. The technical foundation is a Shared Payment Token (SPT) —an encrypted token—combined with cryptographic identity verification, real-time fraud detection, context-based authorization, and spending limits.

More detailed information in the blog post: Agentic Commerce Protocol: Redefining E-Commerce with OpenAI and Stripe

This internal check value provides insight into the completeness, consistency, and machine-readable structure of product data required for an agent-ready catalog. Incorrect or incomplete data are among the most common reasons why an agent cannot capture a product and, as a result, cannot recommend or purchase it.

The open standard published by Anthropic in late 2024 is a universal, model-agnostic interface through which AI assistants and agents can access internal systems such as CRM, ERP, knowledge databases, and product catalogs. MCP is not a payment or checkout protocol; rather, it serves as the underlying data and tool layer through which commerce protocols such as ACP or UCP can access commerce data.

More detailed information in the blog post: Model Context Protocol: MCP as Infrastructure for AI Integration

This term describes how reliably and consistently a retailer fulfills its delivery, shipping, and return promises across all channels and reports them in a machine-readable format. This includes, for example, shipping options, cutoff times, geographic restrictions, and return deadlines. In Agentic Commerce, promise accuracy is a measure of fulfillment readiness and a key selection criterion for agents. If the information is unclear or inconsistent, there is a risk that an agent will skip the offer and exclude it from recommendations.

This KPI shows what percentage of the product range is technically accessible and available for transactions via the Agentic Commerce protocols (ACP, UCP).

The SPT is an encrypted token that handles payment authorization in agent-based payment processes without revealing card data. Among other things, it serves as the technical foundation for automated payment processing via the ACP.

The Unified Incentives Protocol (UIP), initiated by Talon.One, was launched in January 2026 as a specialized standard for the machine-readability of loyalty programs, discounts, and promotions. Unlike ACP and UCP, which govern payment and checkout processes, and MCP, which governs data access, the UIP specifies which loyalty and discount logic an AI shopping agent may take into account when making a purchase decision.

The Universal Commerce Protocol (UCP) is Google’s counterpart to the ACP. It connects merchant catalogs and checkout with AI agents in the Google ecosystem (Web Search, Gemini). Unlike the ACP, which enables AI agents to make purchases through a dialog-based system, UCP builds on Google’s existing search and shopping infrastructure and expands it with agent-based capabilities. The protocol launched in January 2026 with over 20 connected payment service providers (PSPs).

Verifiable Credentials (VC) are a data format for digital credentials standardized by the World Wide Web Consortium (W3C) that allows a statement about a person, a company, or a device to be verified in a tamper-proof manner without the verifier having to contact the issuer. The model involves three parties: the issuer, the holder, and the verifier. An issuer cryptographically signs a statement and transfers it to a holder, who stores the credential in their own digital wallet and, when necessary, presents it to a verifier, who can independently verify the signature. The second version of the standard, VC 2.0, was officially adopted as a W3C standard on May 15, 2025. In Agentic Commerce, Verifiable Credentials form the technical foundation for the Agent Payments Protocol (AP2): The intent and cart mandates used there are issued as Verifiable Credentials. This makes it possible to prove retrospectively, completely, and tamper-proof that a user has authorized a specific purchase intent.

Zero-click checkout refers to a purchase transaction that an agent automatically initiates based on pre-authorized permissions (including specifications regarding budget, preferences, and payment methods). At the virtual checkout, no manual clicks by human users are required for the transaction. This requires tokenized, pre-authorized payment methods (e.g., SPT, Visa Intelligent Commerce, Mastercard Agent Pay) with defined spending limits, as well as programmatic APIs.

The ACP from OpenAI and Stripe expands conversational AI assistance to include shopping features. In contrast, Google’s UCP builds on the existing Google Search and Shopping infrastructure and supplements it with agent-based capabilities. Both protocols aim to enable automated transactions via AI agents. They differ only in terms of their sponsors, level of integration, and maturity.

The MCP, ACP, and UCP protocols complement one another at different levels of an Agentic Commerce tech stack and address various issues in the process chain: MCP connects AI agents to a company’s internal data sources (e.g., product catalog, CRM, ERP), while ACP and UCP handle the actual purchase and payment process between the AI agent and the merchant. An agent can therefore use MCP to read product data and rely on ACP or UCP to complete the purchase.

While MCP connects a single agent vertically with tools and data sources, A2A manages horizontal communication between multiple agents. Both standards facilitate data access for agents and complement each other.

Since in Agentic Commerce an AI agent makes purchases on behalf of a person, tokenized mandate procedures replace traditional payment authentication: In ACP, the Shared Payment Token performs this function. At the same time, UCP-based payment integrations and card network-specific methods such as Visa Intelligent Commerce and Mastercard Agent Pay have been developed for payment authentication.

Legacy payment systems that rely on manual input or CAPTCHAs are “blind spots” for agent transactions and block automated purchases. Agent systems make decisions based on criteria different from those of humans. Rather than a functionally and visually sophisticated user experience (UX), agentic commerce places greater emphasis on the “developer experience” (DX), stable APIs, and programmatic access.

PSPs must prepare their service infrastructure to be compatible with the new standards and agent systems. Necessary adjustments focus, among other things, on API-first and event-driven architectures, robust metadata models, tailored fraud detection mechanisms, and integration with AI platforms. In Agentic Commerce, large language models (LLMs) act as a new intermediary layer, positioning themselves between users, merchants, and PSPs.

New standards such as the Agent Payments Protocol (AP2) govern, regardless of the payment method, how AI agents are permitted to initiate digital transactions (e.g., via credit card, real-time bank transfer, or stablecoin). In this process, an AI agent acts as an authorized representative on behalf of a cardholder. This is made possible by tokenized powers of attorney: cryptographically signed and thus tamper-proof mandates based on W3C Verifiable Credentials form a chain of mandates: An“Intent Mandate”bindingly records the user’s original purchase intent; a“Cart Mandate”permanently locks in the items and price once the shopping cart has been filled, before the payment is initiated.

This mandate chain is intended to resolve three key points of contention regarding autonomous, agent-initiated payments:

1. Did the user consent to the transaction?
2. Does the agent accurately reflect this intent?
3. Who is liable in the event of an error or fraud?

Technically and legally, the cardholder remains responsible for agent-initiated purchases.

New vulnerabilities are emerging due to automated ordering patterns and identity delegation: e-commerce and backend infrastructures (e.g., monitoring, security, and payment systems) must be able to distinguish legitimate agent purchases from malicious bot attacks. Industry-wide standards and procedures are currently being developed.

Since traditional fraud heuristics are designed for human click behavior and automated ordering patterns by agents must be evaluated differently, the first step is to introduce new metrics for collecting and measuring the fraud and error rates in agent-initiated purchases. Failed, aborted, or fraudulent agent transactions must be consistently monitored and reviewed as part of Quality Assurance (QA) in Agentic Commerce. In addition, it must be ensured that the commerce architecture is compatible with new security software and enhanced payment protocols.

Active and continuous monitoring, along with documented policies for agent identities and access rights, are essential. First and foremost, robust identity verification is required through tokenized or attested procedures: Mastercard Agent Pay combines cardholder, agent, and mandate identities into a single credential that enforces spending limits even before authorization. Visa’s Trusted Agent Protocol attaches signed attestation headers to every merchant request. Google’s AP2 works with verifiable “Intent” and “Cart” mandates that users sign with their own key. It also requires narrow authorization scopes based on the so-called “AND gate” principle, whereby an agent may only act if both the agent itself and the delegating individual are authorized for that action. In addition, there are granular audit trails at the tool-call level, as well as clear revocation and offboarding processes with automated deprovisioning.

AI is considered a tool without its own legal personality. Its actions are attributed to the user. The human remains the legal purchaser. The right of rescission under §§ 355 et seq. of the German Civil Code (BGB) also continues to apply unchanged to AI-mediated purchases. General civil liability rules also apply to AI-related errors. This means that Agentic Commerce does not provide a blanket exclusion of liability for consumers or providers.

By converting catalogs into machine-readable, semantically enriched formats, organizations prepare their product data for LLM-based discovery: Structured feeds, unique attributes, consistent categorization, and ongoing data quality checks are essential for AI to read, understand, and recommend products.

1. Complete required attributes instead of free text (dimensions, material, compatibility, variants)
2. Structured markup according to Schema.org/Product or comparable standards
3. Consistent categorization and taxonomy across all channels
4. Keep prices, availability, and attributes up to date, ideally synchronized in real time
5. Multilingual support without manual translation effort (AI-powered translation)
6. Image and text analysis for automated attribute extraction from unstructured sources (PDFs, data sheets, images)
7. Capture trust signals (reviews, certifications, proofs of origin)
8. Duplicate removal and consistency checks between ERP, PIM, and storefront
9. Ongoing monitoring of data quality instead of a one-time data cleansing project

Various tools support the preparation of product data, including those for OCR-based text recognition from images and PDF files, language models for attribute generation, and automated translation tools.

AI agents can only reliably populate shopping carts if bundle logic, tiered discounts, and availability are accessible as structured API objects. Every item in the shopping cart (product, quantity, price, applied discount) is recalculated via the Commerce API with every change (addition, removal, quantity adjustment) and returned as an updated total shopping cart. This allows the agent to verify the result against the user’s original intent before checkout.

Complex promotions (e.g., tiered discounts, bundle prices, “3-for-2”) must be defined as deterministic rules in the promotion engine. Availability checks are performed in real time for each item via the inventory system to prevent failure during the final checkout step.

Since there is no uniform industry standard for this yet, retailers are advised to adopt an API-first, composable commerce architecture that can be flexibly adapted to new agent protocols.

An agent-ready catalog based on a powerful PIM system is a prerequisite for agent referrals: To ensure that your products are mentioned and recommended in AI responses, you need complete, structured product data, verifiable trust signals (reviews, certifications), and open, machine-readable interfaces that agents can use to retrieve and compare products.

For AI shopping agents, brands are only visible if their products and offers appear in structured data sources. AI agents search and query, for example, product feeds, PIM exports, marketplace catalogs, and structured data (Schema.org/Product) on a brand’s own website. This means that brands remain visible when product attributes (title, description, price, availability, variants) are maintained in a complete, consistent, and machine-readable manner across all channels. In addition, trust-building signals—which AI agents factor into their recommendations—are becoming increasingly important, such as customer reviews, real-time availability, and product identifiers (GTIN/EAN) for unique identification. Content structuring also plays a crucial role in digital visibility: Brands that clearly structure their content and take GEO criteria (e.g., question-and-answer structure) into account increase the likelihood that agents will correctly understand their products and actively recommend them.

Traditional search traffic is changing in two ways:

1. An increasing proportion of product discovery is shifting from click-based Google searches to direct answers and recommendations from AI agents. Visibility is then defined less by click-through rates on the search results page and more by whether an agent mentions the product in its response or directly suggests it for purchase (“zero-click” effect).
2. At the same time, a new traffic channel is emerging in the form of inquiries sent directly to the store from agent platforms. For brands, this means that traditional SEO must be supplemented with criteria and guidelines for Generative Engine Optimization (GEO).

Both traffic sources (organic search traffic and agent traffic) must be measured and tracked separately to understand how purchasing decisions and conversions evolve over time.

The shopping journey is increasingly concentrated on a single channel: the chat window with AI assistants or platforms such as ChatGPT. Here, agents search for and compare solutions and products on behalf of a user, narrow down options, and make purchasing decisions. Metrics and attribution for monitoring and tracking must be aligned with this new, central touchpoint. Traditional marketing metrics such as website traffic, engagement, and conversion rates, as well as click-based multi-touch attribution (e.g., based on UTM parameters), provide no insight into agent interactions, mentions, or hit rates in AI-based discovery, and fall short in Agentic Commerce.

The first prerequisite is the technical visibility of agent access: This requires, among other things, the detection of agent traffic, benchmarking against competitors in AI responses, and checking for missing metadata and unstructured content. Tools such as the Adobe LLM Optimizer help map these factors. Building on this, content and data gaps can be closed more effectively, increasing the likelihood of being mentioned and recommended in LLM responses.

It is not only marketing metrics that are relevant, but also sales KPIs and metrics related to product data quality. Examples of new relevant KPIs include

• Agent Traffic Share
• Agent Conversion Rate
• Data Quality Score
• Log coverage in the catalog
• Fraud or error rate for agent-initiated purchases

At its core, performance measurement in agentic commerce encompasses combined metrics for digital visibility, data-driven performance marketing, and data and risk analyses.

Specialized AI agents are increasingly taking over operational steps in A/B testing (e.g., hypothesis generation, test setup, quality assurance of the experiment, and evaluation of results). Instead of individual test steps, specialized workflow agents can orchestrate entire process chains.

Detailed information in the blog post: Data-Driven Optimization: A/B Testing with Optimizely & Agentic AI

Data on specifically identified Agentic Commerce case studies is still scarce across the industry, which is understandable given the early stages of adoption and market development (ACP since 09/2025, UCP since 01/2026). AI projects in related industries provide indicative signals regarding the impact and ROI of Agentic Commerce initiatives:

• Shorter search times and higher conversion rates are the result of AI-powered product search at the online specialty wholesaler COLONS.
• At PIK AG, a leading provider of media and event technology, more efficient sales processes, higher data quality, and processing times reduced to just a few minutes can be attributed to AI implementation by valantic.

Even though AI “talks” about brands and compares them to competitors, personalization remains a key requirement for digital visibility and relevance. AI assistants, with which customers now interact almost daily, are changing usage habits and expectations regarding the shopping and customer experience: Immediate, personalized solutions in response to situation-specific questions are the new standard for building rapport, trust, and relevance. Product data, customer feedback, and brand messages must be prepared in such a way that AI models prioritize them and incorporate them into their recommendations. True personalization works on a personal level by taking context into account and creating one-to-one relevance.

In loyalty programs, the conceptual focus is shifting from purely points-based systems to the question of how well a system addresses customers’ interests after the purchase decision has been made. The previous mechanism—earning points and redeeming them after a purchase—is rendered ineffective when an agent makes the purchase decision. Loyalty signals and rewards must therefore be integrated into the agent’s decision-making process. Technically, this is made possible by standards such as the Unified Incentives Protocol (UIP). At the same time, retailers must establish a centralized control system that regulates which discounts AI agents are allowed to include and display in order to avoid margin losses.

Shipping options and costs, delivery windows and cutoff times, geographic delivery restrictions, and return policies must be consistent across all channels and available in a machine-readable format. Inconsistent or unclear information can cause an AI agent to skip an offer and exclude it from the selection or recommendation.

With Agentic AI, post-purchase, service, and support processes in after-sales can be made more proactive, personalized, and case-specific: Instead of rigid rule sets, agents decide on and evaluate individual cases based on context or specific events when determining returns and solutions; for example, they automatically factor in delivery delays when processing tickets, and tailor and manage information regarding return options (exchanges, store credit, refunds) based on customer profiles and company strategy, provide real-time responses, suggest alternatives and recommendations, or escalate complex, difficult cases to human service representatives—all presented in clear case reports and ticket summaries, complete with specific suggestions (Next Best Action).

Even the best language models cannot compensate for a flawed or fragmented data model. Data quality is the foundation upon which every downstream investment is built. The IT roadmap for any agent-based commerce strategy should be aligned with four building blocks and prioritized in this order:

1. Agent-ready product data and catalogs with PIM as the foundation
2. Protocol integration for data access (MCP) as well as checkout and payment services (ACP/UCP)
3. Governance and employee training to build the necessary skills
4. Performance measurement, iterative optimization, and scaling

Essentially, four points in the agreement should be clarified in advance and clearly defined:

1. Ownership and usage rights to data: Shared product and customer data must remain the property of the retailer and may not be used to train competitors’ models
2. Liability rules with exceptions for data protection, legal, and IP violations: Flat-rate liability caps often fall short when it comes to AI system errors.
3. Audit rights regarding the platform’s pricing and recommendation logic: Personalized pricing (such as in the Universal Commerce Protocol) requires oversight and a notification requirement
4. The agent’s role under data protection law: Data processor status, joint or separate responsibilities, and GDPR compliance must be clarified in advance through a data flow audit, including retention periods after the contract ends

Without updated contracts, merchants risk losing control over price display and brand presence on the platform.

Data sovereignty and independence from global platform giants are a hot-button issue for European companies. According to a valantic study, 72 percent of large German companies see a structural dependence on U.S. hyperscalers; 73 percent view the U.S. CLOUD Act as a risk, and 90 percent are considering or planning a migration to sovereign cloud models. For Agentic Commerce integrations, this means that EU data boundary requirements must be incorporated into the IT roadmap for Agentic Commerce from the very beginning.

The key term is “Agent Literacy”: the ability to outline clear tasks for AI agents (Task Design) and give them result-oriented instructions (“Prompt Discipline”), systematically review the results and evaluate their quality (Verification), and orchestrate them across multi-stage workflows—these skills will be increasingly in demand. As a result, the responsibilities and focus of existing roles are shifting from operational tasks to the strategic orchestration of AI agents, AI governance, and coaching. At the same time, new, cross-functional roles will emerge within organizations, such as Category Data Partner, Cross-functional Activation Lead, or Trust and Policy Steward.

Managed services, marketplace connectors, and standardized product data feeds eliminate the need to develop custom protocols or integrate them technically. Here’s a practical example: Shopify’s “Agentic Plan” enables integration with established AI platforms such as ChatGPT, Microsoft Copilot, and Google AI Mode/Gemini—even for merchants who don’t have their own Shopify store. Product data is managed in a consolidated manner within Shopify Admin, automatically categorized and enriched for all connected AI assistants. Marketplace integration platforms like ChannelEngine work in a similar way: A central interface connects merchants to more than 1,300 channels as well as initial UCP and ChatGPT Shopping integrations; product attributes are automatically maintained with AI support. The priority here is data quality, not technical depth. It is particularly advisable for small and medium-sized businesses to first test assistive AI within their own channel before investing in third-party agent ecosystems.

In the B2B sector, Agentic Commerce is primarily focused on reducing friction in complex sales and procurement processes. The greatest potential lies in sales, service, and aftermarket operations: Agents can handle repetitive tasks such as order tracking, price inquiries, and reorder reminders, while human advisors focus on strategic tasks. To do this, AI agents must handle multi-step approval workflows and exceptions, RFQ processes with supplier evaluations, negotiated prices and contract terms, as well as compliance checks—all integrated into existing ERP and procurement systems.
Specific use cases include, for example, automated quote generation, CPQ (Configure, Price, Quote) and guided selling, the management of contracts and inventory, the assignment of free-text or photo descriptions to corresponding spare part SKUs, and autonomous reordering based on individual consumption patterns. In manufacturing, maintenance processes and malfunctions can be predicted using sensor data (IIoT, Predictive Maintenance). An advanced use case is so-called agent-to-agent procurement, in which purchasing and sales agents from different companies negotiate directly with one another, agree on prices, and conclude transactions within defined governance guidelines. Forrester estimates that by the end of 2026, about one-third of B2B payment workflows will use AI agents, and 20 percent of sales employees will be involved in agent-assisted quote negotiations.

Where agent systems are making purchases, the direct-to-customer (D2C) relationship seems to be taking a back seat. In the short term, the impact of agentic AI on D2C brands is more positive than threatening: pure in-chat checkout has not yet caught on. OpenAI deactivated direct purchases within chat after about half a year because hardly any of the connected retailers were using it at all. Instead, when a user expresses an intent to purchase, assistants like ChatGPT typically redirect them to the brand’s website, which can generate measurably better conversion rates than traditional traffic (according to Adobe, about 42 percent; according to Salesforce, up to eight times better than via social media). It generally also means higher revenue, because the users arriving on the website are primarily those ready to buy.

The direct customer relationship thus remains with the brand, both legally and technically. The far greater influence of agent-ready AI lies in the stage prior to this: whether a D2C brand appears in an agent’s response is determined by the models’ citation logic. Younger target audiences in particular—Gen Z and Millennials—are increasingly replacing traditional search with AI interfaces.

Brands that invest early in agent-ready data and their own agent systems can thereby effectively safeguard their direct customer relationships.

AI-powered trading on marketplaces is possible in two ways:

• Option 1: Marketplace operators can integrate their own shopping agents into their platform. In this case, the marketplace itself becomes the agent interface through which both the process steps and conversions take place. A prominent example: Amazon’s agent Rufus already has over 250 million users, who complete purchases 60 percent more often than those without the assistant.
• Option 2: External agents such as ChatGPT or Google Gemini access marketplace catalogs via protocol interfaces (UCP, ACP), allowing them to find and purchase products directly from there. Anyone wishing to trade via marketplaces in this way must make their product data structured and discoverable for external AI agents.

In principle, marketplaces with a large product assortment could have an advantage, as agents prefer to access comprehensive, complete catalogs—provided they are structured and machine-readable. Marketplaces without their own agent—which serves as a conversational platform—risk becoming nothing more than backend infrastructure for third-party AI assistants.

• valantic Blog, Milko Filipov, November 27, 2025, https://www.valantic.com/en/blog/agentic-commerce-protocol/
• valantic Blog, Robert Rekece, March 10, 2026, https://www.valantic.com/de/blog/agentic-commerce-webshop-strategie/
• Linux Foundation, “Announcing the Formation of the Agentic AI Foundation (AAIF),” December 9, 2025, https://www.linuxfoundation.org/press/linux-foundation-announces-the-formation-of-the-agentic-ai-foundation
• Axios, “A2A Protocol and the Agentic AI Foundation,” August 17, 2026, https://www.axios.com/2026/08/17/a2a-agentic-ai-foundation-open-ai-standards
• Google Developers Blog, “Announcing the Agent2Agent Protocol (A2A),” April 9, 2025, https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/
• Google Cloud Blog, “Announcing the Agent Payments Protocol (AP2),” September 16, 2025, https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol
• W3C, “W3C Publishes Verifiable Credentials 2.0 as a W3C Standard,” May 15, 2025, https://www.w3.org/press-releases/2025/verifiable-credentials-2-0/
• valantic Press Release, October 13, 2025, https://www.valantic.com/de/presse/start-der-agentic-commerce-alliance-aca/
• valantic Blog, Robin Spechtenhauser, June 24, 2026, https://www.valantic.com/en/blog/model-context-protocol-mcp-for-ai-integration/
• McKinsey, “The automation curve in agentic commerce,” January 28, 2026, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-automation-curve-in-agentic-commerce?
• Agentic Commerce Alliance (official website), as of September 2026, https://www.agentic-commerce.org/

• valantic Blog, Felix Reuß, March 16, 2026, https://www.valantic.com/de/blog/agentic-commerce-payment-infrastruktur/
• valantic Blog, Milko Filipov, December 18, 2025, https://www.valantic.com/en/blog/agentic-commerce-payment-service-providers/
• valantic Blog, Felix Reuß, June 4, 2026, https://www.valantic.com/en/blog/digital-payment-in-agentic-commerce/

• EY-Parthenon, “When AI Does the Shopping: Why Agentic Commerce Is Forcing Brands to Rethink Their Strategies,” February 20, 2026, https://www.ey.com/de_de/insights/ai/wenn-die-ki-einkauft-warum-agentic-commerce-marken-jetzt-zum-umdenken-zwingt
• valantic Blog, Elias Henrich, May 22, 2024, https://www.valantic.com/en/blog/perfect-ai-driven-product-data/
• valantic Blog, Myrjam Dobesch, November 20, 2025, https://www.valantic.com/de/blog/complete-the-look-cross-selling-ki-ai/
• valantic Case Study COLONS, updated February 24, 2026, https://www.valantic.com/en/case-studies/colons-product-search-with-ai/
• valantic Case Study PIK AG, https://www.valantic.com/en/case-studies/pik-ag-ai-productdata/
• valantic Blog, Tobias Mathar, August 6, 2024, https://www.valantic.com/en/blog/cpg-report-how-ai-influences-purchase-decisions/

• Digital Applied, “Adobe Digital Insights, Q1 2026 Report,” April 2026, https://www.digitalapplied.com/blog/ai-traffic-converts-42-percent-better-2026-channel-strategy
• NielsenIQ, “The New Gatekeeper in Shopping: AI Is Playing an Increasingly Important Role in Decision-Making,” 2026, https://nielseniq.com/global/de/news-center/2026/der-neue-gatekeeper-beim-einkauf-die-ki-entscheidet-zunehmend-mit/
• valantic Blog, Janette Baumann, March 26, 2026, https://www.valantic.com/en/blog/a-b-testing-with-optimizely/
• valantic Blog, Vera Hofer, October 16, 2025, https://www.valantic.com/en/blog/adobe-llm-optimizer-geo/
• valantic Case Study COLONS, updated Feb. 24, 2026, https://www.valantic.com/en/case-studies/colons-product-search-with-ai/
• valantic Case Study PIK AG, https://www.valantic.com/en/case-studies/pik-ag-ai-productdata/

• Chain Store Age, “When AI Agents Shop For Us, How Will Loyalty Programs Keep Up?”, February 4, 2026, https://chainstoreage.com/when-ai-agents-shop-us-how-will-loyalty-programs-keep
• Digiday, “WTF is agentic commerce? How brands can make loyalty and promotions agent-ready,” June 16, 2026, https://digiday.com/sponsored/wtf-is-agentic-commerce-loyalty-promotions/
• nshift, “The Rise of Agentic Commerce: How AI Agents Will Shop For Your Customers in 2026,” February 25, 2026, https://nshift.com/blog/agentic-commerce-ai-shopping-agents-2026
• Narvar, “From Reactive to Agentic: Narvar’s Post-Purchase Predictions for 2026,” January 7, 2026, https://corp.narvar.com/blog/from-reactive-to-agentic-narvars-post-purchase-predictions-for-2026
• valantic Blog, Julia Saswito, July 16, 2026, https://www.valantic.com/en/blog/what-hyperpersonalization-means-in-retail/
• valantic Blog, Matthias Bös, December 16, 2025, https://www.valantic.com/en/blog/scaling-customer-loyalty-agentic-ai-self-service/
• valantic Blog, Ali Oezdemir, June 26, 2025, https://www.valantic.com/en/blog/salesforce-agentforce-transforming-businesses-from-service-to-commerce/
• valantic Blog, Sarah Groß (Interview with Volker John, Adobe), October 23, 2025, https://www.valantic.com/en/blog/agentic-commerce-interview/

• valantic Blog, Julia Saswito, June 8, 2026, https://www.valantic.com/en/blog/agentic-commerce-why-the-next-18-months-will-decide-the-next-ten-years/
• valantic Blog, Susanne Gmachl, June 23, 2026, https://www.valantic.com/en/blog/ai-is-changing-the-customer-journey/
• valantic Press Release, June 2, 2026, https://www.valantic.com/en/press/valantic-study-sovereign-hypercloud/
• valantic Blog, Sascha Göpfert, July 15, 2026, https://www.valantic.com/de/blog/autonomous-enterprise-hausaufgaben/
• valantic Blog, Dr. Sven-Erik Willrich, August 26, 2026, https://www.valantic.com/en/blog/successfully-scaling-claude/
• Händlerbund Guide, “Agentic Commerce: Contract Formation by AI Agents,” August 12, 2026, https://www.haendlerbund.de/de/ratgeber/recht/agentic-commerce
• Clifford Chance, “Agentic AI: The Liability Gap Your Contracts May Not Cover,” February 9, 2026, https://www.cliffordchance.com/briefings/2026/02/agentic-ai–the-liability-gap-your-contracts-may-not-cover.html
• Path to Purchase Institute, “3 Ways AI Will Bring Category Management Into a New Era,” January 9, 2026, https://p2pi.com/3-ways-ai-will-bring-category-management-new-era
• Forrester, “The State of Agentic Commerce in Mid-2026,” Q2 2026, https://www.forrester.com/blogs/the-state-of-agentic-commerce-in-mid-2026/
• Pinsent Masons, “Retailers Need to Respond to the Rise of Agentic AI in Commerce,” April 17, 2026, https://www.pinsentmasons.com/out-law/analysis/retailers-respond-rise-agentic-ai-commerce
• Artificer Legal, “What to Look for in an AI Vendor Contract for Your Retail Business,” August 7, 2026, https://artificer.legal/ai-vendor-contract-clauses-retail-australia
• U.S. Senator Elizabeth Warren, Letter to Google/Sundar Pichai Regarding UCP Partnerships, February 3, 2026, https://www.warren.senate.gov/imo/media/doc/warren_letter_to_google_on_its_ai_shopping_partnerships.pdf
• MindStudio, “What Is Agent Literacy? The Core Skill Every AI Builder Needs in 2026,” June 11, 2026, https://www.mindstudio.ai/blog/what-is-agent-literacy-core-skill-ai-builders
• BizTech Magazine, “AI Agents, Data Governance, and Workforce Shifts Redefine Retail in 2026” (citing IDC), December 23, 2025, https://biztechmagazine.com/article/2025/12/ai-agents-data-governance-and-workforce-shifts-redefine-retail-2026
• eco.com, “Agent Identity Verification: How AI Agents Authenticate Purchases in 2026,” July 16, 2026, https://eco.com/support/en/articles/15192005-agent-identity-verification-how-ai-agents-authenticate-purchases-in-2026
• Iden, “AI Agent Identity Management in 2026: Standards, Players, and the Governance Gap” (citing the CSA-Oasis report), July 15, 2026, https://www.idenhq.com/en/blog/ai-agent-identity-management-2026
• Shopify, “The Agent-Driven Commerce Platform: Shopify Connects Any Merchant to Every AI Conversation,” January 11, 2026, https://www.shopify.com/news/ai-commerce-at-scale
• ChannelEngine, “As AI-referred e-commerce sites surge 393%, ChannelEngine launches tools to help marketplace sellers stay visible” (citing Adobe Analytics), May 12, 2026, https://www.channelengine.com/en/blog/channelengine-launches-ai-tools-for-agentic-commerce-and-marketplace-visibility
• PYMNTS, “What Happens to Stores When AI Agents Do the Shopping?”, February 18, 2026, https://www.pymnts.com/artificial-intelligence-2/2026/what-happens-to-stores-when-ai-agents-do-the-shopping/
• earninglivingonline.com, “AI Shopping Agents: Why Checkout Stalled and Brands Should Focus on Their Own Site,” July 27, 2026, https://www.earninglivingonline.com/ai-shopping-agents-checkout-stalled-buy-on-your-site-2026/
• California Management Review, “The Rise of AI Intermediaries: How Agentic Systems Are Rewiring Customer Relationships,” April 29, 2026, https://cmr.berkeley.edu/2026/04/the-rise-of-ai-intermediaries-how-agentic-systems-are-rewiring-customer-relationships/
• valantic/Handelsblatt Research Institute, “AI at Scale in Retail” (Digital Excellence Outlook 2026, Special Edition), June 2026, https://go.valantic.com/de/whitepaper-ki-im-handel
• Deloitte, “Agentic Commerce: The Future of B2B Commerce,” June 26, 2026, https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/articles/b2b-agentic-commerce.html
• commercetools, “Agentic Commerce in B2B: From Efficiency to Autonomy,” February 19, 2026, https://commercetools.com/blog/agentic-commerce-in-b2b-from-efficiency-to-autonomy