Google Rolls out Agentic Payments Protocol (AP2) – Techie Blog

Yesterday Google rolled out AP2. Key summary bullets

  • I applaud Google’s efforts to advance AP with first focus on enabling a “Trusted Agent Economy”. AP2 (V0.1) on establishing the core architecture and enabling the most common use cases (cards, data payloads to support VC, human in the loop scenarios with step up). 
  • Long list of supporting participants including MA and Amex. However, no other AI platforms, nor Visa, Paze, or US Banks. 
  • Good detailed documentation on initial flows (see Github)
  • Introduction of Verifiable Credentials (VC) as the core of AP2 with a recognition that merchants (who own risk) may also need transaction fraud data. 
  • A twist on the identity provider of VC to become the [Payment] Credential provider, with initial focus on cards, Google has stated goal of designing AP2 to support stablecoin, push payment and other payment types. This “sets up” Visa and Mastercard to retain their roles as the authentication infrastructure for the internet, while also allowing for other networks (India UPI) and seperate identity providers (eID) to operate with the role.
  • My read is that Google has given up hope of making AP2 work in US, as Visa’s intelligent commerce framework is further along.  How tokens, Issuers and networks work within AP2 is not a big technical effort, but there are several things missing from AP2, for example the rule sets (3DS, DAF, TAF, …etc) which the credential (and transaction) operates under. 
  • The framework is solid, authentication will be a huge part of the challenge here.  Payment networks must control how authentication is performed by with their credentials. Visa and mastercard are the authentication infrastructure for the internet for a reason. Its not the technology, it is the governance, standards, enforcement and the operating rules which govern WHO OWNS THE RISK when authentication has broken. See Identity Models and Governance https://blog.starpointllp.com/?p=6470 
  • Of course stablecoins could work here, but guess who owns the risk when something happened that wasn’t authorized? There is no bank to complain to.. Your automated agent made a mistake and you (the consumer) have the loss.
  • AP2 will be successful as the communication protocol for between agents and stakeholders, but it requires credential providers with strong governance and operating rule constructs. Visa, MA, Amex, UPI/UIDAS and PayPal all fit that bill.  The challenge with this dependency is that the control points for progress are complex, as any change in a network requires buy in from existing stakeholders.
  • Expect Google to demonstrate the technical efficacy of AP2 with Stablecoin or Crypto first, and then look to adapt AP2 needs to credential providers
  • While the EU is the best market for Google to begin with, regulators are not keen on doing anything to help US big tech. My recommendation to Google is work on a US focus plan B that will involve US credential providers (ie Visa and Visa banks). AP2 can be the protocol, but most of it will need to operate within the authentication and rules of the credential provider.

You need to be logged in to view the rest of the content. Please . Not a Member? Join Us

Agentic Commerce – I’ve Seen a Lot of “Revolutions”. This One Feels Different.

Hello from Alberta and the Columbia Ice Fields. During the drive I’ve put together a Case study in how Shopify is Building a key piece of the new model. Also drill down on monetization as the Gordian Knot that will determine how Agentic Commerce Operates.

I’ve been in the payments and retail space for almost three decades. I was there for the dot-com boom and bust, the shift from plastic to mobile wallets, the birth of amazon and Google, and the rise of the platform economy. I’ve seen enough hype cycles to fill a library. Each time, we were promised a revolution that would change everything. And while some of those shifts were significant, they were ultimately evolutionary. eCommerce added a channel, mobile created more shopping interactions and more points to influence,  but were mostly the same commercially. 

This time, I have to admit, it feels different. The rapid advancements in Artificial Intelligence aren’t just creating a new channel; they are actively dismantling the foundational economic bargain of the internet. The core paradigms I’ve operated on for my entire career is being invalidated in real-time.

You need to be logged in to view the rest of the content. Please . Not a Member? Join Us

Agentic Commerce – Killing Small Retailers?

I’ve been in the eCommerce game for a while, last month I outlined why I thought edge use cases where the biggest opportunity for Agentic Commerce. The idea was simple. While the big retailers focus on building great CX with their own data, the real, unsolved problems in commerce exist in the long tail ( the small retailers without massive marketing departments). Think about finding an in-stock item for pickup today, or a slightly used version right in your community. That’s where I saw the gold.

Continue reading

Pricing Agentic: Economic Models for a New Kind of Demand

$4B market opportunity (18 mo), who will lead it?

Today’s blog covers possible pricing models and market structures for agentic transactions, a new type of demand (purchase order with a payment instrument). Retailers may despise the idea of a new aggregator, but they can’t say “no” to a PO by their customer. US retailers spend over $400B on marketing ($90B of which is digital marketing). There is no CAC for an agentic transaction.. While the daily innovations of AI and Agentic is fascinating, it is the economics and structures for pricing value that will influence participation, value creation and market success.

Continue reading

2025 Tech Tsunami will Expand Network Role (and VAS)

A snarky blog. My views on why the role of card networks will grow in the midst of this change (along with Network VAS).

Buckle up buttercups, because the commerce, banking, and payments world is getting a facelift so extreme, it’ll make a Kardashian look like a Luddite. If you thought Web 3.0 and its decentralized pipe dreams were the next big thing, bless your heart. AI and Agentic Commerce are the actual party, and they’re about to flip the table.

Continue reading

Agentic’s Real Economic Opportunity: Edge Use Cases

This blog is a monster 25 pages. Hence the executive summary

Transformation seldom starts at the core of an existing market or business, but rather starts in edge use cases of unmet needs and price performance (ie, Innovators Dilemma). Edge UCs may be the biggest economic opportunity for Agentic, organizing hyperlocal, DTC and the food chain to create a new kind of market. If these edge UCs can be addressed in the US, it also greatly expands the global market opportunity as small regional needs resemble the edge.

The consumer benefit of this approach is an awareness of the very different “purpose” of agentic platforms vs Amazon or Walmart.  Imagine identifying which businesses and service providers could address your needs locally. 

Continue reading

Agentic Commerce Economics and Governance

The “Agentic Era,” promises to revolutionize commerce and customer experience by automating complex purchasing tasks. Agentic’s transformative potential is constrained by the lack of clearly defined shared economic models. 

While the vision of a decentralized Web 3.0 remains unfulfilled, the Agentic Era presents its own set of complex economic questions regarding value creation, distribution, and governance. This blog explores the challenges in establishing shared economic models for agentic commerce by taking a look at Transaction Cost Economics (TCE) and Network Theory to analyze the interplay among consumers, merchants, AI agent platforms, and other stakeholders. We address issues of value attribution, data monetization, trust, risk allocation, permissions and the necessity for robust governance structures beyond mere technical interoperability.

You need to be logged in to view the rest of the content. Please . Not a Member? Join Us

Retailer Actions in Agentic Commerce

Navigating a New Demand Paradigm and Its Two-Sided Imperative

The rise of agentic commerce, where AI agents could potentially execute purchases for consumers, signals more than an evolutionary step in e-commerce; it represents the emergence of a new type of demand: a direct customer buy order, theoretically complete with payment authorization. This presents a two-sided imperative for retailers. Firstly, ensuring your products are discoverable and favorably considered by these AI agents an “SEO for the agentic era.” Secondly, developing the organizational capacity to act on this demand, potentially bypassing traditional e-commerce pathways for direct fulfillment via APIs, with updates to fraud and risk screening.

Continue reading

Big News – Google I/O 2025 – Payments View

Google I/O is going on today and tomorrow and they just announced a vast array of new products and services (Google CEO’s blog). Today’s blog is a quick drill down on payments and specifically what “Buy for me with GPay” means for the payment ecosystem (and agentic commerce).

As discussed in Commercial Models for AI Agents, the network and economics surrounding agentic commerce is far from settled. The lack of a clear commercial model and a robust trust framework has impeded Agentic’s growth as commerce is a multi-sided network. Google’s strategy appears to directly address these deficiencies by deeply integrating GPay as a core component of its agentic offerings. This integration aims to provide the necessary layers of trust, security, and transactional capability that have been missing. It also may provide an additional pricing mechanism for agentic transactions.

As outlined in today’s Stratechery the “lack of payments” were part of the original sin of the internet. A “sin” that Google is fixing in Agentic by creating a complex network that unites search, ML, agent-based action, payments, advertising network, billions of devices, consumer-controlled data for personalization, which will redefine eCommerce (and recapture product search). The price of entry? Merchants need to add the GPay button.

To be clear, merchants will still endeavor to use AI in order to create a better customer experience for those customers that enter their domain. But for consumers, the Google offering will be hard to beat as Google leverages their data and preferences across every device to enable customer interaction through purchase.  While Amazon will likely maintain a solid position, most consumers will not start search within a merchant domain. Agentic originated transactions present a new type of demand, fully qualified consumers with a valid payment instrument and transaction request. A transaction type that should operate in a 100% conversion model (ie no abandonment). With GPay, Google provides the consumer authentication and risk data for merchants to decision the transaction.

Continue reading

Visa Product Drop – Enabling Agentic 

Today Visa provided products, partnerships and demonstrations of “Visa Intelligent Commerce enables AI to find and buy”. This is a very significant effort from Visa that extends their leadership as the payment and identity infrastructure of the internet, to new era of AI and Agentic Commerce. 

What is the new problems it is solving? Payments geeks understand Card Present (CP) and Card Not Present (CNP) transactions, but agents raise the question of WHO authorized the agent and for what purpose?  Visa solves this problem by extending existing facilities like tokenization and FIDO based authentication, with agents receiving finer-grained authorization (from consumer) using limited-use tokens that restrict authorization and consumer control (expiration/time, amount and merchant).   

Within an agentic transaction flow, authentication by the Agent operator is not good enough for the merchant that will bear the risk in the transaction, the merchant must also be able to authenticate the consumer and obtain authorization for the transaction (see Tokens and Binding 101 and Separating Payments and Identity).

Continue reading