Your next customer might be an AI agent news update
Agentic commerce is expanding from simple product discovery into multi-tier autonomous purchasing governed by cryptographic protocols like Google AP2.
By Dillip Chowdary • Nov 21, 2026 • Source: jkfsom.com

At a glance
- Product
- Agentic commerce tiers
- Tiers
- Level 1 to Level 4
- AP2 creator
In this post
Commerce is shifting toward automated software buyers as consumer-facing artificial intelligence evolves from generating text recommendations to completing financial transactions across the web. While merchants conducted transactions across thousands of miles via letter and credit as early as 1900 BCE, followed by sixteenth-century Venetian catalogue shopping and nineteenth-century home shopping, automated systems now evaluate pricing, technical specifications, and merchant reputability to execute purchases on behalf of buyers, according to jkfsom.com's report. Modern agentic commerce breaks down this structural shift into four technical tiers ranging from cross-catalog product discovery to autonomous multi-vendor transaction orchestration.
This analysis examines the operational architecture of these four distinct tiers, detailing the protocols, authorization mechanisms, and technical boundaries governing AI-driven transactions. It is designed for merchants, payment infrastructure engineers, and platform operators preparing for automated checkout interactions where software agents act as the direct customer.
| Level | Role of AI | Example | Human Involvement |
|---|---|---|---|
| 1. Discovery | Researches and recommends products | Find the best electric whisk under £25 | Selects and purchases |
| 2. Assisted purchasing | Facilitates a transaction | Purchase an iPhone via an AI assistant | Approves the purchase |
| 3. Conditional autonomy | Monitors and purchases within predefined limits | Buy Clarks desert boots (tan; UK 12) when they fall below £40 | Sets conditions in advance |
| 4. Orchestration | Plans and coordinates complex, multi-step transactions | Arrange travel to my parents in Austin, TX, on the weekend of Nov. 21st. Preference for daytime travel. | Sets objectives and approves exceptions |
01Cross-catalog product discovery systems
Level 1 systems conduct automated product research without placing orders or negotiating checkout prices. These agents evaluate consumer requirements alongside pre-existing user context to scan multiple retailers and isolate matching items, such as an electric whisk under £25. While legacy recommendations from major marketplaces remain constrained to single proprietary catalogs and browsing histories, AI agents benchmark alternatives across independent merchant inventories.
Consumers use discovery agents to evaluate complex purchase requirements with substantially lower search effort. The agent does not execute financial settlement or transmit checkout credentials directly to the retailer. However, it alters customer acquisition by determining which merchant storefront receives the eventual qualified buyer.
Discovery tools are already broadly accessible across commercial platforms including ChatGPT, Google's Gemini, and Amazon's Alexa for Shopping. While output quality varies across implementations, these systems handle baseline multi-store comparison and contextual filtering.
02Assisted purchasing and merchant checkout handoffs
Level 2 architecture expands the agent role from product filtering to handling transaction mechanics. After surfacing a shortlist, the user instructs the assistant to proceed directly with an order, such as purchasing an iPhone. The agent confirms product availability, inputs customer delivery information, and prepares payment processing on the merchant backend.
The primary operational benefit of assisted purchasing is removing checkout friction, enabling merchants to receive and fulfill orders without the shopper visiting the retailer's storefront. The user retains complete authority over the final payment decision and must provide explicit confirmation before funds transfer.
Adoption of assisted purchasing remains uneven across major platforms. OpenAI launched in-chat purchasing via Instant Checkout in September 2025 before abandoning the mechanism in March 2026 to redirect shoppers to merchant websites. Alternative implementations include Perplexity's Instant Buy, Google's Universal Cart launched in the United States in October 2026 with Google Pay integration, and OpenAI's ChatGPT Wallet for eligible US subscribers.
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
03Cryptographic mandates for conditional purchasing
Level 3 introduces conditional autonomy, shifting human intervention from approving individual checkouts to defining advance rules for future transactions. Users designate specific constraints, such as purchasing a pair of navy Clarks desert boots in UK size 12 when the price drops below £65 inclusive of postage and packaging. The agent monitors retailer pricing feeds and discount codes over days or weeks, executing the purchase instantly once conditions align.
Independent background purchasing requires verifiable trust between retailers, users, and autonomous software. Google's Agent Payments Protocol (AP2) addresses this authorization challenge by issuing cryptographically signed mandates that establish verifiable evidence of consumer consent. These protocol mandates outline explicit bounds for maximum spending limits, product specifications, and transaction rules so merchants and payment processors can verify the agent acts within certified parameters.
Commercial conditional purchasing exists today in restricted environments. Amazon's Alexa for Shopping offers an Auto-Buy tool that monitors items and completes checkouts automatically when target prices trigger. General-purpose conditional purchasing across open web retailers remains gated behind standardized authorization protocols.
04Interdependent planning across multiple merchant vendors
Level 4 orchestration moves beyond single-item checkouts to coordinate multi-step transactions across independent business entities. A user specifies a high-level goal, such as booking a three-night trip to New York departing on 30 December for two people with a £1,500 per-person budget. The agent plans interdependent flight bookings, hotel reservations, and airport transfers without requiring itemized human setup.
Orchestration differs fundamentally from conditional purchasing because individual sub-transactions depend on adjacent booking outcomes. Flight schedules must match hotel check-in windows, airport transit must align with arrivals, and cumulative costs must not exceed predefined spending caps. If a hotel price spikes or an airline seat sells out, the agent dynamically recalibrates connected reservations rather than terminating the workflow.
Note: Fully autonomous multi-vendor orchestration remains largely aspirational across commercial platforms, with general reliability across diverse APIs remaining limited despite niche travel booking deployments.
05Platform deployment and payment protocol evolution
Current market implementations reflect uneven progress across the four transaction tiers. Level 1 tools are universally deployed across conversational interfaces, while Level 2 systems operate through fragmented implementations such as US-based digital wallets and single-platform carts. Commercial Level 3 capabilities remain confined to specialized tools like Alexa's Auto-Buy, though cryptographic mandates under AP2 are laying the groundwork for open-web conditional settlement.
Transitioning to advanced tiers shifts merchant discovery away from traditional search optimization toward structured machine readability. Software buyers prioritize standardized inventory metadata, programmatically verifiable discount structures, and protocol-level API endpoints over visual storefront design. OpenAI Instant Checkout launch: September 2025 OpenAI Instant Checkout sunset: March 2026 Google Universal Cart launch: October 2026
Frequently asked questions
What is agentic commerce?
Agentic commerce is the use of AI agents to research, evaluate, and purchase products on behalf of consumers. Depending on their authorization level, agents range from recommending items to executing complex purchases autonomously across multiple retailers.
What is Google's Agent Payments Protocol?
Google's Agent Payments Protocol (AP2) uses cryptographically signed mandates to verify what an AI agent is authorized to purchase. Merchants and payment processors use these mandates to validate spending limits and product constraints before executing a transaction.
How did OpenAI handle in-chat purchasing between 2025 and 2026?
OpenAI launched in-chat purchasing through Instant Checkout in September 2025, but abandoned the feature in March 2026 in favor of directing consumers to merchant checkouts. The company subsequently introduced ChatGPT Wallet for eligible US subscribers to authorize agent purchases.
What is the difference between Level 2 and Level 3 agentic commerce?
Level 2 assisted purchasing requires explicit human confirmation for every single transaction an agent facilitates. Level 3 conditional autonomy allows the user to set pricing and product rules in advance, letting the agent complete purchases automatically in the background once those conditions are met.
Sources
Author
Dillip Chowdary
Writes Tech Bytes coverage of AI, engineering, and the tools that actually ship. Editor of Tech Pulse Daily.
Related on Tech Bytes
Advertisement