Agentic Commerce Protocols: ACP, UCP, AP2 and Verifiable Intent

Traditional e-commerce was built for humans. Every interaction assumes a person is browsing a screen, reading descriptions, comparing options, and clicking buttons. When the customer is an autonomous AI agent - a machine customer - none of these assumptions hold.

Machine customers do not browse websites. They query APIs. They do not read marketing copy. They parse structured data. They do not click "Add to Cart." They execute transactions through programmatic interfaces. This means that the entire interaction layer of commerce must be redesigned for machine-to-machine communication.

Agentic commerce protocols are the answer. They provide the standardised interaction patterns that enable AI agents to engage with businesses at scale - discovering products, evaluating offerings, negotiating terms, and completing transactions without human intervention on either side.

Agentic commerce protocols operate across four layers, each addressing a different phase of the commercial interaction:

Discovery Protocol. How do agents find products and services? The discovery layer defines how businesses expose their offerings in machine-readable formats - structured product data, API endpoints, capability declarations, and availability signals. Without a discovery protocol, machine customers cannot find you.

Evaluation Protocol. How do agents assess and compare offerings? The evaluation layer defines how businesses communicate quality signals, performance metrics, pricing structures, and terms of service in formats that agents can systematically compare. This is where Signal Clarity - the first pillar of AXD Readiness - becomes operational.

Negotiation Protocol. How do agents and businesses negotiate terms? The negotiation layer defines the interaction patterns for price negotiation, bundling, volume discounts, delivery terms, and conditional offers. Machine-to-machine negotiation happens in milliseconds and can explore far more options than human negotiation.

Transaction Protocol. How are purchases completed and confirmed? The transaction layer defines payment processing, order confirmation, delivery tracking, and receipt generation for autonomous transactions. It must handle the unique requirements of delegated purchasing - including authority verification and constraint enforcement.

The most immediate protocol requirement is Signal Clarity - making your business legible to autonomous agents. This is not traditional SEO. It is the practice of structuring your commercial information so that machine customers can discover, parse, and evaluate it programmatically.

Signal Clarity requires:

Structured product data. Machine-readable product descriptions using standardised schemas (Schema.org, Open Product Data). Every attribute - price, dimensions, materials, availability, delivery terms - must be explicitly declared in structured format.

API accessibility. Product catalogues, pricing, availability, and ordering capabilities exposed through well-documented APIs. Agents interact through APIs, not web pages.

Performance metrics. Verifiable data on delivery accuracy, return rates, quality consistency, and customer satisfaction. Machine customers evaluate trust through data, not brand perception.

Capability declarations. Explicit statements of what your business can do - product categories, service areas, delivery capabilities, negotiation flexibility. Agents need to know your capabilities before they can evaluate your offerings.

In agentic commerce, reputation is not built through brand marketing - it is built through verifiable reliability. Machine customers evaluate suppliers based on objective performance data, not subjective brand perception.

The key metrics that define reputation in agentic commerce include:

Delivery accuracy. What percentage of orders are delivered on time, to specification, and without error? This is the primary trust signal for machine customers.

Quality consistency. How consistent is product quality across orders? Variance is a risk signal that machine customers penalise heavily.

Dispute resolution. How quickly and fairly are disputes resolved? The dispute resolution protocol is a critical component of agentic commerce infrastructure.

API reliability. How available, responsive, and accurate are your machine-facing interfaces? Downtime, latency, and data inconsistency directly reduce your visibility to machine customers.

Agentic commerce protocols are still emerging. No single standard dominates, and the landscape is evolving rapidly. Several developments are shaping the direction:

Schema.org extensions. The existing Schema.org vocabulary for products, offers, and organisations provides a foundation, but extensions are needed for agent-specific requirements - delegation verification, constraint communication, and negotiation parameters.

Agent-to-agent communication. Protocols for direct agent-to-agent interaction - where a buyer's agent negotiates with a seller's agent - are in early development. These protocols must handle identity verification, authority delegation, and dispute escalation.

Know Your Agent (KYA). Emerging regulatory frameworks require businesses to verify the identity and authority of machine customers before completing transactions. KYA protocols are the agentic commerce equivalent of Know Your Customer (KYC) in financial services.

The AXD Institute tracks these developments through its Observatory research programme and publishes practical guidance through the Four Pillars of AXD Readiness framework - helping businesses prepare for the protocol requirements of agentic commerce before standards are finalised.