Business-to-Agent (B2A) is the model that emerges when AI agents become the interface between businesses and their customers. This guide explains what changes at each stage of the customer journey, and the infrastructure a business needs to participate.
Why now?
Three things arrived at once. Capable agents: assistants from OpenAI, Anthropic, Google and Perplexity can now browse, reason and take multi-step actions. Agent-accessible payments: new rails let agents transact securely, including Stripe and OpenAI's Agentic Commerce Protocol and Google's Agent Payments Protocol (AP2). Structured exposure: businesses are increasingly publishing machine-readable information and interfaces agents can consume. Fact
The demand signal is large. Gartner describes non-human buyers as “machine customers” and projects they will influence or participate in roughly $30 trillion of purchases by 2030. Fact
The B2A journey
When the customer interface becomes an agent, every stage of the funnel changes shape.
Discover
The agent finds candidates by reading structured data, search results, feeds and knowledge sources, not by scanning a visual page. If you are not machine-legible, you are not in the shortlist.
Understand
It parses what you offer: products, attributes, prices, availability, policies. Ambiguity or missing fields get you dropped.
Evaluate
It compares you to alternatives against the user's constraints (budget, delivery date, ratings, returns). Objective, structured facts win.
Recommend
It presents or selects an option. Being the cited, trusted source is the new “ranking”. See optimising for AI agents.
Transact
It completes the purchase through an agent-accessible checkout and payment rail, ideally without a human re-entering details.
Service
It handles follow-up: order status, returns, support. Gartner expects agentic AI to resolve a large share of routine service autonomously.
How an agent transacts with you
A concrete path: a person asks their assistant to “find a wireless charger under $40 that arrives by Friday with free returns.” The agent queries structured catalogues and search, reads each candidate's machine-readable product data, filters on price, delivery and return policy, and shortlists those it can verify. It then completes checkout through a protocol such as ACP, passing a scoped, tokenised payment credential. The business never sees a human session, only a well-formed request.
What businesses need to change
- Machine-readable information: expose facts as structured data, not just rendered HTML.
- Structured product and service data: Schema.org Product/Offer, availability, pricing, policies.
- Agent-accessible interfaces and APIs: catalogue, cart, checkout and order status callable programmatically.
- Payments: support agent-initiated, tokenised transactions (ACP, AP2, Visa Intelligent Commerce, Mastercard Agent Pay).
- Identity and permissions: verify which agent is acting, for whom, and within what limits.
- Measurement: attribute agent-driven discovery and revenue as a distinct channel.
Full checklist: how to become agent-ready.
The B2A stack
The infrastructure required for Business-to-Agent interactions, from the bottom up.
How agents find you: search, feeds, structured data, llms.txt and crawler access.
Machine-readable product, service, price and policy data an agent can trust.
APIs and protocols (MCP, A2A) that let agents query and act on your systems.
Verifying the agent, its principal, and the scope of what it may do.
Agent-initiated checkout and tokenised payment rails (ACP, AP2).
Tracking agent-driven discovery, decisions and revenue as its own channel.
See the companies filling each layer in the B2A Landscape.
Assess your readiness
See how ready your business is for the agent economy.