Salesforce found that 24% of consumers are already comfortable with AI agents shopping for them, and among Gen Z that number jumps to 32%. Kantar predicts agent-driven commerce could rewrite category shares in 18-24 months. If you are still building your go-to-market strategy around human eyeballs clicking through landing pages, you are optimizing for a shrinking share of buying decisions.
B2A, short for Business-to-Agent, is not just a new acronym to file next to B2B and B2C. It is a structural shift in who evaluates, selects, and purchases your product. This post is not about how to make your API agent-friendly (we covered that in our B2A API design guide). It is about the business strategy question: how do pricing, marketing, and brand loyalty change when a growing share of your customers are autonomous software?
The Numbers: Why B2A Is Not Optional
The scale of the shift is hard to overstate. According to Forrester’s 2026 predictions, one-third of B2B payment workflows will involve AI agents by the end of this year. That includes invoicing, reconciliation, spend control, and supplier selection, all handled autonomously. Vendors like Basware, Coupa, HighRadius, and Ramp are shipping agent capabilities into production right now.
On the consumer side, the infrastructure is already live. Google launched the Universal Commerce Protocol (UCP) in January 2026 at the National Retail Federation, an open standard that lets AI agents handle the entire shopping journey from product discovery to payment to post-sale support. UCP launched with endorsements from Shopify, Target, Walmart, Wayfair, and Etsy, and integrates with Google Pay for agent-initiated transactions. Shopify published their own UCP engineering integration.
The global AI agents market hit $7.63 billion in 2025. Projections put it at $182.97 billion by 2033, a 49.6% compound annual growth rate. The companies building infrastructure for agent commerce are not making speculative bets. They are following the money.
Marketing to Machines: What Changes for CMOs
Here is the part most marketing teams have not thought through. When an agent decides which product to buy, it does not see your brand video. It does not feel the emotional resonance of your tagline. It reads structured data, compares specs, checks latency, and picks the option that best matches its instructions.
Jack Smyth at Jellyfish calls this the shift from CTA to CTAi: “Call to Agent.” Your marketing now has two audiences. The human who sets the agent’s preferences and constraints, and the agent that executes the purchase. Winning the human is necessary. Winning the agent is what closes the deal.
Metadata Is the New Ad Copy
Kantar’s research shows that brand differentiation in B2A depends on metadata marketing: codifying your brand value into structured data that agents can parse. ESG claims, product attributes, certifications, sustainability scores, and loyalty program benefits all need to be machine-readable. An agent comparing two similar products will pick the one whose structured data proves it meets the buyer’s constraints, whether that is “certified B Corp” or “free shipping for loyalty members.”
This does not mean brand building is dead. It means brand building splits into two lanes. Lane one is emotional: the human who tells the agent “I only buy from Patagonia” or “find me something from a German manufacturer.” Lane two is algorithmic: the agent that compares your product data against a competitor’s and picks the winner based on structured attributes.
Audit What LLMs Think of Your Brand
One of Jellyfish’s recommendations that sounds simple but almost nobody does: ask every major LLM what it thinks about your brand. Prompt Claude, GPT-4, and Gemini with “recommend a [your category] product” and see if you show up. If you do not, your product is invisible to every agent built on those models. That is a marketing problem, not an engineering problem.
The fix involves the same fundamentals marketers already know: authoritative backlinks, high-quality structured content, presence in comparison datasets, and consistent NAP (name, address, phone) data. But the optimization target is different. You are not optimizing for a human scanning Google results. You are optimizing for an LLM’s training data and retrieval context.
Pricing When Agents Do the Buying
Per-seat SaaS pricing assumes a human sits in the seat. When an AI agent is the “user,” that model breaks. An agent might make 10,000 API calls in a day, run 500 comparison queries, or complete 200 transactions. Billing it like a single human user dramatically undercharges. Billing it like 200 human users makes agents too expensive to use.
Chargebee’s 2026 pricing playbook identifies three models gaining traction for B2A:
Per-action pricing charges for each workflow an agent completes. A single “action” might involve multiple model calls, RAG lookups, and API requests under the hood, but the customer pays per completed unit of work. Intercom charges $0.99 per AI-resolved support ticket. Zendesk charges $1.50-$2.00 per automated resolution.
Outcome-based pricing ties revenue to business results the customer already tracks: meetings booked, invoices collected, fraud cases prevented, tickets resolved. This aligns incentives perfectly but creates the problem of attribution. What counts as “resolved”? What about partially completed workflows?
Hybrid models combine a base subscription with usage-based components. Chargebee’s own research shows 43% of SaaS companies already use hybrid models, with adoption projected to hit 61% by end of 2026.
The strategic question for B2A is not just “how do we price for agents?” It is “how do we make our pricing agent-parseable?” If an agent comparing three vendors cannot programmatically determine total cost for a given workload, it will pick the vendor whose pricing API returns a clear number. Transparent, API-accessible pricing becomes a competitive advantage.
Google UCP: The Commerce Stack Agents Will Use
The most important piece of B2A infrastructure shipped in January 2026, and most businesses outside of retail have not noticed. Google’s Universal Commerce Protocol is an open standard that creates a common language between AI agent surfaces (like Gemini), merchants, and payment providers.
UCP is not just another API spec. It defines functional primitives for the entire commerce journey: product discovery, inventory checks, cart management, checkout, payment, and post-sale support. It works with existing retail infrastructure and is compatible with the Agent Payments Protocol (AP2) for secure agent-initiated transactions. It integrates with both A2A (Agent-to-Agent) and MCP protocols.
What makes UCP strategically important is who backed it. Google, Shopify, Target, Walmart, Wayfair, and Etsy did not endorse UCP because it was technically elegant. They endorsed it because they see agent-driven commerce as a primary channel within 24 months. When Google’s Sundar Pichai presented UCP at the National Retail Federation, the message was clear: agentic checkout is coming to Google Search and YouTube, and UCP is the protocol.
Meanwhile, Visa and Mastercard are racing to build the payment rails agents will use. Visa enrolled 21 European banks in its Agentic Ready program. Mastercard completed Europe’s first live agent payment with Santander in March 2026.
What B2A Means Beyond Retail
The B2A conversation is dominated by commerce examples because that is where the money is most visible. But the shift applies to every industry where agents make or influence purchasing decisions.
B2B SaaS selection. When a company asks an AI agent to “find us a project management tool that integrates with Jira, supports SOC 2, and costs under $15 per user,” the agent evaluates vendors programmatically. The vendor with the clearest API-accessible feature matrix, compliance documentation, and pricing wins. The vendor with a beautiful marketing site but no structured product data loses.
Financial services. Forrester’s prediction about one-third of B2B payment workflows involving agents extends to supplier selection, invoice processing, and credit evaluation. HighRadius and Ramp are already shipping agent capabilities that autonomously handle AP/AR workflows.
Professional services. When an agent needs to book a conference room, hire a contractor, or select a legal service provider, it will query available options through structured APIs. Professional services firms that expose their availability, pricing, and qualifications through machine-readable formats get discovered. Those that rely on referrals and relationship selling do not.
Healthcare and insurance. Claims processing, provider selection, and benefits comparison are all workflows where agents will increasingly act on behalf of patients and employers. The providers whose formularies, networks, and pricing are agent-accessible will capture agent-driven volume.
Frequently Asked Questions
What is B2A (Business-to-Agent)?
B2A stands for Business-to-Agent, a business model shift where companies must optimize their products, pricing, and marketing for AI agents as customers. Unlike B2B and B2C where humans evaluate products, B2A recognizes that AI agents increasingly make or heavily influence purchasing decisions by comparing structured data, API responses, and machine-readable product information.
How does B2A change marketing strategy?
B2A splits marketing into two lanes: emotional branding for the human who sets agent preferences, and metadata marketing for the agent that executes purchases. Brands need machine-readable structured data (product attributes, certifications, pricing APIs) alongside traditional brand building. CMOs should also audit what major LLMs recommend in their product category, since agents built on those models inherit their knowledge.
How should companies price products for AI agent customers?
Per-seat pricing breaks when agents are the users. Three models are gaining traction: per-action pricing (e.g., Intercom charges $0.99 per AI-resolved ticket), outcome-based pricing tied to business results, and hybrid models combining subscriptions with usage components. Critically, pricing must be API-accessible so agents can programmatically compare costs across vendors.
What is Google’s Universal Commerce Protocol (UCP)?
UCP is an open standard launched by Google in January 2026 that creates a common language for AI agents to handle the entire commerce journey, from product discovery to payment to post-sale support. It is backed by Shopify, Target, Walmart, Wayfair, and Etsy, integrates with Google Pay for agent-initiated transactions, and works with both A2A and MCP protocols.
Does B2A only apply to retail and e-commerce?
No. B2A applies to any industry where AI agents make or influence purchasing decisions. Forrester predicts one-third of B2B payment workflows will involve AI agents by end of 2026. The shift affects SaaS vendor selection, financial services, professional services, healthcare, and insurance. Any business whose products or services can be compared programmatically faces B2A dynamics.