All articles
AI & Commerce··11 min read

Your Next Biggest Customer Will Not Be Human - Here Is How to Win Their Business

AI agents are starting to buy things on behalf of people and companies. The businesses that prepare now will capture a $15 trillion opportunity.

Picture this scenario. It is a Tuesday morning. Your VP of Sales pulls you into a call because something remarkable is happening in the numbers. Revenue from one of your top product categories jumped 22% last month. But website traffic is flat. No new marketing campaigns. No seasonal bump. No press coverage. When the team digs into the server logs, they find the answer: a wave of purchases initiated not by humans browsing your site, but by AI agents acting on behalf of procurement departments at companies you have never spoken to.

No one called your sales team. No one read your case studies. No one sat through a demo. Software evaluated your products against your competitors, decided you were the best option, and completed the purchase - all without a single human being on the buying side ever visiting your website.

If that sounds like science fiction, it should not. The infrastructure to make this happen is being built right now by Google, Shopify, IBM, Stripe, Visa, Mastercard, and dozens of other companies. Gartner projects that AI agents will intermediate more than $15 trillion in B2B purchases by 2028 and that 90% of all B2B purchases will flow through automated exchanges within three years.

This is not a technology story. It is a business strategy story - and for leaders who move early, it represents the most significant new revenue channel since companies first started selling online. The question is not whether agent commerce will arrive. It is whether your business will be visible to these new buyers when it does.

The Protocols Making This Real

Three pieces of infrastructure are turning "AI agents buying things" from a concept into a reality. You do not need to understand how they work technically. You need to understand what they enable, who is building them, and what they mean for your business.

UCP - Universal Commerce Protocol

UCP is an open standard co-developed by Google with Shopify, Etsy, Wayfair, Target, and Walmart. It has been endorsed by more than 40 companies including Stripe, Visa, Mastercard, American Express, PayPal, Best Buy, Sephora, and The Home Depot. It launched in early 2026 and is already being integrated into Google's AI Mode in Search and the Gemini app.

What it does: UCP gives AI agents a standardized way to discover products, check real-time pricing and inventory, manage checkout sessions, and track orders - all without scraping a website or needing a custom API. Think of it as a universal language that lets any AI shopping agent talk to any retailer's system. It uses REST and JSON-RPC, the same foundational web technologies that power most of the internet today.

Why it matters to you: If your product catalog is accessible through UCP, AI shopping agents across the ecosystem can discover, evaluate, and purchase from you automatically. The companies adopting UCP early are positioning themselves the way early e-commerce adopters did in the late 1990s - not by predicting the exact timeline, but by making sure they are visible when the wave arrives.

ACP/A2A - Agent Communication Protocols

The Agent Communication Protocol was developed by IBM Research and contributed to the Linux Foundation in March 2025 as part of the open-source BeeAI platform. In September 2025, ACP merged with Google's Agent2Agent (A2A) protocol under the Linux Foundation umbrella, creating a unified standard for how AI agents talk to each other. The combined effort now has support from more than 100 technology companies.

What it does: A2A (incorporating ACP's innovations) lets AI agents discover one another, delegate tasks, share context, and report results using a common format. In a purchasing scenario, a company's procurement agent might use A2A to coordinate with a price-comparison agent, a compliance-checking agent, and a vendor-relationship agent - each built by different companies, all communicating through the same protocol.

Why it matters to you: The buying decision is becoming a coordinated process across multiple specialized agents. Businesses that structure their product information for this multi-agent ecosystem - clear specifications, reliable inventory data, transparent pricing - will surface naturally in agent-driven evaluation chains.

ATXP - Agent Transaction Protocol

ATXP was created by Circuit & Chisel, a startup founded by former Stripe executives (the former Head of Crypto & AI Partnerships and the former lead of Crypto Engineering). The company raised $19.2 million in seed funding from investors including Stripe itself, Coinbase Ventures, Solana Ventures, Samsung Next, and Cloudflare's CEO Matthew Prince.

What it does: ATXP enables AI agents to handle payments autonomously - including instant micropayments, nested transactions (transactions within transactions), and delegated payments where one agent authorizes another to spend on its behalf. It operates across traditional payment rails and stablecoin infrastructure, giving agents what amounts to their own wallets.

Why it matters to you: This is the piece that makes autonomous purchasing actually work at the payment layer. When an AI agent decides to buy from you, ATXP is the protocol that lets it actually pay. Cloudflare's CEO described ATXP as "an important protocol to help ensure agent-initiated commerce" succeeds "in a secure and trusted way."

These three protocols together - UCP for product discovery, A2A for agent coordination, ATXP for payments - are assembling the complete infrastructure for a world where software buys things from other software, on behalf of humans, at massive scale.
The Agentic Commerce Stack Diagram showing the three protocol layers enabling autonomous agent commerce: UCP for product discovery, A2A for agent coordination, and ATXP for agent payments. THE AGENTIC COMMERCE STACK Three protocols powering autonomous agent commerce UCP Product Discovery Google + Shopify + 40 partners Catalog, pricing, inventory, checkout A2A Agent Coordination IBM ACP + Google A2A via Linux Foundation Task delegation, context sharing ATXP Agent Payments Circuit & Chisel (ex-Stripe) + Stripe, Coinbase Micropayments, delegated wallets THREE PROTOCOLS, THREE ORGANIZATIONS, ONE DIRECTION
FIGURE 1 - The three protocol layers enabling autonomous agent commerce

How Your Business Changes - and How Smart Companies Are Adapting

As a CEO, you do not need to implement these protocols yourself. Your engineering team will handle that. What you need to understand is how your business changes when a meaningful percentage of your buyers are software agents rather than humans - and how the companies leading this transition are already positioning themselves to benefit.

Your sales process evolves into a dual-track strategy

Most B2B sales processes are designed around human psychology. A skilled salesperson reads body language, adjusts the pitch, builds rapport, handles objections, creates urgency. Your marketing team writes emotionally compelling copy. Your website uses social proof, testimonials, and carefully designed user journeys to guide a human toward a purchase decision.

An AI agent evaluating your product does not respond to any of this. It parses your structured product data, compares it against its criteria and your competitors' data, and makes a decision based on price, specifications, availability, and whatever other parameters its human operator defined.

This does not mean your sales team or brand becomes less valuable - human buyers still respond to relationships, trust, and emotional resonance. But it does mean a growing segment of your purchase volume will come from buyers that evaluate you on structured data alone. In practice, the companies preparing effectively as documented in public reports are building a dual-track approach: relationship-driven selling for high-value human buyers, and machine-readable product data and APIs for agent-initiated purchases. Both tracks feed the same revenue line.

Liability requires proactive framing, not reactive scrambling

Consider this scenario. A company deploys a procurement agent with a mandate to reorder industrial supplies whenever inventory drops below a threshold. The agent, operating within its authorized parameters, places an order with your company for a chemical product. But the agent's parameters were poorly configured, and it ordered a quantity that exceeds safe storage limits at the buyer's facility.

Who is liable? The company that deployed the agent? The company that sold the product? The developer who built the agent software? The protocol that facilitated the transaction?

Current legal frameworks do not have clean answers. As the law firm Proskauer Rose noted in a recent analysis: when an AI agent clicks "Accept" on terms of service, the question of who actually consented is legally unresolved. The Lathrop GPM firm highlighted that companies deploying AI agents may face strict liability for agent conduct, whether or not the outcome was predicted or intended. And the EU AI Act is pushing for mandatory human oversight requirements that may affect how autonomous purchasing is structured.

The important thing to understand is that this uncertainty is not a reason to wait - it is a reason to act now. The companies that establish clear terms of service for agent-initiated transactions, define liability boundaries in their vendor agreements, and build audit trails into their agent interactions will be the ones setting the standard. The legal complexity of new commercial channels historically favors the organizations that proactively define the rules rather than those forced to accept someone else's framework after the fact.

Fraud prevention evolves alongside the opportunity

Traditional fraud detection systems were built to catch humans behaving suspiciously - device fingerprints, IP geolocation, behavioral analytics, velocity checks. When the buyer is an AI agent operating from a cloud server, those signals look different.

The payment networks are already addressing this. Visa published a threat analysis in early 2026 specifically about fraud risks in agentic commerce and is developing new detection frameworks. The ATXP protocol itself was designed with security as a foundational concern - agent identity attestation, transaction protocol verification, and spending scope validation are built into the standard, not bolted on after the fact. Cloudflare's CEO specifically endorsed ATXP as a protocol that ensures agent commerce succeeds "in a secure and trusted way."

A pattern emerging from public deployment reports is straightforward: companies that upgrade their fraud and security stack as part of the agent commerce rollout - rather than treating security as a separate initiative - move faster and more confidently. New signals like agent identity verification and delegated spending limits actually provide more structured accountability than the human-browsing-a-website model they replace.

Your competitive moat expands, not shrinks

In human commerce, brand, design, user experience, and marketing are powerful competitive advantages. In agent commerce, what matters is the quality and completeness of your structured product data, the reliability of your inventory signals, the competitiveness of your pricing, and the frictionlessness of your checkout API.

Here is the opportunity that many leaders miss: this is not a zero-sum shift. Companies with strong brands and strong data win on both tracks. The businesses that invest in clean, structured, UCP-compatible product catalogs do not lose their human customers in the process. They gain an entirely new buyer segment. A well-structured product catalog improves your human shopping experience too - better search, more accurate recommendations, fewer inventory mismatches.

Human Buyers vs. Agent Buyers Comparison diagram showing what wins human buyers versus what wins agent buyers, illustrating that companies with strong data win on both tracks. HUMAN BUYERS vs. AGENT BUYERS Competitive advantages are expanding, not replacing each other HUMAN BUYERS Brand recognition Emotional marketing copy Beautiful UX and photography Sales team relationships AGENT BUYERS Structured product data quality Real-time inventory accuracy API reliability and speed Competitive pricing transparency COMPANIES WITH STRONG DATA WIN ON BOTH TRACKS
FIGURE 2 - The competitive advantages that matter are expanding, not replacing each other

Five Strategic Moves for CEOs

You do not need to become a protocol expert. You need to make five strategic decisions that position your business to capture agent-driven revenue as this channel scales.

1. Audit your product data - this is the foundation of everything else. Ask your team a simple question: if an AI agent queried our product catalog right now, would it get accurate, complete, structured data for every SKU? If the answer is anything less than an unqualified yes, this is your starting point. Clean, structured product data improves your human commerce experience and your agent commerce readiness simultaneously. It is the rare investment that pays off regardless of how quickly agentic commerce reaches your industry.

2. Get ahead of agent liability by defining your terms now. Work with your legal team to understand your exposure on both sides: as a seller receiving agent-initiated purchases, and potentially as a buyer deploying procurement agents. Update your terms of service. Clarify your return and dispute policies for non-human-initiated transactions. The companies that define the legal framework early will set the standard for their industry - a significant advantage over those who discover the gaps during a dispute.

3. Ask your commerce platform vendor about their UCP timeline. Shopify, Salesforce, and Stripe have all announced UCP implementation plans. If your commerce infrastructure runs on one of these platforms, you may get UCP compatibility as a platform update. If you run custom infrastructure, you need a plan. Either way, you need to know the timeline. Platform-level protocol adoption historically tends to move faster than most enterprises expect.

4. Upgrade your fraud and security stack for the agent commerce era. Talk to your payments team and your fraud vendor about how they are preparing for agent-initiated transactions. The new generation of fraud signals - agent identity attestation, transaction protocol verification, spending scope validation - are in many ways more structured and auditable than the human behavioral signals they supplement. Companies that adopt these alongside their agent commerce rollout report higher confidence in transaction integrity, not lower.

5. Explore agent-to-agent selling as a competitive advantage. If your customers' procurement agents will be buying from you, should you have sales agents meeting them on the other side? Agent-to-agent commerce - where a seller's AI negotiates with a buyer's AI on pricing, terms, and fulfillment - is not a distant concept. McKinsey has identified it as one of the key scenarios in the agentic commerce landscape. Companies that build agent-side selling capabilities early will have a structural advantage in automated procurement channels - and the institutional knowledge to iterate quickly as the channel grows.

The Window of Advantage

The timelines here are compressed in a way that favors early movers. UCP went from announcement to implementation partners in months. ACP and A2A merged and gained 100+ company support within a year. ATXP went from founding to $19 million in funding and a working protocol in under a year. Google is already integrating UCP-powered checkout into AI Mode in Search.

Regulation is coming - Gartner projects a significant increase in autonomous system oversight requirements by the end of 2026. But the commerce infrastructure is arriving faster than the regulations. That gap is where the first-mover advantage lives. The companies that establish their agent commerce presence now get to shape the norms, build the institutional knowledge, and capture the early revenue before their competitors even begin.

This is not about predicting the exact month when agent commerce becomes mainstream in your industry. It is about making sure your business is visible, accessible, and optimized for the new buyer segment when it arrives - because the investments that prepare you for agent commerce (clean data, reliable APIs, clear legal terms) also improve your existing human commerce operations.

The companies that approach this transition with a clear plan and experienced implementation partners will not just adapt to agent commerce. They will lead it.

This article is general commentary on emerging agent-commerce protocols, not legal advice. Consult qualified counsel before adjusting your terms of service or liability framework for agent-initiated transactions.

Frequently Asked Questions

What are UCP, ACP/A2A, and ATXP in simple terms?

UCP (Universal Commerce Protocol) is a standard co-developed by Google and Shopify that lets AI agents discover and buy products through a common language. ACP (Agent Communication Protocol) was created by IBM and has merged with Google's A2A (Agent2Agent) protocol under the Linux Foundation - it lets AI agents communicate and coordinate with each other. ATXP (Agent Transaction Protocol) was built by former Stripe engineers and handles the actual payments when agents buy things autonomously. Together, they form the infrastructure for AI-driven commerce.

How soon will AI agents actually be buying things for businesses?

It is already beginning. Google is integrating UCP-powered checkout into AI Mode in Search right now. Shopify, Salesforce, and Stripe have announced UCP implementation plans. Gartner projects that AI agents will intermediate over $15 trillion in B2B purchases by 2028, with 90% of B2B purchases flowing through automated exchanges within three years. The infrastructure is being deployed today - the practical question is how to position your business to capture this revenue as adoption scales in your industry.

Who is liable if an AI agent makes a bad purchasing decision?

This is one of the most significant evolving legal questions in agentic commerce. Current legal frameworks were not designed for autonomous purchasing agents, and companies that deploy agents may face strict liability for agent conduct. The EU AI Act mandates human oversight requirements that will shape how autonomous purchasing is structured. The companies that are navigating this well are proactively defining agent transaction terms in their vendor agreements, building audit trails, and working with legal counsel to establish clear liability boundaries before disputes arise - rather than waiting for case law to develop reactively.

Will agentic commerce replace human buyers entirely?

No. Human buyers will continue to drive purchases where relationships, judgment, and subjective preferences matter - luxury goods, complex enterprise services, strategic partnerships. But for routine, specification-driven, and repeat purchases - especially in B2B procurement - agents will handle an increasing share of the volume. The opportunity here is additive: agent commerce opens a new buyer segment that operates alongside your human customers, not instead of them.

What should my company do first to prepare?

Start with your product data. If an AI agent queried your catalog today, would it get accurate, complete, structured information for every product? That is the foundation - and it benefits your human commerce experience too. Then talk to your commerce platform vendor about their UCP timeline, since Shopify, Salesforce, and Stripe are all building UCP support. Finally, have a conversation with your legal team about agent-initiated transactions and your terms of service. These three steps require modest investment and position you well regardless of how quickly agentic commerce reaches your specific market.

How does this affect my existing sales team?

Your sales team will remain essential for complex, relationship-driven sales - and those deals are typically your highest-margin revenue. For transactional and repeat purchases, the buyer increasingly will be software rather than a person your salespeople can call or email. The most effective approach is a dual-track sales strategy: human-to-human for high-value relationship selling, and machine-readable APIs and structured data for the growing volume of agent-initiated purchases. Companies that serve both tracks capture revenue that competitors focused on only one track will miss.

Code Atelier · NYC

Ready to get agent-ready before your competitors do?

Let's talk