Gartner projects that 90% of B2B buying will be AI-agent-intermediated by 2028, pushing $15 trillion through agent-to-agent exchanges. McKinsey estimates agentic commerce could redirect $3-5 trillion in global consumer commerce by 2030.

Those forecasts are getting a lot of airtime. The underreported part is that most enterprise commerce platforms can't serve an AI buyer today. The technology to support AI buyers already exists; the gap is that pricing, promotion, and product infrastructure was designed for humans to interpret, not for machines to consume.

Your Infrastructure Is the Bottleneck

When a human shops your B2B portal, they can read a PDF rate card, call a sales rep for volume pricing, interpret a "call for quote" button, and figure out that the 15% loyalty discount stacks with the seasonal promo but not the clearance price.

None of that is accessible to an AI purchasing agent.

AI agents need structured, machine-readable data at every step: what products exist, what they cost for this specific buyer at this specific volume, which promotions apply, whether those promotions combine, and what the final price is before committing to a transaction. If any of that requires human interpretation, the agent moves on to a competitor that makes it easier.

Most of the current discussion focuses on which protocol will win, whether it's Google's Universal Commerce Protocol or OpenAI's Agent Commerce Protocol, while skipping over a more immediate question: whether your commerce infrastructure can respond to the queries those protocols will generate.

What AI Agents Need From Your Commerce Platform

An AI purchasing agent operating on behalf of an enterprise buyer will need to:

Resolve pricing without a phone call. Tiered pricing, contract pricing, volume breaks, and customer-segment-specific rates all need to be API-accessible. If your pricing engine requires a sales rep to quote, you're invisible to agents. Broadleaf's pricing services handle this well, returning fully resolved prices by customer segment, volume tier, and currency through standard API calls.

Understand promotion logic programmatically. "Buy 2 get 1 free" needs to be a structured rule an agent can evaluate, not a banner image on a product page. Combinability rules, minimum purchase requirements, and stacking logic all need to be queryable. If your promotions only exist as marketing copy, agents can't factor them into purchase decisions.

Multi-catalog hierarchies, variant-based products, configurable bundles, and product options all need clean API responses too. An agent evaluating whether to buy 500 units of a configurable product across three warehouses needs structured data at every level, not a JavaScript-rendered product detail page.

Can an agent actually complete a purchase without a human stepping in? Cart creation, fulfillment selection, payment authorization, and order confirmation all need to work through APIs. If any step requires manual intervention, the transaction stalls. At that point, headless and API-first architecture crosses from a technical preference into a business requirement.

The B2B Blind Spot

Almost every article about agentic commerce focuses on B2C retail: consumers using AI assistants to find the best deal on headphones. The more consequential shift, though, is happening in B2B, and most enterprise commerce teams have not yet started to address it.

When Gartner says $15 trillion in B2B purchases will flow through AI agents by 2028, they're describing a world where your buyer's procurement agent negotiates with your seller's commerce platform automatically. Contract pricing, approval workflows, account hierarchies, and quote negotiation all need to work without a human in the loop.

For B2B commerce platforms built around manual quoting and relationship-based pricing, this is a structural problem. The platforms that expose their full pricing and ordering capabilities through APIs will capture agent-driven demand. The ones that hide pricing behind "contact sales" buttons won't even be considered.

Why 40% of Agent Projects Get Canceled

Gartner predicts that more than 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls. The common pattern: companies invest in the AI layer (chatbots, recommendation engines, purchasing agents) without fixing the infrastructure underneath.

If your product data is inconsistent, your pricing rules live in spreadsheets, and your promotion logic is hardcoded per campaign, no agent framework will save you. The AI agent is only as good as the data and APIs it can access.

Before any AI strategy pays off, the prerequisite is a commerce platform whose pricing and catalog data is already complete, structured, and accessible through APIs.

Start With the Infrastructure

American Express just launched agent purchase protection for AI-initiated transactions. Google's Universal Commerce Protocol is live with 20+ partners across commerce, payments, and retail. All of these are signs that the protocol layer for agentic commerce is taking shape.

The commerce infrastructure layer underneath is lagging behind. The vendors that get their pricing and catalog data fully machine-readable now will have a real advantage when these protocols reach mainstream adoption. 

A useful test: try making an API call to your own commerce platform to get a fully resolved price, apply a promotion, and complete a checkout. If that requires manual intervention at any step, an AI agent will hit the same wall.

Pricing Strategy Is the New Competitive Moat

In an agentic world, the pricing strategy conversation changes shape. Discounts, tiered rates, contract terms, and promotion logic stop being purely go-to-market decisions and start functioning as the data AI agents use to evaluate you against every competitor in real time. If that logic lives in spreadsheets, sales playbooks, or a rep’s head, it’s invisible to the buyer making the decision. Winning in agentic commerce starts with making your pricing strategy legible to machines, keeping it consistent across channels, and exposing it through APIs that your customers’ agents can call directly.

Getting that pricing infrastructure right is the starting point. Explore how Broadleaf's transaction suite handles pricing, promotions, and ordering through APIs that are ready for the next generation of buyers, human or otherwise.

Pricing Strategy, Re-examined for Agentic Commerce

For most enterprises, pricing strategy was built to support quarterly business reviews, rate cards, and relationship-driven negotiation. Those assumptions quietly break the moment an agent, not a buyer, is the one reading the price.

In an agentic world, your tiers, contract terms, volume breaks, and promotion rules have to function as structured, queryable logic that holds up when an agent compares you against every competitor in the same second. Strategy that only lives in a deck, a spreadsheet, or a sales rep’s judgment doesn’t exist to the buyer anymore, because the buyer never sees it.

Vendors who treat pricing as a machine-readable product of their commerce platform will be the ones agents choose to transact with. Vendors who leave pricing trapped in human workflows will watch agent-driven demand route around them without ever knowing it was there. The $15 trillion projection at the top of this article assumes there’s commerce infrastructure ready to receive it. Whether that describes your pricing strategy is the question worth answering before the next planning cycle.