B2A: Business to Agent — Who Is the Customer?

For 30 years, commerce ran on two categories. Both assumed the buyer was a human. That assumption just broke.

B2C optimised for desire, brand affinity, and the emotional arc from discovery to checkout. B2B optimised for ROI, procurement fit, and workflow integration inside an organisation. Different buyers and different interfaces, but in both cases a person on the other end of every transaction.

A third category is taking shape, and it’s further along than most brands realise. ChatGPT launched Instant Checkout on February 16. Etsy, DoorDash, and Instacart went live immediately; Shopify’s million-plus merchants came next, including Glossier, SKIMS, and Spanx. ChatGPT already processes 50 million shopping queries a day across its 800 million weekly active users. AI-driven shopping orders on Shopify have grown 15 times over since January 2025. The buyer at the other end of many of those transactions isn’t a person making a considered choice. It’s an agent executing one on a person’s behalf.

B2C and B2B were both built around human cognition: attention, emotion, trust formed through copy and imagery, decisions shaped by scarcity cues and social proof. An agent doesn’t have any of that. It evaluates structured constraints. Does this product meet the budget? Does the policy match the principal’s preferences? Can the transaction be reversed if it goes wrong? Where a human buyer discovers through ads and responds to storytelling, an agent discovers through product feeds and schema markup, evaluates through policy compliance, and purchases via a programmatic API call. Agents abort at login walls and abandon at CAPTCHA, because checkout flows designed for human intuition are not programmatic. The whole toolkit of human-facing optimisation, landing pages, urgency signals, loyalty UX, CRO, maps poorly to an agent evaluating structured constraints.

What this creates is a third category: B2A, Business to Agent. B2C and B2B each have their own infrastructure: CRM, CRO, loyalty programmes, landing pages, dashboards, procurement workflows. B2A needs different infrastructure: structured product data, descriptions written to answer agent reasoning queries rather than search keywords, and checkout flows that work without a human in the loop. Amazon blocking ChatGPT crawlers tells you who has most to lose from this shift. Their moat is human-facing discovery. For DTC brands with direct channels, the economics run the other way: the combined take rate for ChatGPT checkout is around 7%, against Amazon’s 25 to 30%.

The AI-in-commerce conversation is mostly about internal efficiency: automating operations and cutting costs. That’s real, but it’s the smaller disruption. The bigger one is who is on the other side of the transaction. Some of your buyers are already autonomous agents, executing purchases on behalf of people who may never visit your site. The commercial infrastructure built for human buyers wasn’t designed for that. The harder update is conceptual, not technical. Who is your customer now?