B2B procurement emerges as the key battleground for AI-driven commerce

The biggest names in technology have spent the past two years trying to define the future of shopping, but the most important battleground may not be the consumer checkout page at all.

OpenAI has repeatedly reshaped its approach, moving from direct in-app purchasing to a stronger emphasis on product discovery. According to recent reporting, ChatGPT now offers richer shopping tools, including side-by-side comparisons, image-led search and a broader Agentic Commerce Protocol, wit...

h retailers such as Target, Sephora, Nordstrom, Best Buy and The Home Depot involved. OpenAI has also introduced shopping research tools that ask follow-up questions, scan the web and generate personalised buying guides for categories ranging from electronics to home and garden. The company says the features are aimed at helping people make better decisions, not just faster ones.

That shift reflects a wider truth about agentic commerce: the consumer problem is visible, but not necessarily the easiest place to start. Discovery and checkout are the areas attracting the most attention, yet they are also the parts of retail that brands have spent years refining. When an AI agent steps in, it may remove friction, but it also strips out the carefully engineered nudges, recommendation engines, loyalty prompts and data-capture moments that retailers use to move shoppers towards a purchase.

The results so far have been mixed. Reuters has reported that Walmart saw a steep decline in conversions when agents sat between the shopper and the purchase. Amazon’s “Buy for Me” also drew criticism from retailers that had not opted in. Even where AI can surface products quickly, there is still little evidence that it consistently beats the familiar shopping experience on trust, convenience or value.

That is why B2B procurement may prove more consequential. Corporate buying is rarely a simple matter of browsing, clicking and paying. A mid-sized manufacturer ordering industrial parts may have to work through approved suppliers, contract pricing, compatibility rules, internal sign-offs and purchase-order systems tied to enterprise software. In that environment, much of what a sales representative does is not high-level judgement but the application of rules and institutional knowledge that could, in theory, be encoded and automated.

In B2B, the data is often messy, but it is also more structured in ways that machines can use. Product catalogues may be fragmented across PDFs, legacy ERP exports and sales staff’s heads, yet the underlying logic is usually explicit: SKU relationships, account-specific prices, stock levels, contractual terms and compatibility matrices. That makes the setting better suited to agentic systems than consumer retail, where product data is often aimed at persuading people rather than informing software.

The scale is significant too. The author argues that global B2B commerce is several times larger than B2C by transaction volume, meaning even modest gains in efficiency could have outsized effects across procurement. If agents can shorten sourcing cycles, reduce manual intervention and make complex ordering easier, they could reshape how distributors, manufacturers and buyers work long before consumer shopping is transformed.

For now, the central challenge remains the same in both markets: product data has not been built for machines. AI agents need clear, structured information on availability, pricing, compatibility and buyer-specific terms. If that data is missing or ambiguous, the agent is unlikely to wait around for clarification.

That is why the first meaningful winners in agentic commerce may not be the companies reinventing the consumer basket, but those turning B2B procurement into something software can actually understand. Consumer shopping may eventually follow. But the more immediate prize may lie in the less glamorous, more broken world of enterprise buying.

Source: Noah Wire Services