In agentic commerce, the agent won’t ask—it will judge

Earlier this year, I was in a room with a group of CEOs in Istanbul. A few weeks later, on a call with board members from a European grocery chain. A month later with investors in Australia, then another call with investors from North America. And most recently, an operating team in the U.S. Different markets, different competitive pressures, different stages of AI maturity. Same conversation every time.

They all wanted to talk about AI-powered shopping agents, and competing on the agentic commerce front end.

That focus is understandable. The signals are real. But the executives who are better positioning their organizations are asking a different question: How do I become the retailer an agent selects?

Agentic commerce stands to change how people shop, but also who actually makes the purchase decision. When a shopping agent assembles a basket under constraints of price, availability, loyalty value, and delivery speed, it will make decisions that used to belong to the customer. The agent will select the retailers judged most worthy of the transaction.

Most retail organizations are not ready for that evaluation. And the reason has nothing to do with their technology stack.

EARLY, BUT NOT EARLY ENOUGH TO WAIT

Agentic commerce is real but nascent. Fully autonomous shopping agents are not yet operating at scale. Most deployments today are narrow, assisted, and still maturing. Executives who feel like they have time are right about the current state, but wrong about the window.

The pattern is visible. Instacart and similar services already factor availability, substitution rates, and delivery reliability into how they rank and route retailers. Retailers are currently being evaluated by systems they do not fully see or control. What changes as agentic commerce matures is the scope of that evaluation and the autonomy of the systems conducting it.

The retailers positioned to compete are making foundational decisions right now: how they structure data, govern decisions, and negotiate platform partnerships. Those decisions compound. This is not a call to panic, but to sequence correctly. Your organization will engage with agentic commerce, but will you engage on your own terms or someone else’s?

OPERATIONAL FRICTION SENDS CUSTOMERS ELSEWHERE

Working inside 84.51°, one of the largest retail data science organizations in the country, I got an early read on where agentic AI would create real value. Not on the front end, but in the back office, innovation pipelines, and the unglamorous work of connecting data to decisions at scale.

Food manufacturing is a good example. The opportunity was connecting customer segment data directly to product formulation decisions, simultaneously optimizing tradeoffs across taste, quality, shelf life, margin, and production feasibility. An agentic system can hold all of those constraints at once, simulate scenarios, and surface options a human team would take weeks to evaluate. That is a different operating model.

The connection to agentic commerce runs through operations. Today’s fulfillment platforms already use fill rates, substitution logic, and delivery reliability as ranking signals. As agentic systems grow more capable, the evaluation expands to service commitments, loyalty economics, and how well a retailer handles exceptions. Each of those signals is produced by internal operations. If those operations are fragmented or inconsistently executed, the system routes demand elsewhere. Internal excellence is not the consolation prize. It is the entry requirement.

THE REAL RISK IS CUSTOMER DISINTERMEDIATION

This pattern has a name, and other industries have been through it. The term is customer disintermediation, and retail is about to learn what those industries learned the hard way.

In financial services, disintermediation did not arrive all at once. Algorithmic tools inserted themselves between firms and their clients, replicated core offerings at a fraction of the cost, and made the original relationship feel unnecessary. Firms that moved early protected what automation could not replicate. The ones that waited lost ground that was hard to recover.

The retail parallel is nearly identical. The agentic platform becomes the interface. The retailer becomes the inventory source. If you do not protect the data relationship and surface your full value proposition before scale sets in, you end up competing on price and availability against every other retailer on the same platform. That is a race to the bottom.

Retailers need full visibility into how they are being ranked, why demand routes the way it does, and what signals the platform is reading. The negotiation needs to cover operating rights, not just commercial terms, before dependency makes the conversation academic. Do not publish a catalog. Publish what makes you the best choice.

In an agentic environment, agents choose the retailers that can get the right product at the right price to the customer without friction—the front end. Operational and fulfillment excellence will differentiate the brand—the back end.

EXECUTION DISCIPLINE IS THE SCARCE ASSET

Agentic AI rewards execution. Agentic systems expose what was already broken: fragmented data, unclear decision rights, and accountability gaps nobody wanted to formalize. When a system needs to act autonomously, those gaps become immediate blockers. The first months of deployment are actually about fixing the enterprise.

The retailers making real progress share one characteristic: The AI agenda is owned by business leaders, not the technology team. Operations, merchandising, and supply chain set the priorities. Technology enables. Value is owned by the people accountable for the outcome.

Anchor AI investment in operating problems with real economic weight. Build the internal foundation before you negotiate the external platform deal. Negotiate that deal early and aggressively, with data rights written into the terms before scale creates dependency.

As agentic commerce matures, it will route demand toward the retailers that deserve it by every operational measure a system can evaluate. The retailers treating that as a future problem are already behind on the current one.

Todd James is founder and CEO of Aurora Insights LLC.