FRESH TAKES ON BIG IDEAS

There’s a growing narrative that AI agents are on the verge of taking over the consumer experience—autonomously searching, comparing, deciding, and then buying on our behalf. And to be fair, parts of that future are already here. Many shoppers have embraced AI tools that suggest gifts, compare products, or assemble baskets with surprising speed and accuracy. 

But the idea that we’re heading toward a world of fully autonomous shopping—no humans necessary—is misleading. What’s actually emerging is something more nuanced: a gradual and sometimes uneven shift regarding how much control consumers are willing to hand over to AI.

Think of it less as a leap and more as a curve.

At one end of that curve are the kinds of purchases we’re already comfortable outsourcing. Routine, low-stakes, low-regret decisions (such as restocking household essentials, reordering groceries, or replacing everyday items) are natural candidates for automation. When the task is repetitive and the downside is limited, efficiency wins. If an AI agent can handle it reliably, most people are happy to let it run in the background.

Move along the curve, however, and things start to change.

In higher-value and more complex categories—travel planning, consumer electronics, home goods—people are willing to delegate parts of the process but not all of it. An agent might do the heavy lifting: researching options, comparing specs, assembling a shortlist, and even building a purchase-ready basket. But when trade-offs become meaningful (price versus quality, convenience versus flexibility) humans tend to step back in. Not because the technology can’t decide, but because the decision itself matters.

And then there are the categories where delegation largely stops.

Consider a luxury purchase, a milestone item, or anything tied closely to identity. Here, shopping isn’t just about the outcome. It’s part of the experience. Consumers may happily use AI to analyze options or surface alternatives, but the final decision remains firmly human. 

This is the central insight of what we might call the “automation curve” in agentic commerce: Not every shopping experience is meant to be automated at the same rate or to the same degree. 

Two forces tend to shape where consumers land on that curve. The first is trust, which builds over time. As agents prove themselves—making good recommendations, executing reliably, and handling exceptions gracefully—people become more comfortable delegating larger portions of the journey.

“Not every shopping experience is meant to be automated at the same rate or to the same degree.”

The second is context. Willingness to hand over control varies widely depending on factors like price, emotional significance, and the risk of getting it wrong. In some moments, we want speed and certainty. In others, we want deliberation, control, and maybe even a bit of serendipity.

The implication is that the future of commerce won’t be fully autonomous—it will be selectively autonomous. That distinction matters, because it changes how companies should respond. For years, digital commerce has been built around capturing human attention: optimizing search results, refining recommendations, improving conversion. But as AI agents begin to mediate more of the journey, many of those moments collapse. Search, comparison, and consideration increasingly happen in a single, machine-driven interaction. 

In that world, visibility alone is no longer enough. If an agent can’t interpret your product data, evaluate your offering, or transact with your systems, you effectively disappear—no matter how strong your brand might be. 

At the same time, not all value shifts toward automation. In categories where human involvement remains essential, the opportunity is different. It’s less about speeding up the transaction and more about enriching the decision—giving agents the information they need to inform, rather than replace, human judgment.

For retailers, this creates a dual challenge. On one hand, they need to become “agent-ready,” with structured data, clear policies, reliable inventory, and seamless transaction capabilities so AI systems can act on behalf of customers. But at the same time, merchants need to be deliberate about where not to automate. Some of the most valuable moments in commerce—those tied to identity, aspiration, or experience—depend on human engagement. Strip that away, and you risk eroding the very thing customers are willing to pay for.

Success, in other words, won’t be simply about pushing for maximum automation. It will be about understanding where automation creates value and where it doesn’t. Because the future of shopping isn’t about removing humans from the process. It’s about redesigning the balance between human judgment and machine execution—one decision, one category, one moment at a time.

—Edited by Larry Kanter, senior editor, New York

ABOUT THIS AUTHOR

Katharina Schumacher is a partner in McKinsey’s Munich office.